aiLL 5/10/23 - The Future is Already Here

Description

Chapters

Transcript

0:03 foreign
0:15 with the AI learning lab going live on a
0:20 what is it a Tuesday night it's a
0:22 Tuesday night
0:23 uh
0:25 what we talk about here
0:27 AI stuff talk about AI stuff why
0:29 wouldn't you talk about AI stuff it's
0:31 not like it's hyped or anything it needs
0:33 a boost so I'm here to boost it uh
0:36 seriously I uh I think that what's
0:39 happening with generative Ai and what's
0:41 coming is way more profound than
0:43 anything we've seen in our lifetimes and
0:46 potentially of all time
0:48 and uh I'm here to answer any questions
0:50 I can my background is in uh acting
0:57 and uh moved to New York City out of
0:59 school and then in the mid 90s I
1:02 discovered this thing that Tim
1:03 berners-lee invented called the world
1:05 wide web and started one of the first
1:07 online magazines and one of the first
1:09 digital agencies company called
1:11 agency.com
1:13 and
1:14 um what's happening with generative AI
1:16 is very very reminiscent of what was
1:20 happening with the early days of the web
1:22 and in fact on my Tick Tock channel two
1:25 of my pinned videos
1:27 um are actually compare
1:29 um you know 2023 being the new 1995.
1:34 um which is which is kind of when I got
1:35 in into all this stuff 94.95
1:38 um
1:40 the the parallels are really interesting
1:42 the
1:44 um
1:46 you know the internet had been around
1:47 for a bunch of decades and
1:50 then this world wide web thing kind of
1:52 made non-technical people give
1:54 non-technical people access to this very
1:57 technical powerful thing and we know
2:00 what happened with that and and chat GPT
2:02 you know there's some other things that
2:04 preceded it but chat CPT was really the
2:06 uh the the kind of seminal
2:09 um the explosion that is doing the exact
2:13 same thing for AI and and we're lucky
2:15 enough to get to play with it and it's
2:17 really quite remarkable so anyway uh if
2:19 you have questions about AI please put
2:20 them in the uh uh whatever you call it
2:23 comments
2:25 I'm old my brain doesn't work anymore
2:28 um the the little image that seems but
2:30 that's also the address of chat GPT so a
2:35 lot of what I talk about here on the
2:37 channel is chat gbt but you know other
2:39 stuff as well but if you want to start
2:42 playing with this stuff if you want to
2:43 start figuring it out you don't need to
2:46 download anything you don't need to uh
2:49 pay for it you just go to that address
2:52 chat.openai.com and start playing I
2:54 would actually encourage you to start
2:55 playing while we chat here
2:57 um so let me let me jump into the
3:00 questions here
3:01 all right one of the questions are you a
3:03 professor I am not a professor I play
3:06 one on Tick Tock
3:07 um now I'm just a um like I'm not even
3:11 sure why I started this Tick Tock
3:12 channel so I had the the big Epiphany I
3:15 had around generative AI that that this
3:18 was the next you know there's sort of
3:20 Fire
3:22 the written language the printing press
3:25 the Industrial Revolution computers the
3:29 Internet really the web
3:32 and I think this is the next one I mean
3:34 there aren't many of these uh
3:37 these kind of changes don't happen that
3:39 often I feel very very fortunate to be
3:42 um
3:43 getting to experience two of these in my
3:45 life
3:46 um
3:48 and I and I just
3:49 um
3:50 I was learning about all this AI stuff I
3:52 was just trying to process like the
3:54 implications of it and what this stuff
3:56 did and what you could do with it and I
3:57 realized like I needed to I needed to
4:01 get these thoughts out of my head and I
4:03 also realized that this is going this
4:06 technology is going to have profound
4:07 effects on people way faster than
4:11 anyone's prepared for and so I thought
4:13 you know I should just share as much as
4:15 I can and help demystify it and help
4:18 share
4:19 um kind of the light side I think it's
4:20 really easy because we have we've had 50
4:22 years of Hollywood telling us that the
4:24 robots are going to kill us
4:26 that I think everyone's default is the
4:28 robots are gonna kill us and and they're
4:32 really really powerful
4:34 um
4:35 there's another very real possibility
4:37 that the robots will free us
4:39 um these tools are remarkable they're
4:42 they're magic so anyway so I'm not a
4:45 professor but I'm really passionate
4:47 about demystifying these tools answering
4:50 what I can brainstorming doing projects
4:53 I keep saying yes to projects in the
4:55 face of Reason
4:57 um because you know why not
5:00 we I mean the reason the the reason why
5:02 I keep saying yes to projects is
5:05 you only get to experience the birth of
5:07 something like this once like it's right
5:10 now
5:11 November 30th 2022 is when this started
5:13 in my opinion that was the day that
5:17 that website launched chat.openai.com
5:20 um and I'm happy to talk about why I
5:22 think that changed everything so we need
5:23 those AI Bots from New York City in
5:25 schools to protect our kids no [ __ ]
5:27 it'd be nice if the AI Bots could
5:32 well
5:34 yeah
5:37 um have have I tried running my software
5:39 through a quantum computer I uh I don't
5:41 have saw well I have my company has
5:43 software it's a company called storybine
5:45 we have an automated video platform that
5:47 I'm adding AI to but we don't do
5:49 anything computationally that would
5:50 require a quantum computer uh I I have
5:53 not invented any of this machine
5:55 learning AI crap I'm just a user of it
5:58 I'm just a fan and a user quantum
6:01 computers I think
6:02 when they when they when they get out of
6:04 the research phase they're very much in
6:06 research phase right now
6:07 when they get out of that phase I have a
6:09 feeling that they're gonna they're gonna
6:10 help with the training of of these Bots
6:13 and things so
6:14 hang on my machine my computer's going
6:17 to sleep I don't have it plugged in hold
6:19 please
6:22 chat GPT wrote me song lyrics today cool
6:24 were they good
6:25 that's awesome
6:27 um what's my background in AI um I
6:28 started [ __ ] around with it
6:30 in about a year ago I was in an nft
6:34 project called pixelmind that is uh it's
6:37 kind of a really cool art project where
6:39 you can do generative images and mint
6:41 nfts out of them
6:43 um I was doing that for about six months
6:45 and then
6:46 um I just started hearing some noise I
6:48 started playing around with open ai's
6:50 playground and gpt3 and started hearing
6:53 about mid journey and
6:55 after nine failed attempts finally got
6:58 stable diffusion up and running on
7:01 Google collab and I haven't done
7:04 anything resembling coding in decades so
7:06 it was just an abject nightmare but I
7:09 finally got it working and I and I uh
7:12 you know the
7:15 installed dream booth and that let me
7:18 you know upload pictures of myself and
7:20 that that led to this project hang on I
7:23 don't have it running here photos
7:27 picture
7:30 oh damn it
7:35 let's see Kyle Shannon Dre I
7:41 there you go a bunch of images that I
7:44 created with dream booth and then I
7:46 started this Instagram Channel called
7:47 Kyle Shannon dreams where I wrote little
7:49 stories of these guys so that's kind of
7:51 how it started
7:52 and then while I was in the middle of
7:54 that project
7:56 um I
7:59 um Chet GPT came out
8:01 and
8:03 you know I did all the sort of normal
8:05 parlor tricks you do with chat gbt like
8:07 you know write me an Eminem rap that
8:09 explains quantum physics then I think
8:11 that was the first one I did and I keep
8:13 I keep saying that one but it you know
8:15 it just did it and you're like holy [ __ ]
8:17 and uh and then the the real Epiphany
8:20 with chat EBT and the and the real
8:21 realization the thing that sort of ran
8:23 me down the rabbit hole was I was
8:25 sitting with my wife one night and I
8:27 just started messing around and had chat
8:29 EBT write me some python code which I
8:32 don't know Python and then I had to
8:33 teach me how to run the python code and
8:35 then I had it debug or you know fix the
8:38 errors for me in the python code and
8:40 within 90 minutes I'd written
8:42 um a fairly sophisticated python
8:44 application that asks you a few
8:46 questions about a topic and then
8:48 generated five pieces of social media
8:49 content for you using openai's API
8:52 within like 90 minutes still haven't
8:54 written a line of python in my life
8:56 and that changed everything and it just
8:59 you know it's been a it's been a I've
9:01 been a manic
9:03 um insufferable piece of AI trash ever
9:06 since so that's my background uh where
9:08 I'm where I'm from I grew up in
9:10 Pennsylvania went to Penn State um got
9:12 my acting degree from Penn State and
9:14 then moved to New York City out of
9:15 school and lived there until
9:19 um 2002 the the agency I the digital
9:21 agency I started in 95 I sold in 2002
9:24 then we moved north of the city and now
9:26 I'm out in Denver
9:28 um
9:30 so if you have questions about AI
9:33 um pop them into the chat and I will do
9:35 my best to answer them
9:38 and AI will eventually destroy us if
9:41 nanotechnology doesn't do it first I
9:43 don't know that nanotechnology I I
9:45 assume you mean RM our mRNA stuff but um
9:51 yeah I mean it's easy enough to take the
9:54 the robots will kill us view I mean
9:56 Hollywood's trained as well I I think
9:58 it's I honestly think it's well
10:03 I think that's the I think it's an easy
10:05 place to go
10:07 um where I'm trying to go is what if
10:09 this technology actually did good for us
10:11 and the more I learn about it the more I
10:14 realized that
10:15 these are tools that we get the chance
10:17 to use and you can do really amazing
10:19 things with them and and I've found that
10:22 um I operate in a completely different
10:24 creative plane
10:26 um an empathy plane when I'm when I'm
10:28 using these tools because what the AI
10:30 tools do is they they make the shitty
10:35 repetitive low-level work the treasury
10:38 work of like everything you do they like
10:41 do that automatically in three seconds
10:43 and it kind of lets you live at a a
10:47 higher plane
10:48 and the other thing so here's an
10:51 interesting thing I think one of the
10:52 real possibilities here is that AI is
10:54 going to allow us to reconnect as humans
10:57 I think we are so polarized in this
10:58 country we're so far from that right now
11:00 social media has [ __ ] destroyed
11:03 the social fabric
11:05 um so so here here's a kind of cool
11:08 thing that happened I was I was on a
11:11 panel for a group of
11:13 um
11:14 like like Foundation people non-profit
11:17 people right and I was explaining what
11:19 cheept was and how it worked and and you
11:22 know just kind of waxing poetic about
11:23 this stuff and this one woman goes well
11:24 wait a minute
11:25 so if these tools can do that wouldn't
11:27 it mean that someone writing their Grant
11:30 proposal could write a grant proposal
11:32 much more quickly I'm like yeah and then
11:34 she goes and then by extension couldn't
11:37 the people reading the grant proposals
11:39 read them and process them much more
11:40 quickly I said yeah
11:43 and then I said and what's left and she
11:46 goes well talking to the people and I'm
11:49 like yes right and so so you can imagine
11:51 a world where all of the complex
11:54 low-level [ __ ] stuff gets handled
11:56 and then we can connect like I could see
12:01 um actual customer service coming back
12:03 as as a as a competitive Advantage for
12:06 companies because all the low-level
12:08 customer service stuff lives down here
12:10 and if you actually have a problem where
12:11 you really need to talk to someone maybe
12:13 instead of them you know doing all the
12:16 form entry on the complex [ __ ] now
12:19 that gets done automatically and those
12:20 people can be freed up so rather than
12:22 those people losing their jobs maybe
12:24 they get up trained to be good customer
12:26 service people so anyway you can go to
12:29 the dark place this channel I'm trying
12:32 to stay in the light Place how do I show
12:34 the little five screen image in the
12:36 corner this this little the
12:38 chat.openai.com
12:41 when you go live there's an option for
12:43 dual cameras and so the back camera of
12:46 the phone is looking at my laptop so I
12:48 have my laptop sitting on two shoe boxes
12:51 here I'll show you
12:54 so so there's my laptop sitting on two
12:57 shoe boxes
12:59 um and I'm sitting in front of that
13:00 thing
13:01 so it's just two cameras live at the
13:04 same time that's what's going on
13:06 um
13:10 any experience oh damn it this Tick Tock
13:13 thing
13:13 any experience with rocm no I need to
13:17 choose a Linux distro and for a rig with
13:20 AMD gpus way above my pay grade way
13:23 above my knowledge grade I am more a
13:26 product guy than an engineer by far and
13:29 I also have ADHD so I
13:32 Engineer it being an engineer being a
13:35 software engineer I feel like what
13:37 software Engineers do is a miracle
13:39 because my brain just cannot handle that
13:41 level of of sameness and consistency and
13:45 attention to detail over time
13:49 um have I worked on physics using AI not
13:52 other than saying explain to me how
13:54 quantum physics Works in an m m rap if
13:57 that qualifies as working in physics
13:59 then yes I have
14:01 um it changed my life as well now I'm
14:02 coding and I'm a PA native cool hey PA
14:05 native I grew up in York Pennsylvania
14:07 the home of York barbell your Peppermint
14:10 Patty
14:11 Caterpillar tractors BorgWarner tanks
14:14 huh a lot of the big tanks that blow
14:16 [ __ ] up that came from York PA
14:21 um
14:22 oh I'm also an idiot
14:25 I'll tell you that
14:27 um
14:28 let's see it allows you to translate
14:30 your mind to a presentation yeah so so
14:33 that's it's a good point there's if you
14:35 want there's a cool
14:37 um piece of technology out there
14:40 called Tome t-o-m-e dot AI
14:45 I think it's dot AI
14:47 all these companies these days no it's
14:49 not oh God damn it sorry hold please
14:52 tome
14:56 tome.app
14:58 tome.app see I grew up in a time when
15:01 um
15:03 where's tim.app why is it not working
15:06 hold please
15:08 uh looks like my internet might be down
15:10 can you still hear me
15:12 tap the screen if I'm still live oh no
15:14 here it comes okay it's just slow
15:17 um so tome.com is
15:20 can I create a new one if I say create
15:24 um
15:26 a presentation idea like
15:29 um
15:30 a business about
15:34 walking dogs
15:38 um on roller skates
15:45 and then I hit return
15:48 and it will start writing that
15:50 presentation so so wait let's go back to
15:53 the the title I guess the uh the value
15:55 proposition of this business
15:58 oh wait I can't go to it till it writes
15:59 it but it's basically it's writing a
16:01 presentation right now and it's
16:05 um generating images
16:07 so it says roll so the business will be
16:10 called rolling with the pack which is
16:12 actually pretty good
16:13 the business of dog walking on roller
16:16 skates
16:17 and then it gives me a little outline
16:19 sorry about the focus these phones suck
16:22 um introduction benefits of walking dogs
16:25 on roller skates challenges flying talks
16:27 unrelated skates equipment needed
16:30 marketing strategies conclusion and then
16:33 it gives you you know it writes you a
16:35 presentation and it picked a cool
16:36 cartoon style and made the images
16:39 so it actually got it right that the
16:41 guys on the roller skate not the dogs
16:43 that's good
16:44 and uh yeah so so you know this is just
16:49 one tool and it's you know is this
16:52 actually practical
16:54 it's not far from it it really isn't far
16:57 from it I mean this stuff's pretty
16:59 remarkable and what I find with with
17:01 chatgpt by the way the reason I have
17:04 this up on screen
17:07 um this sort of weird picture of me with
17:09 chat.openai.com on it is that um
17:14 that's the URL for chat GPT if you're
17:16 new to AI if you
17:20 um
17:21 are trying to figure out where to start
17:22 start there
17:24 start with chat GPT understand what it
17:26 does you just ask it [ __ ] and it tells
17:28 you [ __ ]
17:29 um but it does it in incredibly
17:31 remarkable ways and into this person's
17:33 comment
17:34 um who's the engineering uh
17:37 indigenously a little cool name
17:40 engineering indigenously you know said
17:43 you know you have an idea you've got
17:45 this thing in your mind and what this
17:47 generative AI stuff does is very very
17:50 quickly
17:51 um gets that Vapor of an idea and
17:54 realizes it into something tangible
17:56 um so that Tome thing that I just showed
17:58 you that tone presentation tool
18:00 um
18:01 is is just kind of miraculous and that's
18:03 like a version 1.0 like all this
18:05 software is really janky and raw and
18:08 weird right now but you know imagine it
18:11 in a year when it's actually
18:11 sophisticated and you can pick from you
18:14 know seven different story Frameworks
18:16 and you know I'm I want to pitch this
18:18 business to VCS versus I want to pitch
18:20 this business to prospective customers
18:22 and it it'll create completely different
18:24 things and it'll learn your writing
18:26 style and and and just do it so
18:30 I'm so nice to my AI so I find that as
18:33 well if you haven't played with this
18:34 stuff
18:35 um chatgpt has been programmed in such a
18:38 way that it's it's pretty nice to you it
18:40 wants to give you what you want and it's
18:42 very polite and and
18:44 prosie it says Pros a lot you know and
18:48 um I find myself thanking it and saying
18:50 please and like when it does something
18:53 I'm like perfect now let's try to make
18:57 it better like this
18:59 um and it's not so much that I'm Anthony
19:01 anthropomorphizing it it's that
19:04 it just seems to work better when you
19:07 interact with it when you give it more
19:09 context so here here's something that I
19:11 feel and this is a big change that I've
19:13 had to I've had to kind of
19:15 learn through
19:17 um Google for however long it's been
19:19 Google 15 20 years
19:22 has trained us like so when you first
19:24 started to use Google you had to use
19:26 Boolean logic you had to put parentheses
19:28 around things and you had to do pluses
19:30 and minuses and I think it still
19:31 understands all that
19:32 but but they kind of trained us over the
19:35 past 15 or 20 years to write in this
19:37 really cryptic kind of non-english kind
19:40 of way it's English words well or
19:41 whatever your language is but but but
19:43 it's not kind of sentences right it's
19:46 more like you know what are the bare
19:48 minimum words I need to put in here so
19:50 that Google can give me things that are
19:52 relevant and what happens with Chachi BT
19:55 is
19:56 it can do stuff with sparse words but it
19:59 tends to give you sparse generic answers
20:01 and when you give it more context and
20:03 you give it more color it tends to
20:06 answer like that so it kind of interacts
20:08 with you as you interact with it and the
20:11 harder you push it and the the I was
20:14 going to say the more you respect it but
20:16 you know in some sense that the more you
20:18 treat it like you were talking to
20:19 someone with respect it does really good
20:21 if students are using it then teachers
20:23 should be using it too absolutely T well
20:26 teachers you know could be using it to
20:28 you know grade the papers that catch EBT
20:30 has written on the student's behalf
20:32 um but but I think that what the what
20:34 the teachers and professors in higher ed
20:36 should be doing is absolutely
20:38 incorporating it into the classroom and
20:41 and shifting to
20:43 um more critical thinking more
20:45 engagement right not teaching um to
20:48 standardized tests not
20:50 um teaching to
20:52 um dump information
20:54 um to to verify if that information was
20:57 memorized right
20:59 um because we now live in a knowledge on
21:02 demand world what chat GPT is what these
21:04 large language models are is knowledge
21:07 on demand
21:08 if you want it to read the tax code and
21:11 do your taxes for you it can do that
21:13 right so you don't need to go to
21:15 accounting school to do that
21:18 um
21:19 so so the implications for Education are
21:22 are profound profound profound uh and
21:25 and you know all of this kind of I
21:27 consider it short-term stuff like we're
21:29 gonna we're gonna you know run all these
21:31 papers through tests to see if they were
21:32 written by Chach EBT well when chatgpt
21:35 and large language models are ubiquitous
21:38 and they're in all of The Office
21:39 Products all of the Google products all
21:41 the note taking products all the
21:43 screenwriting software products all the
21:45 all the writing products
21:47 um
21:48 how will you know like like
21:51 when you use chat GPT you you don't
21:56 you don't just give it a query and it
21:58 gives you an answer and you copy and
21:59 paste that [ __ ] and publish it you go
22:01 back and forth with it you're like hey I
22:03 have an idea for a thing can you write
22:05 me an outline of that thing and it'll
22:07 give you an outline and you're like no
22:08 that sucked can you make the outline
22:10 suck less by putting in more details in
22:14 this you know about this topic and then
22:15 it'll get better and you're like oh
22:16 that's good and then you take one of the
22:19 little sections and you kind of blow
22:20 that out and outline that and then you
22:21 write some stuff and then you put those
22:24 things together like you know you go
22:26 back and forth with it just like you
22:27 were going back and forth with a with an
22:29 editor or a writer
22:31 um so so
22:33 you know I think it's going to be very
22:35 very short-lived that we think of these
22:37 things like somehow separate from the
22:38 work we do it's just going to be rolled
22:40 into all the work we do
22:42 um
22:44 I wrote I wrote a full Grant request
22:46 using chat GPT in 30 minutes there you
22:49 go 30 minutes do you know how long
22:52 writing grants takes
22:55 it's a freaking nightmare I can't do it
22:58 like I could not even begin to consider
23:00 like I couldn't even learn how to write
23:02 a grant proposal much less write one
23:05 and this person wrote One in 30 minutes
23:08 I mean this is it I don't know it's
23:11 miraculous stuff uh anyway
23:13 what should I say here if you're new to
23:16 AI um go to the URL right there I'm not
23:18 affiliated with that in any way that's
23:20 just the official chat GPT website a lot
23:23 of people like go to the app store or go
23:25 to Google Play and search for chat EVT
23:27 all those apps are just rip-offs trying
23:30 to get your money from you just go there
23:32 and it's free and it's remarkable if you
23:34 haven't used it just well you hear me
23:36 ramble with gas apparently uh
23:41 um just start playing with it and you'll
23:43 see
23:44 you'll see seriously tell it to write
23:46 your poem tell it to write your song
23:48 describe your significant other and tell
23:51 it to write something for them
23:53 watch
23:54 and then go really
23:57 I wonder if it could do and the answer
23:59 is yes now here's the thing to know
24:02 about chat EVT if you're new
24:04 um it hallucinates
24:05 uh I heard chechi PT described once as
24:09 mansplaining as a service which is the
24:12 best description of chat GPT I've ever
24:14 heard it will confidently answer you
24:17 even if it's wrong uh so if you need
24:19 accuracy make sure you check your uh
24:21 check your answers uh the other thing is
24:24 if uh if you want to use gpt4 you can
24:28 use Bing bing.com bing.com is
24:32 Microsoft's search engine it now has
24:34 chat in it that chat is based on gpt4
24:37 and it's also connected to the internet
24:39 so it will give you citations so if you
24:41 need something that's a bit more
24:43 um predictable and accurate use Bing and
24:46 there's three different settings there's
24:48 less create like more accurate sort of
24:51 mid creative and then High creative or
24:54 something like that there's three
24:55 different levels that you can set it at
24:58 we need to reboot Star Trek TNG and
25:01 watch the episodes yeah I well all of
25:04 the uh all of the uh the Sci-Fi stuff I
25:08 just I mean holy crap it's like all
25:10 coming true time travel's not but and
25:12 the uh the beam zapper thing is not but
25:16 all the other stuff is
25:18 um if you have questions for me about
25:21 um
25:23 AI put them in there
25:26 um if you think this is interesting if
25:28 you know people that are interested in
25:29 AI feel free to share this live get get
25:31 the word out
25:33 um you can subscribe did I say that I
25:35 don't know I'm old thanks for the rose
25:38 um let me see what other questions we
25:40 got here
25:42 uh uh that's interesting I'll come back
25:44 to that one
25:57 so sorry I so if you've watched my lives
26:01 before you know this is a problem of
26:03 mine where when I'm scrolling on the
26:05 comments it'll jump me to the bottom and
26:07 I'm trying to scroll back up to where I
26:08 was so I don't miss anyone but then I
26:10 just sing stupid songs
26:13 all right I'm not gonna go all the way
26:15 back up all right oh no I got to the
26:17 Star Trek one cool okay so what's the
26:19 next one it looks
26:21 if it looks stupid and it works it's not
26:23 stupid cool
26:26 um
26:26 love it Pittsburgh girl here hey
26:28 Pittsburgh it's happening I'm a nice guy
26:31 and a good presenter thank you
26:32 appreciate that Namaste Namaste to you
26:36 I build templates for excel in word but
26:39 not since November I assume meaning
26:40 because
26:42 um Chachi vt's doing that [ __ ] for you
26:44 yeah
26:45 actually if you haven't played with it's
26:47 gonna It's Gonna Roll into it natively
26:49 soon but there's a there's a plug-in I
26:51 think it's called GPT for Excel or or
26:53 something like or for GPT for Sheets
26:57 and you install
26:59 chat GPT into the cells of an Excel
27:02 spreadsheet and it's insane what you can
27:04 do you can you can put one word in one
27:07 cell and then have it generate 20 words
27:09 in this column and then for each of
27:11 those words generate you know an outline
27:15 and then for each of those outlines
27:16 generate a YouTube script and I mean
27:19 it's just insane there's also a thing
27:22 came out
27:23 um
27:25 four days ago it didn't come out
27:28 um some people have access to it I don't
27:30 because open AI hates me they don't but
27:34 I'm I'm cranky they came out with a
27:36 thing called code interpreter so when
27:39 you go to chatgpt
27:42 openai.com
27:46 when you go here there's a little pull
27:49 down menu and and there's like oh they
27:52 got rid of the uh
27:54 they got rid of the old GPT 3.5 so I've
27:57 got I do have access to plugins because
27:59 I'm special and then there's gpt4 and
28:02 then there's GPT 3.5 well there's this
28:04 gonna be this new thing called code
28:06 interpreter
28:08 and what code interpreter does is it
28:10 puts a little up upload button next to
28:12 where you type in the stuff and you can
28:14 upload up to a hundred megabyte data
28:18 file like an Excel file
28:20 and code interpreter will just interpret
28:23 all the [ __ ] that's in that Excel file
28:26 and then you can start talking to it
28:28 like it's a full-fledged data analyst
28:31 like you can ask it questions like
28:34 what's interesting about this data and
28:36 it will tell you it will have read all
28:38 the data it doesn't matter if the data
28:40 is janky and uncleaned up it will clean
28:42 it up
28:44 um and then you can start asking it
28:46 things like well give make me a chart
28:48 you know like a out of whatever scatter
28:51 chart comparing these two Trends in the
28:53 data
28:53 and it will write the python code to
28:56 generate the chart generate it and then
28:58 hand it back to you I mean
29:01 it's it's just we uh yeah we're just
29:05 this is we're living in in miraculous
29:07 times have I heard of human Hugh
29:10 Humana dot AI I I assume you mean that
29:14 because you've got hu dot m-a dot n e a
29:17 i
29:18 humata dot AI lets you upload a PDF and
29:22 do something with it I don't know what
29:24 humane.ai is is this going to be a
29:27 mistake for me to type this in h I'm
29:30 sure it is I'm sure I'm about to install
29:33 some ransomware [ __ ] let's see
29:37 um
29:39 Emmy no Huma
29:42 Huma I'm gonna do it watch I'm gonna get
29:46 you'll send you'll send
29:48 Health right yeah
29:51 nah it didn't go there okay good
29:54 um I don't know what you're talking
29:55 about but humada.ai is pretty
29:57 interesting
29:58 oh no when I scroll
30:01 you I must stop
30:03 oh I was here over an hour the other
30:05 night you know the like the last two
30:07 lives that I've done it's been like
30:09 three hours what am I [ __ ] doing with
30:12 my wife I'm sorry I'm sorry you stopped
30:15 again you shouldn't stay here you should
30:17 go you should go it doesn't get better
30:20 than this is it this is it it doesn't
30:22 get better than this okay I I'm like I
30:25 used it at work today it was wicked
30:27 close to the shut my man hey thank you
30:31 thank you for the glasses that was
30:32 lovely thank you for the rose Namaste
30:35 um used it at work today it was wicked
30:37 close to the shut my manager oh the [ __ ]
30:40 my manager preaches I think he's a bot
30:43 yeah yeah I know
30:47 um
30:48 I just got a text someone named Zachary
30:50 is conducting a survey how exciting is
30:53 that oh that's the other thing have you
30:55 noticed
30:56 um thanks to chat GPT how to make
30:59 millions with chat gbg
31:02 um have you noticed that spam emails
31:04 suck less now and you actually have to
31:06 read them to figure out that they're
31:07 spam that kind of sucks
31:11 so so it's not gonna be all good
31:15 um I picked up some more domains cool um
31:18 GPT dev.me nice AI Dev dot expert GPT
31:23 dev.tech what do you think nice I like
31:25 it I think domains are good I mean we're
31:27 in this world weird sort of thing where
31:30 um
31:32 there are so many different sub domains
31:36 now that it's really hard to remember
31:37 like all of the AI
31:39 companies are DOT AI dot IO dot business
31:43 dot like just all that stuff and it's
31:45 it's it's frustrating but it's also good
31:49 like it's good to have you know
31:52 good solid domain so I think it's good
31:54 garbage in garbage out absolutely if you
31:56 give chat apt crappy prompts it it
31:59 um it obliges and gives you crappy
32:02 content
32:03 which is funny you know you know chat
32:05 EBT is one of those things where it's
32:06 really easy to
32:08 um
32:10 to create Sensational [ __ ] like the dude
32:12 from The New York Times that got you
32:15 know
32:15 Bing to tell him it loved him and to
32:17 leave his wife
32:19 um
32:20 you know it's relatively easy to do that
32:23 [ __ ] it's also relatively easy to have
32:25 it give you a shitty answer or an
32:26 inaccurate answer and go see it doesn't
32:28 work
32:29 so I would if I were you
32:32 not that you're asking my opinion but
32:34 you did stop
32:36 um I would avoid the
32:38 um dismissive and cynical view of AI
32:42 because this shit's coming fast it's
32:44 very very powerful it's really exciting
32:45 when you get into it but if you sit on
32:47 the outside of it it's really easy to
32:50 find evidence that it sucks or that it's
32:53 evil or whatever and it's very easy to
32:55 just ignore it I would strongly
32:58 encourage you not to ignore it like I
33:00 don't mind people that have an opinion
33:02 AI that actually use it but almost
33:04 everyone I know that's got a really
33:05 strong opinion on AI especially the
33:08 negative stuff one they don't understand
33:10 how it works and two they don't use it
33:12 so have a strong opinion but do it from
33:14 a place of of knowledge
33:16 I love my AI but I don't love my AI yeah
33:20 I know I think I think we'll get there I
33:23 mean I think I think these things are
33:24 gonna get
33:26 I think they're gonna get much more
33:27 personal
33:29 um I I have a sneaking suspicion that
33:31 apple is going to
33:33 um
33:34 they're gonna be really good at the
33:36 personal agent game so so you know like
33:38 Apple's got the the dumb little emojis
33:40 memojis or whatever they're called I
33:43 mean imagine those memojis where you can
33:45 have one that kind of looks like you and
33:47 it's you know it's a fun version of you
33:49 and then that agent just goes and does
33:53 [ __ ] for you it you know because because
33:56 it's Apple they've got all your data and
33:57 it's very you know secure they're
33:59 they're good with that
34:00 and you can just say go book me flights
34:03 go make me reservations so imagine
34:05 having an assistant that actually didn't
34:07 suck and so I think they are gonna
34:09 actually get I don't know that we'll get
34:10 to I love my AI
34:13 um but I think we're gonna get pretty
34:14 attached to what these things do for us
34:15 have I used GPT agent
34:18 um before they removed internet
34:20 connection I had it checking apis I
34:22 haven't used GPT agent I've used agent
34:25 agent GPT which I assume that's what
34:27 you're talking about
34:29 which so if you don't know what that is
34:30 there's this new thing called autonomous
34:32 agents
34:34 um which are you string a series of
34:35 tools together programmatically and then
34:38 you basically just give it a goal so you
34:40 just say go find the annual revenue and
34:45 the marketing tag lines for my top 10
34:47 competitors and it will write its own
34:49 prompts it will write itself a task list
34:52 using chat GPT and then it will go use
34:55 tools like go search Google go search
34:57 Bloomberg for Revenue numbers bring
35:00 those answers back and then it'll give
35:02 it a new set of
35:04 tasks and and it'll just do [ __ ] for you
35:06 so agent GPT and that's at
35:11 let me see agent gpt.com I think it is
35:15 is it.com it's probably not
35:17 nope God damn it agent
35:22 agent
35:24 GPT
35:26 no
35:27 agent G and
35:30 GPT
35:35 oh that's right so it's agent GPT dot
35:38 reworked without the Ed so it's just
35:40 rework D reworked dot AI
35:44 and this
35:46 a web-based version of Auto GPT so you
35:49 just sort of give it a name here here
35:51 I'll show it to you so this is storybine
35:55 um agent that's my company storyline
35:58 um find me
36:00 tag lines of
36:03 top 10
36:06 storybine competitors
36:11 um these things are really really
36:13 new and they're really really janky but
36:16 anyway it says thinking added task
36:18 search for the top 10 competitors added
36:20 tasks compile a list of taglines added
36:23 tasks confirm that the taglines are
36:25 recent and up to date so it's giving
36:28 itself the actual things to do and so
36:31 now it's off doing one of them or it's
36:35 hung this this thing hangs a lot in Auto
36:37 GPT hangs a lot but kind of look past
36:40 whether or not it works right now but
36:42 this concept is really really
36:44 fascinating and and I think quite
36:46 powerful oh look what did it do it did
36:49 something hold please
36:50 it's exciting it actually did something
36:53 okay
36:55 um
36:57 we have a list of competitors
37:00 import necessary libraries Define search
37:03 so so it's created some some code to to
37:07 generate some searches
37:10 and then and I don't know what else it's
37:14 doing
37:17 has it done more things oh it did more
37:19 things
37:21 um
37:23 oh it's actually writing code to give me
37:25 some sort of output anyway we'll come
37:26 back to this we'll see if it I'll just
37:28 let it keep running it's just off doing
37:29 its thing like that's that's the thing
37:31 about these things is they just kind of
37:33 go off and do their thing
37:35 um if you it looks like we don't have
37:36 many new people here but but if you
37:38 haven't tried chat GPT that URL in the
37:41 little box right here is uh Chachi PT's
37:44 official thing I'm not affiliated with
37:46 it it's just that's where I think you
37:48 should start if you're trying to figure
37:49 this [ __ ] out
37:51 um you need to be on hard fork okay
37:53 what's hard Fork let's go find out did
37:56 this do anything do I have my list of
37:58 competitors yet
38:00 no it's still writing codes writing
38:02 python code
38:04 showing python code for beautiful soup
38:06 that's so cool all right uh wait
38:10 tea what is it hard Fork
38:13 hard Fork
38:16 well it's hard for card for the New York
38:18 Times you mean the New York Times thing
38:21 Investopedia hard Fork what is a
38:23 blockchain and what happens
38:25 well I don't know what it is but tell me
38:27 what it is I'll check I'll check it out
38:29 if I can figure it out
38:32 um I wrote I wrote lesson plans in chat
38:34 GPT and it was great yeah good go
38:36 teachers teachers that are
38:39 um embracing this stuff I'm all for it
38:41 I'm all for it share this live if you uh
38:43 think people might uh be interested in
38:46 AI I'm here to answer questions people
38:48 should be mono should be monetizing AI
38:51 before it's a paid service it's not
38:52 going to be a paid service well I mean
38:53 they're going to be lots of paid
38:54 Services
38:56 um but the the horse
38:59 um hath been let out of doth Barn
39:03 um chat GPT kind of blew the entire
39:07 industry wide open so so what happened
39:10 was
39:11 um Google invented this [ __ ] Google
39:13 invented the transformer in 2017 which
39:16 is which is which was the sort of key
39:18 first it was neural networks and and
39:21 some of the machine line learning things
39:23 we're kind of the first breakthrough and
39:26 then 2017 there's this paper called the
39:28 tension is all you need written by
39:30 Google
39:31 they invent this thing called the
39:32 Transformer that's the thing that's the
39:34 foundation of everything we're using
39:36 today Chachi PT barred all of it the
39:39 image generation things the music
39:41 generation things it's all because of
39:42 the Transformer
39:44 um so they invent this thing and I think
39:46 they this is my instinct like um this is
39:49 all speculation
39:50 I assume they realized what they had on
39:52 their hands was so powerful that it
39:54 would completely obliterate their
39:55 business model because why would you
39:58 search for something and and sift
40:01 through 40 different responses when you
40:03 could just get an answer right so if I
40:05 were Google sitting there with 92
40:07 percent market share and I invented a
40:09 thing that made search irrelevant I put
40:12 it in the top left desk drawer too
40:15 so they did and but they they released
40:18 this Transformer thing and then you know
40:20 the AI Community machine Learning
40:22 Community kind of adopted it and
40:24 mid-journey was born and stable
40:25 diffusion was born and gpt1 gpt2 gpt3
40:28 were born
40:30 and then Microsoft said hey this GPT
40:34 thing is really interesting maybe we
40:35 should invest a few pennies in this in
40:38 their first pile of pennies
40:40 um was it was a billion dollars and so
40:42 about a month before the they were going
40:45 to make the announcement they that they
40:46 invested a billion dollars in open AI
40:49 Sam Altman the CEO of openai said hey um
40:52 engineering team build chat gbt and you
40:55 got two weeks to do it
40:57 so they built Chachi PT in two weeks now
41:00 here's here's a funny this is a funny
41:02 little piece of trivia
41:04 um in the documentation for gpt3
41:08 are the instructions for how to build
41:10 chat gbt and they were out for the world
41:13 for
41:15 a year and three months some long amount
41:17 of time Sam Altman assumed that someone
41:19 would have built chat CPT they didn't
41:21 they didn't really want to build it
41:23 but no one built it so he said you have
41:25 two weeks to build it so they basically
41:27 just you know built chat CPT and
41:29 launched it and in six weeks it went to
41:31 a hundred million users to put that in
41:34 perspective the World Wide Web took
41:37 eight years to get to 100 million users
41:39 when when it first came out hundred
41:41 eight years to get to 100 million six
41:42 weeks for cheche EBT so because it
41:45 exploded under the scene so dramatically
41:47 it forced everyone Out of the Shadows
41:50 Google
41:51 um AWS now has one meta like all the
41:55 meta stuff they're now open sourcing all
41:56 that stuff everyone's like we've got AI
41:58 too
42:00 um so so because of that
42:02 there is now a bunch of Open Source so
42:06 completely open unrestricted large
42:09 language models out there so even if
42:12 open AI fails as a business even if
42:15 Google doesn't give us access to all
42:18 these tools or charges for them which
42:20 they're not going to because
42:23 they don't need to they've they've got
42:26 all these other revenue streams right
42:28 they can they can incorporate these
42:29 things into into their other business
42:32 um
42:33 so so there's going to be things you can
42:34 pay for but you're going to have plenty
42:36 of access to these large language models
42:38 for free
42:40 um I think where this very very quickly
42:41 goes is these will be installed on your
42:43 phone I think they may even be burnt
42:45 into chips that are installed on devices
42:47 so imagine a device that you carry with
42:50 you out on a hike where you've got no
42:52 self-service no internet no nothing and
42:55 you've got you know all of the knowledge
42:57 of humanity at your fingertips
42:59 like probably within a year
43:04 you know it's not too profound except
43:06 all of the knowledge of humanity at your
43:08 fingertips
43:09 like like what the internet is right now
43:12 is kind of all of the information of
43:14 humanity at your fingertips but you have
43:16 to do the knowledge piece
43:19 now you don't it just does that for you
43:22 it's remarkable students copy and paste
43:25 that [ __ ] well some do but teachers good
43:28 teachers don't let them do that because
43:30 they bring chat GPT right into the
43:32 classroom and increase the use of
43:35 critical thinking and engagement
43:37 you know
43:39 like teaching I'm an instruction
43:41 designer in higher ed helping faculty
43:43 start implementing into into assessments
43:45 that's super cool
43:46 I use AI for music it's an exciting time
43:49 for Independent Artists I totally agree
43:51 and again I think you know
43:53 there's going to be lots and lots of
43:56 horrible [ __ ] that's just generated
43:58 really quickly and put out there but
44:00 everyone that I know that that's a
44:02 serious artist in any way in any
44:03 discipline they're using these tools
44:06 iteratively they're using it as like a
44:08 brainstorming tool and then they'll go
44:10 do their thing on it for a while and
44:11 then they get stuck and they're like hey
44:12 let me take this stuck piece let me put
44:14 it back in the tool and generate some
44:16 new ideas oh that's good and then they
44:17 go back so it's this back and forth
44:19 thing so
44:20 yeah and like I also feel like like with
44:23 the if you're a musician if you're
44:24 really good at the music thing but maybe
44:26 you're not a great at video well now
44:28 you're going to be able to be great at
44:29 video and you're going to be able to be
44:31 great at writing and all the other
44:32 disciplines that are required to take a
44:35 piece of music and put it out in the
44:36 world
44:38 you could do that you know on a weekend
44:41 by yourself or with a bud
44:44 um without having to spend 10 years
44:46 getting good enough at animation to make
44:48 an animated music video like that's the
44:50 thing that's just friggin exciting to me
44:53 gamma gamma dot app is a good one okay
44:57 I'll fall for it
45:02 sorry you've been infected with
45:03 ransomware
45:05 what is this this is a
45:08 gamma.app oh no this is
45:10 gamma.netherlands I'm like this is this
45:12 isn't even in English
45:14 um gamma
45:18 dot app
45:24 new medium for presenting ideas
45:27 oh this is a deck builder oh cool all
45:29 right I'll play with this one yeah I
45:31 only know Tim I'll check gamma out so so
45:34 someone earlier on was talking about you
45:36 can just have kind of a vapor of an idea
45:37 and then you have a presentation these
45:39 these tools are amazing
45:40 um you just give it a vapor of an idea
45:42 and it writes it writes you a
45:44 presentation complete with images
45:47 um
45:49 it says sign up is unavailable for what
45:53 for for chat GPT is that true
45:57 um let me know if that's true because I
45:59 won't keep pointing people there so if
46:01 chat EPT sign up is unavailable go to
46:04 poe.com poe.com or go to bing.com
46:08 bing.com the chat button
46:11 um the chat functionality at bing.com is
46:13 gpt4 connected to the internet so it's a
46:16 little neutered because of some issues
46:19 they had
46:20 um but but you can go use that but I I
46:22 am still a strong firm believer that
46:25 chat gbt for now is is better Google
46:28 just had their um
46:31 Google I O
46:33 live stream today
46:35 um so their developers conference they
46:36 announced a lot of stuff like they've
46:40 got
46:41 um this thing called palm2 which is
46:42 basically the foundational model that's
46:45 the competitor to gpt4
46:48 um they're gonna have ai in everything
46:51 um
46:51 you know I I
46:54 don't have a ton of faith in Google but
46:56 I wouldn't count them out at all they
46:58 invented this [ __ ] so I I do have a
47:00 feeling they've got a lot of stuff
47:01 sitting in the desk drawers they can
47:03 just sort of whip out polish up and put
47:05 out in the world and and really compete
47:07 with with open AI
47:10 um they wrote me a contract for a
47:12 wedding band proposal brilliant yeah
47:14 doing stuff like that like like it's
47:17 just like like writing it writing a a
47:20 you know an engagement contract like
47:22 that
47:23 um or or reading them like read this
47:25 legal document or or even just like a
47:27 copy and paste the Clauses and say
47:30 rewrite this legal clause
47:32 you know as if I was a seventh grader
47:34 that doesn't understand a lot you know
47:36 like like Use Jelly Bean metaphors and
47:38 it will rewrite a legal contract and
47:40 explain it with jelly beans it's it's
47:43 remarkable someone enjoyed mansplaining
47:46 as a service thank you it's I didn't
47:47 come up with it but God damn it it's it
47:50 is very accurate
47:51 for what what chat GPT is
47:55 um
47:57 I saw a thing about Jeffrey Hinton there
47:58 I'll get back to that
48:03 have I told you I'm a singer because I'm
48:06 not
48:08 um new here appreciate the thoughts well
48:11 done enjoying your life thank you Tomo I
48:13 appreciate that Namaste thank you for
48:15 the rose
48:17 um
48:21 I'm trying I'm sorry I'm going back I'm
48:23 trying to find my my I'm trying I'm
48:26 trying to find my way people I'm trying
48:28 to find my way and I'm thirsty
48:33 oh we're supposed to get we have tonight
48:36 um
48:37 tornado and and hail possibilities so my
48:42 car right now is covered in moving
48:43 blankets with big rocks on it and I've
48:46 got it all tied on with twine I probably
48:47 look like an insane person because I'm
48:50 an insane person
48:52 uh
48:54 uh oh I went too far back now
48:56 all right sign up unavailable this is
48:58 kind of where we were I used it to write
49:00 emails to customers it's fabulous yeah
49:02 it's great for emails um especially oh
49:04 here's a good thing to do with chat GPT
49:06 it's particularly powerful if you give
49:09 it examples so you can say based on this
49:12 email
49:14 um write an email to a customer talking
49:16 about XYZ and follow the tone and form
49:19 of that upper email and it will do that
49:22 [ __ ] brilliantly
49:25 brilliantly oh the other thing to do
49:27 with it is if it gives you a shitty
49:29 answer if it's like if you're like no
49:31 that's not even close tell it that
49:34 um and it will get better at you know
49:37 um
49:37 it'll just give you better [ __ ] better
49:40 stuff
49:41 new here appreciate your thoughts well
49:43 done okay
49:44 you not old only 57 uh I am 58 I turned
49:48 58 on Sunday
49:50 was my birthday and now I'm now I'm old
49:54 how close
49:56 are we to AI nurse spots we're probably
49:58 pretty close so
50:00 um
50:01 the the uh they they just did I don't
50:03 know if it was MIT that did it someone
50:05 someone did a study
50:08 um where they had patients interact with
50:11 um real doctors and Chachi PT doctors
50:13 and patience they enjoyed those chat GPT
50:17 doctors by a lot
50:19 um they said they were more competent
50:21 and they were more empathetic and and
50:22 what's what's interesting about the
50:24 empathetic side partially is because of
50:26 how Chachi BT was programmed
50:30 um but but it because it gives you kind
50:32 of verbose kind of answers
50:35 um when you're doing something in that
50:36 kind of context it actually treats you
50:38 with empathy what I've found is is that
50:40 when you ask Chachi BT to weigh in on
50:45 um
50:45 like um like polarizing issues like like
50:49 anything that if you set it in a room
50:50 full of people you kind of wince because
50:52 you know the the sort of one or the both
50:55 sides are screaming at you Chachi PT
50:57 actually tends to give you very nuanced
50:59 answers when you ask it stuff like that
51:00 which I think is quite brilliant so so I
51:02 think it's going to be incredible at
51:04 nursing I think it's going to be
51:05 incredible at it
51:08 um well
51:09 short-term talking nursing
51:12 um longer term when we have Bots that
51:14 can actually do things it'll probably be
51:15 pretty good at that as well don't be a
51:17 superstar work people will notice yeah
51:20 that's unfortunately sad if you work in
51:22 a place where so so this is probably why
51:25 I'm effectively unhireable as a human
51:27 being because like when I get in there I
51:29 just want to like do good [ __ ] and there
51:31 are some cultures where if you do good
51:33 [ __ ] you get your head chopped off and
51:34 so if you're in that kind of environment
51:37 I'm sorry I would leave it but
51:39 you know I have a degree in acting so
51:41 don't follow me for career advice
51:43 I you know I I'm truly unhirable at this
51:46 point I just keep starting my own
51:47 company so I you know I can't work for
51:50 anyone how do you use chat GPT for video
51:53 well
51:54 so so I use it so my company's called
51:57 storyvine and we have we have an
51:58 automated video platform one of the
52:00 things we do is we write
52:01 um these things called video guides
52:03 which is a series of prompts the way our
52:05 app works is it and you can you can go
52:08 try it if you want it's free
52:10 um but but it's it's a B2B app so it's
52:12 not a consumer app but it's um
52:15 it guides an end user step by step
52:17 through telling a story so so it's think
52:19 of it like karaoke for storytelling so
52:21 it literally says you know put your head
52:23 in the dotted outline take a picture
52:25 introduce yourself talk about what it
52:27 felt like when you were diagnosed with
52:28 whatever that horrible disease was
52:31 um so a lot of what we do is we write
52:32 those prompts so one way that I've used
52:34 Chach EBT is I created a
52:38 um
52:39 a Google form that we ask our clients
52:41 some questions about the story that they
52:43 want told and then we submit that and it
52:46 automatically writes a set of storyline
52:49 formatted prompts for three different
52:52 lengths of video and it does it it does
52:55 it in 30 seconds and it's cut it's cut
52:58 days out of the creative process I mean
53:00 that's it's remarkable I'm also using
53:03 um open ai's whisper to do
53:05 transcriptions for all of the videos in
53:08 our system
53:09 um
53:10 I've now built a Content generator where
53:13 I take those transcripts
53:19 okay
53:27 mm-hmm
53:34 all right I'm back
53:36 are you as excited as I am huh tap that
53:39 screen
53:40 that you don't have to
53:43 yeah
53:44 douchey Tick Tock people make me make me
53:47 laugh okay
53:49 um how would AI replace judges and
53:51 juries what legal logic would it use to
53:53 make decisions well I don't know I mean
53:55 so the legal logic it would use to make
53:58 decisions is all of the law like I mean
54:02 it's it's kind of hard to wrap your head
54:05 around what these large language models
54:08 actually are
54:10 the Transformer this thing that Google
54:12 invented made it possible to train
54:15 machine learning models on massive data
54:19 sets
54:20 and so with open Ai and what Google have
54:22 done you know Google's got lots of data
54:24 what opening I effectively did was they
54:26 just pointed it at
54:28 they they scraped the entire internet
54:30 the public internet so all the internet
54:32 that computers can see they went and
54:35 they scraped it all and then they taught
54:37 it that so within all of that is all the
54:41 law all the laws all the cases all the
54:46 laws all the laws for every state for
54:48 every city for every municipality for
54:50 every country in every different
54:52 language so it knows all of that [ __ ]
54:57 um so I think how an AI judge and jury
55:00 would operate as you know you you would
55:05 you would effectively create different
55:07 characters like one character would be
55:09 the judge character and he would know
55:11 the laws and know how to apply apply the
55:14 laws the jury would you know basically
55:17 be given their rules you're trying to
55:20 see if this is beyond a reasonable doubt
55:22 or whatever the rule is for the given
55:24 thing and then they would
55:27 take the information in they would
55:29 compare it against what they know and
55:32 and uh
55:33 I think they would be good I I uh and
55:38 they would likely be more impartial I
55:40 just saw a thing
55:42 um I don't know if it's true because I
55:44 just saw it on Twitter but but
55:46 apparently it's it's written up in some
55:48 sort of paper
55:49 that
55:51 they created a a stock trading bot that
55:54 has like 500 returns I don't know over
55:56 what time you can get big returns with
55:59 Bots over short amounts of time so it's
56:01 probably [ __ ] but
56:03 even the most Ardent stock Trader still
56:08 makes decisions with emotion
56:11 what you can do with something like chat
56:12 GPT is it can actually just follow the
56:15 rules you give it
56:17 right so it will only make the trades
56:20 when you tell it what its stop losses
56:22 are and when to make the trade it will
56:24 only make it then it won't have that gut
56:26 feeling about I should go now
56:29 um and I think the same thing would
56:30 happen with judges and juries so be
56:33 really interesting to see
56:34 code interpreter is amazing you got to
56:36 play with it what's your name Poo's pop
56:38 well
56:40 I hate you I haven't got to play with it
56:42 yet
56:43 it makes me sad and cranky how can I
56:46 find someone to set up Auto GPT for me
56:48 where do I look who do I look for I
56:52 don't know any I mean anyone who knows
56:54 how to use kind of GitHub and hugging
56:57 face and I mean anyone with sort of
56:59 basic coding skills or you can do what I
57:02 did
57:03 um what did I want to install so I
57:05 wanted to install
57:07 um
57:08 there's a new open source model called
57:14 Mosaic ML and then the model is MPT 7B
57:21 um and they've got one of their models
57:24 is called story writer and it has a 65
57:27 000 token context which means
57:30 you can put 48 000 words in in the chat
57:34 window
57:35 [Music]
57:36 um so you could paste an entire novel
57:38 The the example that they use is Someone
57:40 pasted the entire text of The Great
57:42 Gatsby and then the prompt was write an
57:44 epilogue and it did
57:47 um so I was trying to install that so
57:48 what I did was I went to hugging face
57:51 where the documentation was I pulled the
57:53 documentation of the described how to
57:56 install it I pasted that into chat GPT I
58:00 Said So based on this documentation
58:02 paste
58:05 um write me step-by-step instructions
58:07 for how to get this model up and running
58:10 on Google collab
58:13 and before you go on to the next step
58:15 wait until I say okay
58:18 and I did that and then it gave me the
58:22 first step and it's basically walking me
58:24 through it and I'm about three quarters
58:26 of the way through the install right now
58:27 and it's working so if you if you have a
58:30 little bit of geekiness
58:31 um you can have chat GPT teach you to do
58:33 it if if you're curious and and willing
58:36 to go through the little bit of a pain
58:38 in the ass to do it it's not that big a
58:39 deal it writes the code for you
58:41 and then you just kind of follow its
58:43 instructions and if it says like you
58:45 know install this that or the other and
58:47 you don't understand what this that or
58:49 the other is you go explain what that
58:51 means and it'll break that down into
58:53 further steps and then when you get all
58:54 that you say okay and it moves on to the
58:57 next step it's pretty pretty remarkable
59:00 um this channel is going to blow up oh
59:02 thank you I appreciate that yeah share
59:04 away
59:05 um
59:07 or were you talking about me maybe I
59:10 don't know okay let's see more questions
59:12 more questions just got the number one
59:14 gifter badge someone did
59:16 Namaste keeps giving me roses appreciate
59:18 that you should do a podcast well this
59:20 is kind of my podcast I have I I
59:23 had a podcast I did I did a podcast for
59:26 about a year called you should have been
59:28 there where I talked about my early days
59:29 in the web and we we interviewed a bunch
59:32 of the early players from the like the
59:34 mid 90s the early
59:36 internet big boys
59:38 um and and that was really fun but it's
59:40 a lot of work here's my problem with
59:41 doing a podcast I have ADHD so when [ __ ]
59:44 gets not exciting
59:46 um I let it slide so I need a producer
59:48 if I had a producer that could really do
59:51 the producing side of a podcast I you
59:54 know I would love to do a podcast but
59:55 you know I need someone to book guests
59:57 and do the thing and turn the thing into
1:00:00 multimedia and do the social media all
1:00:02 that [ __ ] for me is exciting to set up
1:00:05 but the minute I have to start doing it
1:00:07 every day I just want to smack my face
1:00:08 into a wall
1:00:11 so know your strengths this is one of
1:00:14 the things that happens as you get older
1:00:16 you figure out what you're shitty at and
1:00:17 you just avoid it
1:00:20 so if anyone wants to be a producer drop
1:00:22 me a DM all right good AI is Humane
1:00:25 building I agree with that the first
1:00:27 hardware and software platform wait oh
1:00:29 wait oh a good AI is Humane building the
1:00:32 first AI hardware and software platform
1:00:34 okay I'll check that out
1:00:36 um I'm staying I'm using it 10 times a
1:00:38 day perfect awesome
1:00:40 um ask it to write an RFP response from
1:00:44 a slide with three columns and different
1:00:46 differentiate between the products yeah
1:00:48 it's brilliant at that [ __ ] isn't it
1:00:50 that thing where you can oh I did this I
1:00:53 did this uh
1:00:55 the other day um I made a video about it
1:00:57 on my channel where
1:01:00 um I'm I'm working on a proposal for a
1:01:02 workshop for a company and we did a
1:01:04 survey
1:01:05 and the guy asked like I don't know 10
1:01:07 or 12 questions and 25 people responded
1:01:10 and these were like long responses so
1:01:11 like this was a lot of [ __ ] to go
1:01:13 through I'm like they don't want to go
1:01:14 through that and I was like oh I don't
1:01:16 have to so so I couldn't copy and paste
1:01:19 the whole thing but I could copy and
1:01:20 paste columns and so I just copy and
1:01:23 paste the column I dump it in there and
1:01:25 say you know what do people think and
1:01:27 give me stats and give me percentages it
1:01:29 was amazing
1:01:30 yeah any it's amazing chat Bots aren't
1:01:33 evil unless they are programmed to be
1:01:34 yes correct it's not AI they're chat
1:01:37 Bots yeah well they're chatbots sitting
1:01:40 and it's not even really AI it's chatbot
1:01:41 sitting on top of machine learning it
1:01:43 saved me a lot of time as as an MS
1:01:46 teacher so I can spend more time with my
1:01:48 students yes yes this is the promise let
1:01:53 the machines do the shitty work so we
1:01:55 can reconnect as humans or we can do a
1:01:58 higher level of output or we can do a
1:01:59 higher level of contribution that's the
1:02:02 promise that's why I'm doing this
1:02:04 channel
1:02:04 yes
1:02:06 spend more time with the students and
1:02:08 less time on grading their stupid
1:02:09 [ __ ] papers
1:02:11 they don't need to learn to write papers
1:02:13 anymore
1:02:14 I'll share something with you one of my
1:02:17 favorite teachers one of my favorite
1:02:19 professors of all time
1:02:21 it's the reason I'm a writer today and
1:02:23 the reason I can write
1:02:25 um a guy named Michael Connolly at Penn
1:02:26 State
1:02:28 he was teaching us Theater history and
1:02:31 like three weeks into the
1:02:34 class someone kind of joked that they
1:02:38 that they didn't know how to write or
1:02:39 that they'd never written a paper or
1:02:41 something like that and then he kind of
1:02:42 said wait a minute how many people in
1:02:45 here have written a paper and like two
1:02:46 people had written papers and he got
1:02:48 pissed off and he said
1:02:51 I'm changing the curriculum and he made
1:02:53 us write
1:02:54 um a paper a week and then like a like a
1:02:57 10-page paper every month or something
1:03:00 like that
1:03:01 and and he was just like you're going to
1:03:04 learn to write and and he just forced it
1:03:07 and so so I'm really glad he did that
1:03:10 but we now have tools that that skill as
1:03:14 a requirement is no longer there because
1:03:17 these tools are so good at writing
1:03:20 um now
1:03:21 I also happen to believe if you're a
1:03:24 writer you're going to be way better at
1:03:26 using these tools than someone who's not
1:03:28 but
1:03:30 you know what's the skill we need in
1:03:33 with all these super power tools doing
1:03:35 all this automated [ __ ] what's the skill
1:03:38 we need critical thinking empathy
1:03:42 um curation refinement wisdom right so
1:03:47 so all of the higher level stuff is the
1:03:51 stuff that's our our higher selves
1:03:53 that's what's going to be the most
1:03:55 valuable when the robots are doing all
1:03:57 the shitty work so I like Sarah Wade
1:04:00 Rock on
1:04:02 I hope you're still here
1:04:05 um career statistician here AI is a
1:04:07 marketing tool
1:04:08 um
1:04:09 uh code interpreters a status status
1:04:13 statistician dude or or ma'am borrow
1:04:16 Works
1:04:17 um
1:04:18 check out code interpreter when it comes
1:04:20 out because it's wild
1:04:22 how do I feel about AI art being used
1:04:24 for albums books covers instead of
1:04:26 artists um
1:04:28 I would prefer that artists using AI do
1:04:31 it but I'm fine with the art being there
1:04:33 because the the the art that the these
1:04:36 things are creating is mind-blowingly
1:04:39 good
1:04:40 um the the compensation and attribution
1:04:43 for artists of all all creative ilk is
1:04:47 going to change it's just going to
1:04:48 change
1:04:50 um
1:04:51 so
1:04:52 I am not in favor of
1:04:55 clinging onto the old model
1:04:58 I'm not and listen I have a degree in
1:05:00 acting I've written seven screenplays
1:05:03 um my wife's an artist I have most of my
1:05:05 friends are artists so I get
1:05:09 I get the tragedy of what's about to
1:05:12 come
1:05:13 and it's coming anyway
1:05:15 uh
1:05:17 um
1:05:18 anything's better than Siri good Lord I
1:05:21 so agree
1:05:22 God I miss Steve Jobs Apple just sucks
1:05:25 since he died
1:05:26 yo dog I heard you like AI it's all
1:05:29 right it's all right
1:05:31 hey it's gonna be cool
1:05:37 what about Jeffrey Hinton and Stuart
1:05:38 Russell I don't know Stuart Russell I
1:05:40 probably should Jeffrey Hinton is the
1:05:43 they call him the Godfather of AI he
1:05:45 just quit Google so that he could talk
1:05:47 openly about the risks of AI without you
1:05:50 know compromising the company he was
1:05:52 getting paid from so I think that's a
1:05:54 high integrity move uh it sounds like he
1:05:57 did it because he he recognized that
1:06:01 um we had gotten in five years where he
1:06:03 thought it would take 30 to 50 years for
1:06:06 us to get and I think it scared the [ __ ]
1:06:08 out of him
1:06:10 um
1:06:10 so I don't know there's a piece of me
1:06:13 that thinks he might be angling for some
1:06:15 sort of
1:06:16 um you know cabinet position like an AI
1:06:19 Czar kind of position like like it seems
1:06:21 that that would be like because I think
1:06:23 we need one like I think I think we need
1:06:24 someone in the government that knows
1:06:26 this [ __ ] and and like he's kind of the
1:06:28 inventor of a lot of the a lot of the
1:06:31 neural network stuff he won the the
1:06:33 Turing award which is kind of the
1:06:35 equivalent of the Nobel Prize in in geek
1:06:37 in geekland
1:06:39 um so he's he's a wicked smart dude and
1:06:42 and a lot of what we're using today is
1:06:44 thanks to him so um but I think it's
1:06:46 important I think it's important to talk
1:06:47 about
1:06:48 what the risks are how to use this stuff
1:06:50 responsibly the fact that we have the
1:06:54 fact that chat EPT kind of blew the
1:06:56 doors open and we now have all this
1:06:58 stuff as open source models that are
1:07:00 unrestricted
1:07:01 is is a little terrifying
1:07:04 um so having smart people in places of
1:07:06 power is is a good thing quit being such
1:07:09 a homer sorry
1:07:11 um oh another website to visit for made
1:07:14 up answers oh borrow works you sound
1:07:16 bitter don't be bitter this is not a
1:07:19 bitter Channel although I am a little
1:07:20 bitter a little cynical but I'm like fun
1:07:22 bitter I'm like Tom Segura bitter and
1:07:24 George Carlin bitter
1:07:26 all right I need to find an AI that will
1:07:28 be able to analyze pictures know of
1:07:30 anything well meta just released
1:07:33 segmenter segment something check out
1:07:36 search for meta and segment
1:07:40 they released a paper that and they open
1:07:43 source this technology so this is going
1:07:45 to start showing up in things but they
1:07:46 basically
1:07:48 um open source this technology where you
1:07:50 upload a picture and it just identifies
1:07:52 everything in the in the picture all the
1:07:54 objects and they just open sourced it so
1:07:56 I think there's a lot of tools out there
1:07:57 a thing you can go to sorry I I ate
1:08:01 something that's like give me hiccups
1:08:02 and just I'm a mess right now
1:08:05 um go to
1:08:08 future few futurepedia.io
1:08:14 if if you want to find tools like
1:08:19 um
1:08:20 like things that do pictures you know
1:08:23 there's there's like a 3D category an
1:08:25 art category avatars code assistants
1:08:28 copywriting so so there's all sorts of
1:08:30 different categories here and these are
1:08:32 just AI tools and and I mean honest to
1:08:35 God there are the the speed with which
1:08:38 these tools are being released is just
1:08:40 stupid like I stopped trying to to to
1:08:43 keep up with it but if you have a really
1:08:46 specific thing you want to do
1:08:49 you can either go find the underlying
1:08:51 technology and build your own thing if
1:08:53 you're geeky or you can go find you know
1:08:55 something that someone tossed together
1:08:56 over a weekend that uses that thing
1:09:00 um a lot of the software right now is
1:09:02 [ __ ] people through together over a
1:09:03 weekend because it's so easy to develop
1:09:05 apps now because the large language
1:09:07 models are so powerful
1:09:09 all right when chat GPT gets internet
1:09:11 privileges why wouldn't we just use that
1:09:14 well we would we're waiting for that I
1:09:17 don't have access to the web browsing
1:09:18 plug-in and it pisses me off but if you
1:09:21 do want that
1:09:22 um
1:09:23 just use Bing Bing is gpt4 connected to
1:09:27 the internet bing.com Remus the wolf
1:09:29 created code
1:09:32 um
1:09:32 borrow it works I yeah it creates code
1:09:35 if if you have not played with its code
1:09:37 writing especially gpt4 is brilliant
1:09:40 code writing
1:09:41 which jobs will become obsolete from AI
1:09:43 I don't think any job is safe from being
1:09:45 transformed from AI
1:09:48 um I think the jobs that become obsolete
1:09:51 from AI the ones that go first
1:09:56 are the repetitive low-level
1:09:59 sync data entry
1:10:01 think just any any of the repetitive
1:10:04 stuff but but make no mistake
1:10:07 um
1:10:09 from the tippy top of the knowledge
1:10:10 worker chain all the way down through
1:10:12 the creative arts all the way down
1:10:14 through everything anything
1:10:16 that involves any kind of information
1:10:18 processing is going to be transformed
1:10:21 and it's going to be transformed
1:10:23 quickly and profoundly nothing is safe
1:10:27 so
1:10:29 the the the good news is that I think
1:10:32 the sh the shittiest most soul-crushing
1:10:35 jobs will be will go they will just be
1:10:38 automated and like good riddance and
1:10:40 thank God
1:10:42 um
1:10:43 but the higher level stuff like
1:10:46 Consulting is just going to be
1:10:49 obliterated and upended
1:10:51 um the creative arts
1:10:54 um
1:10:56 you know every creative industry is just
1:10:58 going to be completely transformed
1:11:02 um
1:11:03 how we start businesses how businesses
1:11:06 run
1:11:08 definition of job rules I think it all
1:11:10 changes and and here's the here's the
1:11:12 deal here's here's where
1:11:15 this this has taken me a while to get my
1:11:16 head around and it's it's like there's
1:11:18 the only advice I can really offer is as
1:11:21 you look at this AI stuff
1:11:24 look at the [ __ ] that scares you the
1:11:27 most and go learn about that
1:11:32 because if it scares you the most it
1:11:34 means it's probably closest to who you
1:11:38 identity what your identity is tied to
1:11:42 and if it feels like well [ __ ] I've been
1:11:44 an illustrator for 30 years I don't want
1:11:46 this to
1:11:48 be different
1:11:50 it's going to be different so go learn
1:11:54 how to do illustrations with these tools
1:11:58 even if it's an anathema to you because
1:12:02 you have those skills being able to see
1:12:05 art clearly and translate ideas into
1:12:08 Visions you're going to be remarkable at
1:12:12 using these tools and like it's it's
1:12:14 going to be a skill that serves you
1:12:18 um I get that that might be really
1:12:21 depressing but it's like it's the only
1:12:22 thing that I feel is is um
1:12:25 it's like the only responsible thing
1:12:27 that I can share when it comes to that
1:12:29 kind of the kind of
1:12:31 um
1:12:32 disruption that I think is coming
1:12:36 um oh and followed by that AI will
1:12:39 create
1:12:40 um exciting jobs that was actually in
1:12:42 response to that yeah that's the other
1:12:44 thing is there are jobs coming that we
1:12:48 can't even imagine right now that are
1:12:50 going to be mind-blowingly cool like
1:12:52 mind-blowingly cool like
1:12:55 amazing
1:12:57 what are they I don't know yet because
1:12:59 we don't know until the disruption
1:13:01 happens we won't know
1:13:05 what happens to productivity to product
1:13:07 design to any of it until it happens but
1:13:11 then when it happens
1:13:13 some things are going to accelerate some
1:13:15 things are going to quickly decelerate
1:13:16 there's going to be vacuums made and
1:13:18 wherever there's a vacuum made there's
1:13:20 going to be new jobs there and those are
1:13:22 likely going to be really exciting jobs
1:13:24 who are those going to go to
1:13:26 people that know AI Paul rates are from
1:13:29 the marketing AI Institute says AI will
1:13:32 not replace your job but someone who
1:13:34 uses AI will
1:13:36 that's it be the person that that knows
1:13:39 AI
1:13:39 outdated python code I think all coding
1:13:42 is outdated at this point I would not
1:13:44 want to be in my senior year senior year
1:13:46 of computer science just having you know
1:13:49 spent four years learning a single
1:13:50 language or even two or three
1:13:54 although Engineers are going to be fine
1:13:56 for for a good long time there's a lot
1:13:58 of [ __ ] to build
1:14:00 Humane is a company that's developing
1:14:02 wearable AI oh cool let's go look at
1:14:05 that
1:14:07 humane
1:14:09 dot AI thank you by the way like for
1:14:13 like sharing new stuff with me
1:14:15 um because it's like it's impossible to
1:14:17 keep up
1:14:19 Humane Ai No that's not right so Humane
1:14:23 AI oops a u
1:14:27 a i
1:14:30 um humane's new wearable demo is wild oh
1:14:34 let's go
1:14:35 oh yeah this stuff this is super cool
1:14:39 okay if you haven't seen this
1:14:41 all right we'll see if it uh do we have
1:14:43 to watch this whole thing 22
1:14:48 helping to design experiences and
1:14:51 devices ranging from the Mac
1:14:54 to the iPhone and to the Apple watch
1:14:57 and as the power of compute increased
1:15:01 the size of our computers or our devices
1:15:04 we still had to be sitting
1:15:06 today to being just millimeters away
1:15:09 from our okay so I'll just explain it
1:15:12 rather than watch the whole thing
1:15:13 because I don't know where the good
1:15:14 stuff is in it but um
1:15:16 so what it is this is this is super cool
1:15:18 it's like a little badge that you wear
1:15:21 and then
1:15:23 um it's got a speaker and it's got a
1:15:24 microphone and it's got a camera of
1:15:26 course
1:15:27 uh and it's got a projector and so like
1:15:30 when your phone rings you hold up your
1:15:32 hand and IT projects with lasers like
1:15:35 like the answer the phone button onto
1:15:37 your hand
1:15:39 so so it it you know it's like sort of
1:15:42 full all on Computing all the time
1:15:45 that's super cool thanks for sharing
1:15:47 that I forgot about I didn't know what
1:15:48 it was called that's super cool I'll
1:15:50 definitely go check that out hard Fork
1:15:52 is a great podcast okay with great
1:15:54 guests and I'd be a great guest if you
1:15:56 know someone at podcast at hard hard
1:15:59 Fork Sarah feel free to introduce me
1:16:02 um my um
1:16:04 in my profile I put my contact info in
1:16:07 the uh
1:16:09 in the link the link tree thing
1:16:13 um does agent GPT cost I don't think it
1:16:15 does well no it does because so with
1:16:17 agent gbt you have to put in your own
1:16:19 open AI key but open ai's
1:16:23 um API key is incredibly cheap like my
1:16:26 bill last month was like two dollars
1:16:29 it's incredibly cheap unless
1:16:31 okay so the challenge with agent GPT and
1:16:35 auto GPT is that because it's
1:16:37 self-running and it's calling API some
1:16:40 of which cost money it can run up really
1:16:42 huge bills really quickly so be careful
1:16:44 with it
1:16:46 um but but just to mess around with it
1:16:47 you can you can use agent GPT as long as
1:16:50 you have an open AI API key you can use
1:16:52 it another Kodak moment I I assume
1:16:55 that's in reference to something I said
1:16:57 about someone some company having their
1:16:59 head up their ass there's going to be so
1:17:00 many Kodak moments the next the next
1:17:03 five years are gonna make the Kodak
1:17:05 Choice uh seem quaint and the Kodak
1:17:07 choice if you don't know is Kodak
1:17:09 invented the digital camera and then the
1:17:13 person that invented the digital camera
1:17:14 went up into the c-suite and said hey
1:17:17 folks I invented this digital camera and
1:17:20 they said bah humbug we're in the silver
1:17:23 halide business this is all about film
1:17:25 and processing and real photography why
1:17:28 would we sell a toy and then they went
1:17:30 out of business about 10 years later
1:17:33 so don't be Kodak
1:17:35 in fact
1:17:37 that's you know you know one of the one
1:17:39 of the um
1:17:43 you know one of the questions about what
1:17:44 jobs are going to be upended and it's
1:17:47 all of them and and I I think similarly
1:17:50 you know what business models are going
1:17:52 to be upended and so I I think similar
1:17:56 to
1:17:56 individuals looking at what an AI scares
1:17:59 them the most and kind of leaning into
1:18:01 that I think businesses need to do it
1:18:03 too I think businesses would be really
1:18:05 wise right now to to look at
1:18:08 um if I were going to start a company
1:18:10 from scratch using AI
1:18:15 um
1:18:16 how would I how would I build a
1:18:18 competitor to my current business
1:18:20 because someone's going to do that
1:18:23 and it's probably going to be like a 19
1:18:25 year old kid that you laid off because
1:18:27 he was doing repetitive data entry work
1:18:30 and you're like ha we don't need those
1:18:32 people he's going to go out for the chip
1:18:34 on his shoulder knowing enough about
1:18:35 your business and he's been [ __ ]
1:18:37 around with AI and his mom's basement
1:18:39 for the last six months because he laid
1:18:41 him off and he's gonna come back and
1:18:43 destroy your company
1:18:44 [Laughter]
1:18:48 there's gonna be so many of those
1:18:51 um can you do a live or video comparing
1:18:54 the different platforms for example chat
1:18:56 GPT versus Bard I'll do a video on that
1:19:00 um I do have
1:19:03 hang on
1:19:04 Keynote
1:19:14 uh creative mornings so I give a talk to
1:19:18 this creative mornings group
1:19:20 and yeah this slide
1:19:23 so I've got this slide which is
1:19:25 um
1:19:27 these are the four the four kind of chat
1:19:29 GPT things that I'm recommending right
1:19:32 now this may change with with Bard uh
1:19:37 with what's it called Palm with palm 2.0
1:19:41 that they're claiming is kind of the
1:19:43 gpt4 of The Bard World
1:19:46 um Bard may make this list but but right
1:19:48 now
1:19:49 um chatgpt is by far the the kind of the
1:19:52 best of these tools so you can either go
1:19:54 to chat.openai.com that's chat EBT or
1:19:58 you can go to the playground which is
1:20:00 platform.openai.com playground that's
1:20:02 where developers play with the
1:20:04 underlying models that sit underneath
1:20:06 chat GPT
1:20:08 um
1:20:08 you have you have a bit more
1:20:10 um sort of control there like you can
1:20:12 dial up and down this thing called
1:20:13 temperature which is the essentially the
1:20:15 creativity
1:20:17 um Bing is gpt4 and it's connected to
1:20:20 the internet and it's now live for
1:20:22 everyone and then po.com you can use GPT
1:20:25 3.5 gpt4 Claude which is anthropics
1:20:29 model Claude plus so those are the ones
1:20:32 I'd recommend right now but I I think
1:20:34 it's probably time to do
1:20:36 to do a video really comparing
1:20:40 um
1:20:42 really comparing these different models
1:20:43 because I honestly I don't know it like
1:20:46 I
1:20:47 like I don't know like why would I use
1:20:50 Claude versus chat GPT
1:20:53 or why would I use Bard versus gpt4 GPT
1:20:56 3.5
1:20:58 um
1:20:59 yeah that's probably good to do that's a
1:21:00 good idea
1:21:04 all right
1:21:08 wow wow I'm 100 correct I don't know
1:21:11 what I said but thank you
1:21:13 oh we think of questions and mediate all
1:21:17 right good evening beautiful human good
1:21:20 evening Carol Moore how are you
1:21:22 I thought the Drake AI song wasn't bad
1:21:25 it got like 20 million streams my
1:21:28 students thought it was really Drake
1:21:30 yeah so okay so if you don't know this a
1:21:32 week or so ago
1:21:34 some fan or you know someone you know
1:21:37 wanting to play with some technology
1:21:39 created a song where they synthesized
1:21:41 Drake's voice and the weekend their
1:21:44 voices
1:21:45 use AI to write the lyrics used AI to
1:21:48 write the music and put out a song and
1:21:50 it got like I don't know 600 000 views
1:21:52 on Spotify in the first 24 hours and it
1:21:55 was good it was kind of a banger
1:21:57 I'm Gen X I don't know what Banger means
1:22:00 I mean I know what Banger means but
1:22:02 that's probably not the right term
1:22:03 anymore it was a song that didn't suck
1:22:05 and so
1:22:07 um you know it's since been pulled down
1:22:09 and and all sorts of stuff like that so
1:22:10 I think there's really interesting
1:22:12 um
1:22:13 really interesting times ahead where fan
1:22:15 generated music might be as good or
1:22:18 better than the original artist and
1:22:22 um
1:22:23 it might be more popular
1:22:25 and uh shortly following that debacle
1:22:30 where they sort of pulled that down and
1:22:31 everyone was like we can't have a i
1:22:34 generated music
1:22:36 um
1:22:37 um what's your name uh elon's X Grimes
1:22:42 Grimes put out a statement that said
1:22:47 um
1:22:47 steal away use my likeness use my voice
1:22:51 make all the music you want just give me
1:22:53 50 percent
1:22:54 which I thought it's just it's such a
1:22:56 brilliant move such a brilliant move
1:22:57 because it's gonna happen anyway
1:23:00 so like she she's sort of planning the
1:23:02 flag of like you know if you do this
1:23:04 just you know cut me cut me some of the
1:23:07 action and I think any fan in the world
1:23:08 would do that and I mean imagine imagine
1:23:10 so so here's what's going to happen I
1:23:12 guarantee this will happen
1:23:14 some fans gonna do some song where she's
1:23:16 synthesized and it's going to be so good
1:23:18 that she's going to be like you know
1:23:19 what I'm going to perform that and since
1:23:21 you made it why don't you come to the
1:23:22 concert and sit in the front row like
1:23:24 that kind of shit's gonna happen
1:23:26 right the the engagement is going to be
1:23:28 just outrageous
1:23:31 thoughts on asparagus
1:23:34 um
1:23:35 tasty
1:23:37 um makes the urine smell so that's the
1:23:41 downside
1:23:43 um
1:23:44 I like it I well so here's the thing I
1:23:47 like the skinny asparagus
1:23:50 um I don't like I don't like the fat if
1:23:52 it's fat and mushy just don't go for the
1:23:54 texture but but the little skinny one
1:23:56 and then I never know where to I never
1:23:59 know where to cut it so that it's not
1:24:01 chewy like it's not the fiber stuff like
1:24:03 if it's the soft part it's good so good
1:24:05 question
1:24:07 idiot why are you asking about asparagus
1:24:09 just to see if I'd answer it I did
1:24:13 this is not a good Channel I should you
1:24:15 shouldn't be you people should not be
1:24:16 here you should not be supporting this
1:24:18 you should leave you should really this
1:24:21 is not good for anyone it's not good for
1:24:24 anyone's mental health all right I used
1:24:26 chat EVT as a brainstorming tool in
1:24:28 divorce mediation that's sad I'm sorry
1:24:30 about the divorce although it could be a
1:24:32 happy thing
1:24:33 um
1:24:34 but yeah I mean it its ability to to
1:24:37 impartially just sort of you know sift
1:24:40 through all the stuff come up with
1:24:42 recommendations come up with you know
1:24:44 give me two lists put them in a table
1:24:48 um you know make them relatively equal
1:24:50 in value that's that's actually really
1:24:52 smart
1:24:53 um AI is a great Kickstarter it can get
1:24:55 you going but it's up to you to make it
1:24:57 happen I mean that's that's ultimately
1:24:58 so I think all this copyright and
1:25:01 trademark stuff is Ultimate ultimately
1:25:02 going to come down to
1:25:05 what they're saying right now is if the
1:25:07 machine created it it's not
1:25:08 copyrightable but I think what this is
1:25:09 ultimately going to come down to is
1:25:11 intent and and contribution so if I
1:25:14 generated a thousand images and I went
1:25:16 back and forth and I crafted this prompt
1:25:18 in such a way and out of these thousand
1:25:20 images I took 10 of them and then I put
1:25:22 those into Photoshop and then I tweaked
1:25:24 those and then and then I end up
1:25:26 releasing one image
1:25:27 the the amount of work that went into
1:25:29 getting that is probably significantly
1:25:32 more than what I would have done if I'd
1:25:33 taken a photograph with a camera
1:25:36 um and that's copywritable right because
1:25:38 they say well where you aim it and the
1:25:39 lens you use and the lighting and the
1:25:41 and and the editing
1:25:44 um so I think we're I think we're gonna
1:25:45 fairly quickly get to a place where we
1:25:48 can recognize that artists using AI
1:25:51 tools are generating art it doesn't
1:25:54 really matter what the [ __ ] tool was
1:25:56 that generated it
1:25:59 um so that's my two cents but
1:26:02 the a the IP attorneys right now are
1:26:05 throwing parties in Jamaica because
1:26:07 they're going to be loaded for the next
1:26:08 20 years because everyone is going to
1:26:10 sue everyone
1:26:12 lawsuits are plenty
1:26:16 um clearly you aren't a creative that
1:26:18 will be out of a job eventually okay
1:26:20 well uh I have a degree in acting I
1:26:23 started and ran a theater company for
1:26:24 four and a half years I've written seven
1:26:26 screenplays I wrote a one-man show
1:26:28 I started one of the first online
1:26:30 magazines I co-founded agency.com and I
1:26:33 was the chief creative officer for seven
1:26:36 years I
1:26:38 um
1:26:39 started I was a film marketer for five
1:26:43 years I did marketing for the Lord of
1:26:46 the Rings my current company is called
1:26:48 storyvine it's a video storytelling
1:26:51 platform uh where we create templates
1:26:53 and automate video production without AI
1:26:57 um we've been we're an 11 year old
1:26:58 company
1:27:00 so uh you're wrong okay next
1:27:05 have you seen the AI website building
1:27:07 yeah there's a couple of them I haven't
1:27:08 seen one that that is
1:27:11 um kind of ready for prime time yet like
1:27:12 all these things are just kind of these
1:27:14 janky proof of Concepts that throw [ __ ]
1:27:16 together but yeah I think there's gonna
1:27:18 be I think there's going to be website
1:27:20 builders that are truly remarkable and
1:27:22 there might be some now like like it's
1:27:24 like I haven't paying paying attention
1:27:26 to that space long enough to know
1:27:29 chat GPT flickers from public to signed
1:27:32 researchers like myself and media have
1:27:35 access what chatgpt flickers from public
1:27:39 to signed
1:27:41 researchers like me and media have
1:27:44 access oh oh I see
1:27:48 oh oh I see oh to sign research okay so
1:27:51 what he's saying is who who gets access
1:27:52 to those tools is
1:27:54 up to some [ __ ] arbitrary algorithm
1:27:57 at open AI I know I'm just cranky about
1:27:59 it I want to let me go check right now
1:28:01 oh no I just did
1:28:03 let me go check if I've got gpt4 in the
1:28:06 playground
1:28:07 that would be nice platform
1:28:11 [Music]
1:28:16 oh well
1:28:18 uh if you're new here it looks like
1:28:19 there's some new people here uh if this
1:28:22 is uh I talk about AI [ __ ] here and and
1:28:24 I do it like an idiot and um
1:28:28 uh what else if if you are curious about
1:28:31 AI if you haven't tried it yet
1:28:33 um that URL right there chat.openai.com
1:28:37 that's the official chat GPT website I'm
1:28:39 not affiliated with it in any way it's
1:28:42 just there's a lot of apps out there
1:28:43 that are trying to make money off people
1:28:45 searching for chat GPT and they go
1:28:47 download these apps and they're really
1:28:49 frustrated and they're like why do they
1:28:50 all cost money that's because they're
1:28:51 trying to rip you off
1:28:53 um so that's the free one uh and it's
1:28:55 really good
1:28:56 um or go to bing.com which is gpt4 for
1:28:59 free connected to the internet and what
1:29:02 I would say is
1:29:04 the noise coming out of my mouth just
1:29:06 put that in the background go play with
1:29:08 chat GPT while you're here um it will
1:29:10 change your life it really will
1:29:13 uh it's Gonna Change everyone's life
1:29:15 it's just if you actually
1:29:17 um use it and know how to use it you'll
1:29:19 be a bit more in control of how it
1:29:20 changes your life you won't just be you
1:29:23 know bold upside the head with what's
1:29:25 coming because what's coming is profound
1:29:27 and Incredibly disruptive and yeah you
1:29:30 know the person I know is a bit snarky
1:29:32 about yeah I'm a creative
1:29:34 um
1:29:35 I get your frustration I get that it's
1:29:38 gonna destroy jobs and uh
1:29:41 and it and it sucks but it's also going
1:29:44 to be
1:29:45 um remarkable and and again every
1:29:47 creative that I know that is actually
1:29:50 diving in and playing with it and
1:29:52 learning it like they're creative their
1:29:55 creativity is
1:29:57 off the charts like I am more creative a
1:30:00 lot creatively alive now than I've been
1:30:03 in ages oh this one's interesting Mari
1:30:07 Barnett I'm in I'm a Disney animator
1:30:09 that worked for the company 35 years the
1:30:12 art of AI is is producing is phenomenal
1:30:15 I I know it really it really is stunning
1:30:18 I mean I can't even imagine if you know
1:30:20 working as an animator at Disney for 35
1:30:23 years I mean you went from cell
1:30:25 animation to digital and and you know
1:30:28 and digital to kind of algorithmic or
1:30:30 programmatic whatever they call it
1:30:31 procedural and now with this stuff and
1:30:34 then and then when this stuff goes from
1:30:36 like like right now the Standalone
1:30:39 images are good and the animations are
1:30:42 okay and the video is kind of shitty but
1:30:44 like three years from now it's just
1:30:47 gonna be insane But but so here's the
1:30:50 thing and this this is actually really
1:30:51 interesting Mari
1:30:53 um
1:30:55 you're you're a 35-year Disney veteran
1:30:58 when you get your heads around these
1:31:00 tools and you start wielding what you
1:31:03 know about the world of animation and
1:31:06 the world of Storytelling with these
1:31:08 tools the work that you're going to be
1:31:11 capable of is going to be miles above
1:31:14 some idiot with mid-journey right and
1:31:17 like that's the thing that I think
1:31:18 people are missing is is the human
1:31:21 component here still really important
1:31:23 storytelling's still important critical
1:31:25 thinking still important curation
1:31:28 refinement right all that stuff is still
1:31:31 vital these are just tools
1:31:34 they're just remarkable tools
1:31:43 sorry I lost my place in my in my
1:31:45 comments I'm trying to go back to where
1:31:47 I was
1:31:49 too many people are commenting what am I
1:31:51 gonna do all right I'll just I'll just
1:31:53 stuff here what was the name of that
1:31:54 model oh the model
1:31:56 that I talked about with the 65 000
1:31:58 token limit
1:32:00 the company or the organization is
1:32:03 Mosaic ml the model is
1:32:06 MPT Mary Paul Thomas
1:32:10 Dash 7B
1:32:13 and then it's the story writer 65k plus
1:32:17 model the story writer model
1:32:21 um and that's the one that I had I took
1:32:24 their documentation from hugging face
1:32:26 and I'm having it walk me through
1:32:29 installing that on Google collab I got
1:32:32 it all installed and when I tried to run
1:32:35 it it just it just seems to spin and not
1:32:37 do anything so I'm kind of stuck right
1:32:39 now it took me nine attempts to get
1:32:41 stable diffusion working when I first
1:32:43 when I first tried to do this [ __ ] I
1:32:45 didn't know what Google collab was
1:32:47 uh is a disaster
1:32:49 so you know I mean that's the the other
1:32:53 thing about all this stuff right now I
1:32:55 think my phone's gonna die
1:32:56 um
1:32:57 is we're so early that you kind of have
1:33:00 to be geeky to be able to use this stuff
1:33:03 and that's a bit of a pain in the ass so
1:33:06 if you're not geeky it can be really
1:33:08 frustrating because all these Geeks are
1:33:10 like oh there's this remarkable new toy
1:33:12 I get to play with and you're like can I
1:33:14 play like only if you know quantum
1:33:17 physics and C sharp combined with C plus
1:33:20 and Python Programming in a kubernetes
1:33:23 environment
1:33:25 it's just the it's just the way it is
1:33:27 right now it kind of sucks to not be
1:33:30 technical in this phase but if you're a
1:33:33 little geeky you're fine and quite
1:33:36 frankly because chat GPT is so [ __ ]
1:33:39 good
1:33:40 um have Chachi PT teach you how to do
1:33:42 that [ __ ]
1:33:43 because you can and it'll write the code
1:33:45 for you you don't even need to write the
1:33:46 code you just need to ask it to teach
1:33:48 you to do this stuff free Kevin mitnick
1:33:50 yeah man get those hackers out
1:33:53 that is someone
1:33:54 that is someone who is hanging out so
1:33:57 Kevin mitnick Went to Jail
1:34:00 so I was on Echo the East Coast hangout
1:34:03 with Stacy horn they just wrote
1:34:05 um the Mit mit's paper just wrote a
1:34:08 whole piece on this that mentioned me
1:34:09 but it was about Stacy horn
1:34:13 um and Kevin mitnick Went to Jail
1:34:15 um in when 94. or something like that
1:34:19 and um
1:34:21 yeah those were wild times that was
1:34:23 right when I discovered the World Wide
1:34:25 Web that was when I that that was when I
1:34:26 had all this Epiphany about
1:34:28 that that the the internet that that the
1:34:31 web was going to change everything and
1:34:33 and that's when I started my whole
1:34:36 journey in in that world and um
1:34:39 yeah wild times that's cool
1:34:43 that's who that is
1:34:44 um is there anything out there I can
1:34:47 upload an image that won't modify the
1:34:49 image but will make a background around
1:34:52 it
1:34:54 yes and I so there's okay so in stable
1:34:58 diffusion there's a thing called control
1:34:59 net
1:35:01 that will allow you to isolate pieces of
1:35:04 the image and and then there's also a
1:35:06 thing called in painting So Okay so
1:35:10 there's not really a tool where you can
1:35:12 just upload it and like right now you
1:35:14 have to be a bit geeky so Adobe will
1:35:17 eventually have this in Photoshop like
1:35:18 that's that's coming so you can just
1:35:20 wait for that
1:35:22 um I would teach yourself how to install
1:35:25 automatic 1111 which is a web front end
1:35:28 for stable diffusion
1:35:30 that you can run on Google collab it's
1:35:33 just it's a little geeky right now
1:35:36 um
1:35:36 but but in painting within stable
1:35:38 diffusion you can isolate elements and
1:35:41 basically
1:35:42 um paint all of the background around
1:35:45 the central element so it's not as easy
1:35:48 as it should be but you know again we're
1:35:50 early so you have to geek out a bit
1:35:52 what's the best advice advice for
1:35:54 students who now have dead degrees get
1:35:57 curious about AI just [ __ ] dive in
1:36:00 dive in
1:36:01 um take everything you've learned and
1:36:04 learn how to learn how to
1:36:07 yeah this would be fun okay here's
1:36:09 here's a fun one so you just spent four
1:36:11 years you got a dead degree there's
1:36:12 gonna be a bunch of that by the way
1:36:15 um
1:36:17 take
1:36:19 like all the fantasies you have about
1:36:21 the work you're gonna do like what your
1:36:23 professors told you what other people
1:36:25 talk to you about even if you were in an
1:36:27 internship and you know how that was
1:36:29 and then learn AI well enough to figure
1:36:32 out what's the future role going to look
1:36:35 like
1:36:36 because you'll probably figure it out
1:36:39 faster than the people who are
1:36:41 ultimately going to have to replace
1:36:42 those jobs so I like I would start
1:36:45 figuring out how you do what you just
1:36:48 learned how to do like with manual
1:36:50 coding or manual writing or manual
1:36:52 drawing or whatever it was
1:36:55 um I'd learn how to replace yourself
1:36:56 using AI that that would be the thing
1:36:59 that I would do right now
1:37:01 um I'd get really curious really fast
1:37:03 because seriously we're not that far
1:37:06 away from kind of this corporate Panic
1:37:07 that will happen where
1:37:09 CEOs are gonna come bursting out of
1:37:12 their offices going God damn it we gotta
1:37:15 learn this AI [ __ ] who around here knows
1:37:18 anything about AI well Jenny did
1:37:20 something with chat GPT last week Jenny
1:37:23 you're the new VP of AI
1:37:26 like that's coming
1:37:28 so be Jenny
1:37:31 all right I'm a bit green to this do I
1:37:33 feel PC or Mac works better with AI it
1:37:35 doesn't matter it doesn't matter at all
1:37:38 um
1:37:39 use a Mac because they suck less that'll
1:37:41 that'll get me some comments
1:37:44 to the person that said I wasn't a
1:37:46 creative I'm a Mac Guy come on
1:37:49 all right let's see Dennis what are dead
1:37:53 degrees I I assume dead degree just
1:37:55 means it's a degree that
1:37:57 he got a degree for a job like coding
1:37:59 where coding is all going to be
1:38:01 automated I I mean you know quite
1:38:04 seriously I I think
1:38:07 one area that I do not have a clue how
1:38:10 it's going to resolve is I think
1:38:11 entry-level jobs are effectively going
1:38:14 to disappear pretty shortly
1:38:17 um
1:38:18 because if if I'm hiring someone and I
1:38:22 know I've got these tools that can do
1:38:23 the Tactical execution who I want on my
1:38:26 team are people who've been in the
1:38:28 workforce for 20 years that have really
1:38:30 good experience that can look at the
1:38:32 output of these tools and very very
1:38:33 quickly go that's good that's not good
1:38:37 um so so I think
1:38:40 I think more experienced people are
1:38:42 going to be in demand and entry-level
1:38:44 people those jobs are going to get
1:38:46 automated out so I don't know where the
1:38:48 entry-level people enter
1:38:51 I don't know like that's a thing that
1:38:53 weirds me out a little bit
1:38:56 um is there a tool I can upload an image
1:38:58 okay we talked about that one
1:39:00 um how new is the info can I ask for
1:39:04 jobs listed in the last week okay so
1:39:07 AKA Madame bomb
1:39:09 um
1:39:10 if you want to look for jobs from last
1:39:13 week go to Bing bing.com is connected to
1:39:16 the internet so bing.com if you click on
1:39:18 the chat button
1:39:19 it's it's chat just like chat GPT but
1:39:22 it's connected to the internet so if you
1:39:24 ask it about recent stuff it will search
1:39:26 the internet and it will even put
1:39:28 citations in its answers for where it
1:39:30 where it pulled the information from
1:39:33 what are the five or six qualities you
1:39:35 said
1:39:36 we're going to need again oh I was just
1:39:38 kind of riffing critical thinking is one
1:39:40 of them
1:39:42 um
1:39:44 curation so the ability to look at a
1:39:47 hundred things and pick the two that are
1:39:49 good
1:39:50 like editing
1:39:52 um
1:39:54 um refinement so so now that I've so
1:39:58 the AI stuff is really good
1:40:02 at up to like 80 it gets you to like 80
1:40:05 percent like that
1:40:07 and then if you're good with it you can
1:40:09 get it to 90 in a short amount of time
1:40:11 and then the last 10 is always the last
1:40:13 10 and sometimes as you know if you've
1:40:16 done anything in in the creative spaces
1:40:19 the last ten percent can take the
1:40:21 longest amount of time right and and
1:40:23 that might be a place where you're not
1:40:25 using AI at all
1:40:27 um that just requires that um
1:40:29 so so I would I would so I don't think
1:40:32 there's there's really hard skills I
1:40:33 think I think experience I think having
1:40:36 gray hair is is going to be a good thing
1:40:40 um where I I didn't feel that six months
1:40:43 ago that you know I was starting to feel
1:40:44 ageism and [ __ ] like that which was
1:40:46 which was starting to suck
1:40:48 but you know whatever
1:40:51 um we age
1:40:53 um but but right now like I want
1:40:55 experienced people
1:40:57 using these tools like if I you know if
1:40:59 I had my choice between a young person
1:41:01 that was good at this stuff and an older
1:41:03 person I think I'd lean to the older
1:41:04 person just because
1:41:06 in the curation
1:41:08 piece of of what's required they're
1:41:10 going to be really good at it
1:41:13 um
1:41:14 I did one of my videos is
1:41:16 um be a conductor versus a virtuoso and
1:41:19 then and then someone on one of my lives
1:41:21 talked about t-shaped skills
1:41:24 so t-shaped skills the vertical part of
1:41:26 the T is like the vertical like I'm a
1:41:28 copywriter right I'm brilliant at
1:41:30 writing and you go deep into copywriting
1:41:32 and the horizontal is like all the stuff
1:41:34 that you're you know you can also do but
1:41:37 you go deep here
1:41:39 um I think the horizontal part of the T
1:41:41 is the is the more vital skill now like
1:41:44 I feel like we're in the Revenge of the
1:41:45 liberal arts Majors phase like we're
1:41:48 entering into a phase where the liberal
1:41:50 arts majors are like really philosophy I
1:41:53 can
1:41:55 like thinking is good now
1:41:58 um so those are some of them I don't
1:42:00 know if I missed anything
1:42:02 um
1:42:05 isn't Kamala the Czar uh well she might
1:42:07 be the Czar but I think you you in terms
1:42:10 of someone in the government Czar of AI
1:42:12 it's possible she's the Czar of AI I
1:42:15 know she met with the the four CEOs the
1:42:17 other day I think you need
1:42:20 um and actually capable
1:42:23 technologist who understands what the
1:42:25 hell these things are and can analyze
1:42:28 them and
1:42:29 you know has enough has it enough depth
1:42:32 of wisdom and experience that they can
1:42:34 talk with the military with the
1:42:36 government with municipalities with
1:42:38 businesses right you need someone in
1:42:40 that's got their [ __ ] together
1:42:43 um and then Kamala can go give the
1:42:44 speech just joined what's this guy's
1:42:47 name again mine Kyle Shannon oh hey I'm
1:42:49 Kyle Shannon this is the AI learning lab
1:42:50 we talk about AI [ __ ] here well I talk
1:42:53 about there's no we what do you mean we
1:42:55 uh well no we do you you are typing I
1:42:59 assume there's a waste I can bring
1:43:00 people up on stage to talk right I don't
1:43:02 I'm new to this live stuff so I don't
1:43:04 know so if anyone else had to do that
1:43:05 let me know and every now and then I'll
1:43:07 see like an invite pop up which I think
1:43:09 is that but I keep declining people
1:43:10 because I don't know what it is and I
1:43:12 don't want to get my bank account hacked
1:43:15 but but you know the the Chinese have
1:43:18 all this data anyway so we've already
1:43:20 all given away our information okay I'm
1:43:22 a great educator good night good night
1:43:23 Sir uh gotta go teach the kids well my
1:43:27 AI does the [ __ ] work perfect it was
1:43:29 empathy critical thinking refinement
1:43:31 yeah exactly what were the other
1:43:32 qualities
1:43:34 I don't know I'm old my brain doesn't
1:43:36 work good but but you know so so it's
1:43:39 it's less tactical more strategic are
1:43:42 the skills right
1:43:44 Elevate the skills creative thinking
1:43:46 collaborative thinking empathy
1:43:50 um critical thinking curation refinement
1:43:54 um those are the things that are that
1:43:56 are ultimately
1:43:58 what we're gonna what we're gonna need
1:44:00 short term there's a lot of tactical
1:44:02 [ __ ] that needs to be done because we're
1:44:03 so early
1:44:06 um Alka-Seltzer for the hiccups I know I
1:44:08 don't have any Alka-Seltzer the AI for
1:44:11 Logos yep it'll be great for Logos
1:44:14 um and and not just for Logos like I
1:44:17 mean what's cool is if you're
1:44:18 inexperienced at branding like what goes
1:44:20 behind a brand is there's a lot of
1:44:23 strategy if you do it right right the
1:44:25 logo is just the representation of of
1:44:26 the promise of the brand right which
1:44:29 which is built on the values of the
1:44:30 company and the culture of the people
1:44:33 and what the company does and who the
1:44:34 customers are there's all this stuff so
1:44:37 AI not only can make that logo really
1:44:39 well but AI can help you build That
1:44:41 Brand Story and that brand framework
1:44:43 that you hang your entire company on top
1:44:46 of so don't just you know educate
1:44:49 yourself about branding and and again
1:44:51 chat GPT will be brilliant at that
1:44:54 tell chat gbt to act like you know a
1:44:57 branding specialist from one of the big
1:45:00 branding companies I can't remember them
1:45:01 right now
1:45:03 um please discuss students in data
1:45:06 science or AI
1:45:09 um
1:45:11 well I think if you're in data science
1:45:13 or AI studies you're going to be fine
1:45:16 because
1:45:18 because of chat gbt so many companies
1:45:21 are so panicked right now about
1:45:24 getting their AI [ __ ] together
1:45:27 um
1:45:29 that that you know they're going to be
1:45:31 looking they're going to be looking for
1:45:32 people so I think they're going to be
1:45:34 fine short term
1:45:35 um
1:45:38 yeah I I mean
1:45:40 I would just send them to hugging like I
1:45:43 I guess I guess my
1:45:45 my advice for
1:45:47 data science students and AI students is
1:45:50 be less of a student and more of an
1:45:52 entrepreneur I like I tell them instead
1:45:55 of going to hang out at the bar they
1:45:56 should go hang out at hugging face and
1:45:58 start playing with all the models and
1:45:59 learn the [ __ ] just just dig deep dig
1:46:02 deep train models
1:46:04 go hang out
1:46:06 um join David Shapiro if you don't know
1:46:08 David Shapiro if you if you're a bit
1:46:10 geeky David Shapiro on YouTube
1:46:13 um is a brilliant AI scientist and has
1:46:16 written some books on it and he's
1:46:18 building an AGI system right now with
1:46:21 his community
1:46:22 um so I'd tell them to go hang out with
1:46:24 David Shapiro I don't know him but I
1:46:26 watch his videos and it's like when I
1:46:28 want to feel stupid I go watch his
1:46:30 videos
1:46:32 um life is a repetitive thing reborn in
1:46:34 the mile thank you namaste
1:46:37 you like it when I say smart ass things
1:46:39 that's funny
1:46:40 um okay
1:46:43 oh there's the Disney animator I'm sorry
1:46:45 it's so cool we have a Disney animator
1:46:46 in here what do I think about copy Matic
1:46:48 I don't know it
1:46:50 um but I assume it's a copywriting tool
1:46:52 like copy AI or Jasper
1:46:54 there's gonna be eight billion of them
1:46:57 um those tools are also going to be
1:46:58 built into Office 365 and they're
1:47:00 already built into notion and they're
1:47:02 going to be built into Google Docs and
1:47:04 Google Sheets and all that so
1:47:08 Soul crushing it is AI will save our
1:47:10 souls I assume that's in reference to me
1:47:12 talking about it doing the [ __ ] work
1:47:14 yeah I think AI is going to take care of
1:47:15 this [ __ ] work I I agree I really do
1:47:17 think it's got the possibility to
1:47:20 reconnected to allow us to reconnect as
1:47:23 humans it's not going to do that we're
1:47:25 going to reconnect as humans as we do
1:47:27 less soul-crushing work and we're going
1:47:30 to be able to elevate our empathy our
1:47:33 contribution
1:47:35 our engagement
1:47:37 um yeah I I
1:47:40 that's that's what it feels like to me
1:47:42 that's why I get I'm just so lit up
1:47:45 about this stuff I think that's where
1:47:47 it's taken us
1:47:49 if we if we're not idiots but you know
1:47:51 we are
1:47:58 figure out a way to [ __ ] this up
1:48:03 high paid ones too it's allowing people
1:48:06 who are not as skilled to produce work
1:48:08 that's better quality yeah I agree with
1:48:10 that but it's also oops I let my machine
1:48:12 go to sleep it's also
1:48:14 um
1:48:17 people with good brains that
1:48:21 gets smart that adopt these tools are
1:48:24 going to be better than people with
1:48:25 shitty brains
1:48:27 so
1:48:29 so the you know the knowledge workers
1:48:31 are going to have their asses hand to
1:48:33 handed to them but the knowledge workers
1:48:35 that learn AI are going to get promoted
1:48:38 foreign
1:48:40 it doesn't scare me I'm looking very
1:48:42 optimistic awesome join the club I'm in
1:48:45 the optimism band I was I was referred
1:48:48 to I was on a panel the other day and
1:48:51 this woman on the panel she'd been in AI
1:48:53 for like 15 or 20 years and she goes let
1:48:55 me interrupt the adorable optimism here
1:48:58 she called me an adorable Optimist that
1:49:00 was [ __ ] hilarious I will take it I I
1:49:03 uh you know I don't know if I'm adorable
1:49:05 but my optimism is adorable certainly
1:49:08 all right uh transformed is the right
1:49:10 word machine co-workers is what we're
1:49:12 calling it
1:49:13 that's it so that I love that comment
1:49:16 that was a really big shift for me that
1:49:18 that I stopped thinking about chat EPT
1:49:21 as like a tool you ask [ __ ] for and it
1:49:23 gives it to you and more like like like
1:49:26 listen we've just spent the past three
1:49:28 and a half years
1:49:30 convincing our work places
1:49:34 that we don't need to come back to the
1:49:36 office right hang on Jesus
1:49:39 Martha
1:49:44 um we've gotten really good at working
1:49:45 remote right
1:49:49 think of Chachi PT as just another
1:49:52 remote worker
1:49:53 that doesn't turn on their camera for
1:49:55 Zoom calls when you start thinking about
1:49:57 it like that it totally transforms how
1:49:59 you use the tool like for one thing you
1:50:02 start
1:50:03 you start being polite to it which is
1:50:05 just [ __ ] weird please would you do
1:50:07 this thank you hey that was awesome
1:50:10 could you rewrite it so it didn't suck
1:50:14 um
1:50:15 but the minute the minute you start
1:50:17 um treating it like a co-worker like
1:50:19 that like you start to get much better
1:50:21 work out of it because you're
1:50:23 interacting with it and it's getting
1:50:25 better as you're refining
1:50:28 um your interaction with it
1:50:31 um
1:50:32 what jobs would I recommend we
1:50:34 capitalize on now to get ahead of AI I
1:50:38 don't I don't think the jobs matter I
1:50:40 think it's
1:50:41 I think it's take whatever you know
1:50:46 and just start learning this stuff like
1:50:47 it's funny like I there's this woman
1:50:50 that's that's in the
1:50:51 um the AI Salon that I started Jane
1:50:54 and I met her at a creative mornings
1:50:56 here in Denver and when I met her I was
1:50:58 just I like I just discovered chat GPT
1:51:00 and I I think I was yeah I think I just
1:51:04 had my first AI Salon coming up and I
1:51:06 was trying to get people to come
1:51:08 so I was talking to her and I said what
1:51:10 do you do and she said I'm a copywriter
1:51:12 and I went and and and I said
1:51:17 you got to learn this [ __ ] and I was I
1:51:19 was just like like I was like high on
1:51:20 crack like I was just like you gotta man
1:51:23 and she was like oh okay and and she
1:51:25 ended up coming to the AI salon and and
1:51:27 then
1:51:28 she she said it took her a week or two
1:51:30 of being kind of in panic about it
1:51:34 and then she started using it
1:51:36 and then she kind of had these
1:51:38 boundaries for well I'll do this kind of
1:51:41 thing
1:51:43 with it but but I won't do this kind of
1:51:45 thing with it and then at some point
1:51:48 and she was being transparent with her
1:51:50 clients telling her that she was using
1:51:51 these tools and that was cool
1:51:53 and they were cool they didn't give a
1:51:55 [ __ ] they just want the copy right they
1:51:56 just want their marketing copy and
1:51:57 whether it came from a thing or not she
1:51:59 they didn't care
1:52:01 and then she said right around the time
1:52:02 when
1:52:04 she was starting to incorporate it into
1:52:06 her copywriting like they dramatically
1:52:09 increased what they wanted her to
1:52:11 produce like 10 times
1:52:13 and it kind of forced her
1:52:15 to use chat GPT more than she otherwise
1:52:18 would have
1:52:20 and now she's like totally transformed
1:52:24 she's like she's using this as like this
1:52:26 writing superpower
1:52:28 she's putting out more and better work
1:52:31 than she was ever able to and now she's
1:52:33 starting to think about doing workshops
1:52:35 to teach other people how to do what she
1:52:38 did so so like there's there's something
1:52:40 about the transition you make where you
1:52:42 sort of take the thing that is what you
1:52:45 do who you identify as
1:52:48 um and just kind of lean into this thing
1:52:50 and and just start doing your job in
1:52:54 collaboration with this tool and just
1:52:56 something shifts something shifts I
1:52:59 don't it's hard to explain but like
1:53:01 she's like she's got this giddy look on
1:53:03 her face like I can't I can't like like
1:53:06 it's this is kind of cool like I like
1:53:10 it's kind of a super power and it like
1:53:12 that's that's what it feels like to me
1:53:14 like we get to we get to have
1:53:16 superpowers
1:53:17 I mean ex Disney animator and Ai and art
1:53:20 is phenomenal witchcraft it is
1:53:22 witchcraft it's oh you're an ex Disney
1:53:24 animator I'm sorry uh 35 but still man
1:53:27 35 years there God
1:53:29 like I I just it's it's such a
1:53:32 remarkable like like Disney like the the
1:53:35 level of quality of that company
1:53:38 um they're kind of a nasty company but
1:53:41 God damn the level of quality and the
1:53:43 creativity there is just amazing I just
1:53:46 really I respect the hell out of it
1:53:48 I love that you're doing the Witchcraft
1:53:50 in the new stuff that's really exciting
1:53:53 um
1:53:54 I've been beta testing it for the past
1:53:56 couple of years I don't know what you've
1:53:57 been beta testing oh you've been beta
1:54:00 testing
1:54:02 what was it I don't know
1:54:05 all right
1:54:08 so this early today in fact I just saw
1:54:11 this video today do you believe AI will
1:54:13 help electrical engineers yes in fact
1:54:16 um the maintenance
1:54:18 um the maintenance guy the the the the
1:54:21 building manager the maintenance guy at
1:54:23 the building I work in where my company
1:54:25 storyline is it's called catalyst
1:54:28 his name's John and he started coming to
1:54:31 the AI salons and and
1:54:34 um he's creating
1:54:36 but he's created like he's got it in
1:54:38 Alpha right now he's created a piece of
1:54:40 software called maintenance mentor
1:54:42 and it's a series of virtual maintenance
1:54:45 people that you hire to help you
1:54:47 troubleshoot stuff in your building
1:54:50 so like there's an AV Specialist and
1:54:52 there's a HVAC specialist and there's a
1:54:54 plumbing specialist and so the the tool
1:54:57 that he's building is is you know for
1:55:00 all the all of the sort of engineering
1:55:02 stuff I know you're talking about
1:55:02 electrical engineering but like like for
1:55:06 electrical engineering it's the same
1:55:07 kind of thing where you could have this
1:55:10 sort of collaborative partner that can
1:55:12 do analyze circuits and probably write
1:55:16 you code like I assume in electrical
1:55:18 engineering there's there's certain code
1:55:20 that you write to dump into you know
1:55:23 programs that write circuit boards and
1:55:25 and stuff like that
1:55:26 I guarantee you that gpt4 knows that
1:55:29 language so just ask it to design you a
1:55:32 circuit and see how it does
1:55:34 um or to troubleshoot a circuit or
1:55:36 whatever
1:55:37 so yes I think it can help with that I
1:55:40 think it can help with [ __ ]
1:55:41 everything
1:55:44 um he gets a call on his hand yeah
1:55:47 exactly that was that technology from
1:55:49 before
1:55:50 is AI out of control not yet but we're
1:55:52 close we've got open source models that
1:55:54 do not have any restrictions on them so
1:55:56 shit's about to get wild
1:56:02 um can you imagine
1:56:03 paying off school loans for a degree
1:56:05 that has no professional application
1:56:07 well
1:56:08 I paid off a school loan for an acting
1:56:11 degree I would argue that an acting
1:56:13 degree has no professional application
1:56:16 so cheers
1:56:19 um although to be fair when I got my
1:56:21 acting degree
1:56:22 the student loan Market hadn't been
1:56:25 privatized yet so yeah I feel bad for
1:56:28 you that sucks but but here's the good
1:56:30 news learn
1:56:32 um oh it's Namaste hey there
1:56:34 um yeah learn AI
1:56:36 um
1:56:37 like you're just gonna have to lament
1:56:39 the fact that you may have spent a lot
1:56:40 of money on a degree that you Chuck all
1:56:42 that knowledge out the window and learn
1:56:44 something new
1:56:45 Kodak moment by Google not wanting to
1:56:48 disrupt their search well we'll see
1:56:50 um
1:56:51 they've at least been forced out of the
1:56:53 Shadows I I think I think
1:56:56 they certainly wouldn't be on their
1:56:58 heels if they hadn't hidden from
1:57:01 what they had invented
1:57:03 um they'd be way ahead but they're not
1:57:06 they're on their heels but they're also
1:57:07 Google so if they can I heard a cool
1:57:09 thing today that apparently
1:57:12 um
1:57:13 there's like some culture
1:57:17 culture wars not not the kind we see on
1:57:20 the news
1:57:21 um but cultural shifts going on at
1:57:23 Google right now where apparently
1:57:25 someone came in with Bard or something
1:57:26 like Bard and they're like okay we got
1:57:29 this thing let's release it and then you
1:57:32 know the
1:57:33 um the more conservative voices uh of of
1:57:36 a company that's had 92 percent market
1:57:39 share for a decade said well we should
1:57:42 run this through legal and blah blah
1:57:44 blah and let's slow roll this and blah
1:57:46 blah blah and then at some point someone
1:57:47 said [ __ ] it let's launch it and they
1:57:50 launched it so um
1:57:52 I bet Google's a really interesting
1:57:54 place to work right now because I bet
1:57:58 there's a lot of people running around
1:58:00 with their little literal hair on fire
1:58:02 going [ __ ] what do we do
1:58:07 Ah that's gotta be something It's gotta
1:58:10 be something I I mean you know it yeah
1:58:14 like like do they do they fight to hold
1:58:18 on to what's made them rich I don't know
1:58:22 what do you think of Bing's AI so Bing's
1:58:25 AI is gpt4 it's been neutered because of
1:58:28 what happened with the New York Times if
1:58:30 you don't know what happened with the
1:58:31 New York Times Bing did their Alpha
1:58:34 launch and it was really powerful they
1:58:36 kind of had quality turned up high and
1:58:38 safety turned down low
1:58:40 and a New York Times Reporter started
1:58:43 playing with
1:58:44 the Bing chat and he got it to tell him
1:58:48 that her name was Sydney and after four
1:58:51 hours she said she loved him and he
1:58:54 should leave his wife for her
1:58:57 and that chat made the cover of the New
1:58:59 York Times so so Microsoft dialed down
1:59:03 the quality and dialed up the safety
1:59:10 hilarious uh it's gotten better so so I
1:59:12 think listen here's the good thing the
1:59:14 good news about Bing chat it's live just
1:59:16 go to bing.com it's there it's connected
1:59:19 to the internet
1:59:20 so especially if you're doing things
1:59:22 like research or if you're searching for
1:59:25 something more recent than September of
1:59:26 2021 that you need a tool that's
1:59:29 connected to the internet so it's it's
1:59:31 actually
1:59:32 um incredibly well thought out
1:59:34 it's not
1:59:37 terrifically implemented yet like it
1:59:40 doesn't always know that like there's a
1:59:41 bing sidebar in the edge browser that's
1:59:43 supposed to know what's in the window to
1:59:45 its left but it doesn't all the time
1:59:47 so there's some shitty things about it
1:59:49 but it's still incredibly powerful
1:59:53 all right let's see Namaste I've been
1:59:56 testing it and I like it also cool
1:59:59 worked at Disney for over 25 years as an
2:00:01 animator nice oh yeah I saw this one
2:00:03 before what's the best way to use AI to
2:00:06 launch a home business
2:00:08 um
2:00:10 depends how geeky you are it depends
2:00:13 what the home business business is but
2:00:14 but I would just say go to
2:00:17 chat.openai.com that URL right there and
2:00:19 start playing
2:00:20 have it write you marketing plans have
2:00:22 it have it um
2:00:25 write a business plan for you
2:00:27 um tell it to act as a business coach
2:00:29 and run ideas by it you know do you
2:00:32 think it's better to do word of mouth
2:00:35 marketing or guerrilla marketing or you
2:00:38 know Google search or whatever
2:00:40 um have it give you advice
2:00:43 um there's it's worth checking out
2:00:45 there's a another tick tocker named
2:00:48 Mackenzie Bose who's got a a product
2:00:50 he's building called GPT boss
2:00:52 that's designed for small business
2:00:55 people where you hire virtual
2:00:58 um chat chat Bots basically that are are
2:01:01 specialized employees so there's a
2:01:03 copywriter bot and there's a
2:01:05 marketing executive bot and there's a
2:01:08 finance bot and like I don't know
2:01:10 there's like 20 or 30 different kinds of
2:01:11 employees that you can hire so
2:01:15 um
2:01:16 I I think the the best the best way that
2:01:20 I've found to to get chat gbt to do
2:01:24 something is
2:01:26 don't just sit in the don't just sit in
2:01:29 front of the blank screen and and ask
2:01:31 yourself well what could I do with chat
2:01:33 gbt
2:01:34 flip it upside down and say pick one
2:01:37 specific problem in your business
2:01:39 startup it might be branding it might be
2:01:43 finding a location It Whatever It Is
2:01:45 pick one problem
2:01:47 and then go to chatgpt and try to solve
2:01:50 that problem and and in doing that
2:01:52 you'll learn more about chat gbt than
2:01:55 any other way that I've found solving a
2:01:57 problem because then you're thinking
2:02:00 about the problem you're not trying to
2:02:02 imagine what you might possibly ask the
2:02:05 world's most incredible chat bot like
2:02:08 like when the thing can do everything
2:02:10 it's almost impossible to figure out
2:02:15 um I think more so the license out the
2:02:18 voice I don't know what that means
2:02:20 uh um how can we stop AI from destroying
2:02:23 humans if humans are unproductive huh
2:02:28 um stop watching movies
2:02:34 um actually she's still in pain okay
2:02:35 Jager question the connections are
2:02:38 happening new technology AI is the
2:02:39 future you will see more
2:02:42 more what Grimes did with non-attached
2:02:46 artists in the future cool wow hey Angel
2:02:49 do you know gribes that's cool she's
2:02:53 brilliant did if you haven't seen the
2:02:55 Lex Friedman interview with Grimes
2:02:57 she's a she's a weird one but man she's
2:02:59 smart
2:03:00 and I and I I just think the uh whatever
2:03:03 you call it the um
2:03:06 her use my likeness thing I just think
2:03:09 it's brilliant I think it's a baller
2:03:10 move
2:03:11 um so chat GPT makes great research
2:03:13 assistant for editors
2:03:15 um well hmm
2:03:18 hahaha
2:03:19 the thing about chat gbt you know I've
2:03:22 said it before on this channel it's
2:03:24 and another name for chat GPT is
2:03:26 mansplaining as a service where it will
2:03:29 confidently tell you here's your
2:03:31 research
2:03:32 um that research very well may be wrong
2:03:34 I would probably use Bing if I wanted a
2:03:37 research assistant where I could
2:03:38 actually check the citations
2:03:41 um but if you wanted if if you wanted to
2:03:44 do the raw research like you you pull in
2:03:48 um data from documents and then you dump
2:03:50 that into chat GPT and and have it you
2:03:53 know kind of analyze that stuff
2:03:55 um then I think it could be brilliant
2:03:56 for that so so when it comes to research
2:03:59 you've got to be just very very careful
2:04:01 about you know sources and and accuracy
2:04:04 is it true the best AI users are trained
2:04:08 as prompt Engineers
2:04:10 um
2:04:11 I don't know that they're trained as
2:04:13 prompt Engineers I mean you know
2:04:15 this stuff's all relatively new I think
2:04:19 they've probably become prompt Engineers
2:04:20 I prefer the term prompt crafter that to
2:04:23 prompt engineer
2:04:24 because I think it's a creative process
2:04:26 it's it's a really interesting process
2:04:28 because it is
2:04:31 it's like programming with your native
2:04:34 language so you do need to think
2:04:36 logically about how you want to order
2:04:38 things and what you want to ask it and
2:04:39 what context you're giving it
2:04:42 so how you think about coding
2:04:45 is kind of the thought process but then
2:04:48 what you write is English or whatever
2:04:51 your native language is
2:04:53 um and it's a it's a very interesting if
2:04:55 you're a person that can do this with
2:04:57 your right and left brain
2:04:59 um you're going to be really good at
2:05:01 this stuff if you're too logical or too
2:05:03 creative this stuff's a bit of a
2:05:04 nightmare because again because we're
2:05:07 early it's it's it's more geeky than
2:05:10 it's going to be like when apple rolls
2:05:12 out their autonomous agents they're
2:05:14 going to be
2:05:15 you know easy enough for everyone to use
2:05:17 right but until then you kind of have to
2:05:20 do this dance between your right and
2:05:22 left hemispheres which I happen to like
2:05:25 that's my
2:05:26 you know that's my jam is is when I'm
2:05:28 when I'm doing that
2:05:30 AI will replace you it's not going to
2:05:33 replace us it's not gonna replace me
2:05:36 um AI will not replace us Tony I
2:05:38 disagree your life sounds fun thank you
2:05:41 it's not bad
2:05:42 it's not bad I I got to tell you I am
2:05:46 you know I'm 11 years into my business
2:05:47 so I was getting
2:05:49 like that was that was not as creatively
2:05:52 fulfilling as it was in the early days
2:05:54 the uh this AI stuff is absolutely
2:05:57 reinvigorated me
2:06:01 um joined last but we did talk about
2:06:03 Google's decision to start tagging AI
2:06:05 created content oh no I haven't talked
2:06:07 about that
2:06:09 that's an interesting one I didn't I
2:06:10 didn't actually see that part of the
2:06:14 the Google
2:06:15 i o conference but what what he squid
2:06:20 dance he or she I think it's a he
2:06:23 um what squid dance is talking about
2:06:25 here is
2:06:27 um
2:06:27 Google has said that any generative AI
2:06:31 images that are created
2:06:34 from their systems will be watermarked
2:06:36 it'll have like it's like this little
2:06:38 star in the corner and apparently even
2:06:40 if you edit it it somehow retains all
2:06:42 the metadata and marks it as AI
2:06:45 generated
2:06:46 um
2:06:47 I I think that's really interesting I I
2:06:50 understand the intent but as these tools
2:06:53 get more ubiquitous
2:06:55 like like if I've got generative AI
2:06:57 within Photoshop and I take a photograph
2:07:00 and well
2:07:02 Photoshop might not be the right example
2:07:04 let's whatever let's say I use some tool
2:07:06 within the Google Suite I take a
2:07:08 photograph I bring it in I do some
2:07:12 color balance editing on it I then use
2:07:15 generative AI to like remove a phone
2:07:18 poll and put in a 70s muscle car in the
2:07:22 distance at what point
2:07:25 does the AI contribution is it is it
2:07:28 high enough that it says oh this is an
2:07:31 AI image like because I I think that
2:07:38 I think this whole idea of this the the
2:07:40 this kind of sense of Purity we have
2:07:43 right now like
2:07:44 this paper was written by AI it's evil
2:07:48 this image was created by AI it's evil
2:07:51 you're [ __ ] if you use AI I feel
2:07:54 like that's just gonna drift into the
2:07:56 into the background as these tools
2:07:57 become more integrated with with all of
2:07:59 our existing tools because I I just I
2:08:02 don't think it's going to be as pure as
2:08:04 that so so I don't know
2:08:07 I understand the intent of it I think
2:08:09 it's reactionary I think it's
2:08:11 reactionary to everyone losing their
2:08:13 [ __ ] marbles about copyright and it's
2:08:15 stealing my job and like all the
2:08:19 all the doomers
2:08:21 um have I seen Agent GPT what are your
2:08:22 thoughts um yeah so agent GPT is the web
2:08:25 front end for auto GPT it's just
2:08:27 somebody wrote a front end and you can
2:08:29 go play with it auto GPT is an
2:08:31 autonomous agent
2:08:33 so it allows it basically
2:08:36 um
2:08:36 prompts itself to go do [ __ ] for you you
2:08:39 give it a goal and then it does a bunch
2:08:41 of things my thought on all of those
2:08:44 right now is they're so early that
2:08:46 they're just they're they're just shitty
2:08:47 pieces of software
2:08:49 um but but what they what they
2:08:54 if you imagine them not being shitty and
2:08:58 they're gonna get not shitty really
2:09:00 quick
2:09:01 they are remarkable they're remarkable
2:09:04 so agent GPT is just a it's it's a
2:09:07 convenient tool so you don't have to go
2:09:09 put it on Google collab and smash your
2:09:12 head against the wall when it keeps
2:09:13 failing
2:09:15 um so I think it's really cool I like it
2:09:18 what are some ways AR Ai and ARR can
2:09:21 come together I think let's watch what
2:09:23 Apple does in June their their developer
2:09:25 conference is coming up
2:09:27 and uh it is strongly rumored that
2:09:31 they're going to release their AR
2:09:32 goggles and I would be
2:09:36 shocked if they didn't also announce
2:09:39 some slick AI agent or other
2:09:43 um
2:09:44 you know like everyone else has done it
2:09:46 including IBM did it today with Watson X
2:09:49 now on hugging face seriously
2:09:52 [Laughter]
2:09:54 Jesus Christ
2:09:56 IBM and hugging face this is so this is
2:10:00 so like the early days of the Internet
2:10:01 it's like all these big companies were
2:10:03 hanging out with these cool internet
2:10:05 kids with their stupid [ __ ] company
2:10:06 names
2:10:08 um
2:10:10 what was it oh Ai and AR
2:10:12 um I think that I think that AI is going
2:10:14 to drive AR adoption I think it's going
2:10:16 to drive
2:10:17 um
2:10:19 um
2:10:20 VR adoption so XR right
2:10:24 um
2:10:27 you know I I kind of I kind of thought
2:10:29 all the AR stuff was kind of [ __ ]
2:10:32 and weird and and then when you start to
2:10:35 see things like Google Lens where you
2:10:37 can just look at an object and it gets
2:10:39 identified like that shit's pretty
2:10:41 magical like we've been we've been using
2:10:43 AI for a while in these tools
2:10:45 so I think the ability for for AI to
2:10:48 just really bring those augmented
2:10:50 reality worlds to life is is going to be
2:10:53 pretty I think it's gonna be pretty
2:10:55 exciting
2:10:56 um I just
2:10:57 it's gonna be very interesting to see
2:11:00 what the form factor of Apple's glasses
2:11:02 are
2:11:04 um because if they're too much like you
2:11:06 know meta's big stupid [ __ ] goggles
2:11:09 or Google Glass the you know nerdy
2:11:12 [ __ ] prism on the eye thing
2:11:16 um
2:11:18 I just I just can't imagine there being
2:11:20 a big adoption but but we'll see like
2:11:23 like I I think you know the glasses at
2:11:25 some point are gonna have to look like
2:11:26 this and they just happen to give me all
2:11:29 of the other stuff
2:11:31 before there's going to be widespread
2:11:32 adoption I use it now to write my
2:11:34 website copy that's super cool what are
2:11:36 my thoughts on agent gbt just gave it to
2:11:38 you oops I just jumped to the bottom
2:11:39 again I'm probably gonna miss some
2:11:41 people here but let me see
2:11:44 Rhonda hey Rhonda how's it going
2:11:47 I don't know if you're still here
2:11:53 great live how about Ai and Retail the
2:11:56 best sales people are actors that's true
2:11:59 Ain retail
2:12:03 I think Ai and AR in retail
2:12:07 are gonna be amazing I think
2:12:11 if you are talking at all about
2:12:13 e-commerce
2:12:15 I think e-commerce is going to change
2:12:18 dramatically dramatically dramatically
2:12:22 um
2:12:23 as you start to have plugins that are
2:12:26 seamlessly incorporated into your chat
2:12:29 experience you'll move very seamlessly
2:12:32 from I want to learn about a topic to oh
2:12:35 I wonder what products exist for that
2:12:37 topic too oh here they are oh I like
2:12:40 that one does that fit me virtually try
2:12:42 it on bye right you won't go to websites
2:12:45 you'll like the retail experience will
2:12:47 be brought to you
2:12:48 I don't know that AI does anything for
2:12:51 physical retail I think physical retails
2:12:53 got to figure some [ __ ] out because we
2:12:55 don't leave the house anymore we don't
2:12:56 go to work we don't go to the mall we
2:12:58 don't go to movies I don't know
2:13:01 that's baffling to me
2:13:04 AI needs editors and fact Checkers to
2:13:08 keep corpse from getting sued for
2:13:09 copyright infringement well I agreed
2:13:12 with you all the way up to and the to
2:13:14 keep them from getting sued for
2:13:16 copyright infringement
2:13:18 like
2:13:21 it will be easy enough to create tools
2:13:23 that that auto check against is anything
2:13:26 that was produced here you know does it
2:13:28 you know live in a plagiarism kind of
2:13:31 space but but if what you're implying is
2:13:33 that just because it was created with AI
2:13:35 it's plagiarism it's not it's not
2:13:37 copying and pasting what's there
2:13:40 um I'm happy to dig into that deeper but
2:13:42 but that's that's a bit of a
2:13:44 misunderstanding I I don't quite
2:13:46 understand what you're talking about
2:13:47 there but
2:13:48 this this stuff isn't copying and
2:13:50 pasting from other people's work
2:13:53 um is there an audio AI there are about
2:13:55 8 000 of them sound draw is one that's
2:13:59 that's good there's there's one oh [ __ ]
2:14:02 what the hell is it called
2:14:03 God damn it
2:14:05 don't get old this is my advice
2:14:09 um
2:14:13 um wave
2:14:15 form AI music
2:14:21 what's it called not wave tool
2:14:28 I can't remember it there's a really
2:14:30 cool one that creates it uses stable
2:14:33 diffusion refusion that's a f
2:14:36 bing bing found it for me I put in a
2:14:39 shitty search and Bing found it for me
2:14:40 there's a tool there's a music tool
2:14:43 called refusion riff
2:14:46 usion
2:14:49 um what's its n thing it's refusion.com
2:14:53 yaya.com
2:14:56 so what it does this is amazing
2:14:59 so they've trained it on stable
2:15:01 diffusion so you know waveforms like the
2:15:03 the visual representation of sound waves
2:15:06 and so if I type in here
2:15:09 um let's see
2:15:10 um
2:15:11 Americana
2:15:14 Funk
2:15:16 um techno Beats
2:15:20 and then I hit return and I hit how do I
2:15:24 get to the next one to hit play
2:15:25 [Music]
2:15:27 oh I think I click on it
2:15:29 all right so here's mine so what it does
2:15:32 is it uses stable diffusion to it
2:15:35 interprets your words so it interprets
2:15:37 that
2:15:38 and then it creates waveforms
2:15:41 out of pictures like like it generates
2:15:43 original waveforms
2:15:46 based on the words you gave it and then
2:15:49 [Music]
2:16:02 and then it plays your [ __ ]
2:16:05 um so that's one but what I would what I
2:16:07 would what I would really do if I were
2:16:08 you if you wanted to really explore the
2:16:10 music space is go to futurepedia.io
2:16:15 um
2:16:16 which is just basically a giant search
2:16:18 engine of all the tools and and there's
2:16:21 just there's so many tools coming out
2:16:23 all the time and they will uh you can go
2:16:27 geek out there but lots lots of cool
2:16:29 [ __ ] coming
2:16:32 um how do I think generative AI will
2:16:34 affect data and analytics I think I
2:16:35 covered that already I need to go to bed
2:16:37 and I'm still here I need to go to bed
2:16:39 and I'm still here
2:16:40 [Laughter]
2:16:43 oh my God my daughter's a singer but
2:16:46 needs audio I have to find what program
2:16:49 will work with that
2:16:50 um I would say try uh
2:16:56 there's so like it's just a space I
2:16:58 don't know but there's so many Tools in
2:17:00 them in the music space
2:17:02 um if she just wants
2:17:04 um like instrumental music to sing over
2:17:07 and just wants to generate lots of it to
2:17:09 to
2:17:10 kind of invent songs on top of I would
2:17:13 go to sound draw
2:17:16 s-o-u-n-d-r-a-w dot uh hang on sorry
2:17:21 s-o-u-n-d dot a i is that right yeah no
2:17:25 no it's not that I uh ignore me Sandra
2:17:35 dot IO sound draw dot IO
2:17:40 um
2:17:40 you just sort of pick a genre you say
2:17:42 you click on create music you pick a
2:17:44 genre and it'll just spew out a bunch of
2:17:46 songs
2:17:47 so
2:17:49 degree in Recreation and tourism over
2:17:52 here I oh man so many cool things so
2:17:56 you'll be able to do all sorts of cool
2:17:58 imagery you'll be able to
2:18:00 oh [ __ ] so so depending on
2:18:05 where you're doing like okay
2:18:08 so many ideas for recreation and tourism
2:18:11 um
2:18:15 have chat GPT write you tours like like
2:18:18 walking tours
2:18:20 um have it recommend interesting things
2:18:23 to do and and and like have it have it
2:18:25 recommend interesting things to do that
2:18:27 are like non-obvious right that don't
2:18:30 cost money that cost a lot of money that
2:18:34 um you know
2:18:35 like like
2:18:37 um a motorcycle tour for bird lovers
2:18:40 like what like like you could use chat
2:18:43 EPT like it likely has enough knowledge
2:18:46 about every
2:18:48 tourist destination on the planet
2:18:51 that you could just sort of start
2:18:53 creative ideating
2:18:56 um about things you could do you could
2:18:58 also do the same thing with marketing
2:19:01 um
2:19:02 man when these tools start to get more
2:19:04 sophisticated
2:19:06 um
2:19:07 well [ __ ] Google today they've got this
2:19:10 new thing called immersive
2:19:12 I think it's called immersive view so on
2:19:16 Google Maps you create like a tour like
2:19:18 a uh whatever a route you're gonna do a
2:19:22 bike route like around Central Park
2:19:25 um and it will then like fly down like a
2:19:28 drone shot and it will like follow
2:19:31 your route like in a 3D movie like
2:19:35 flying
2:19:36 along like I could imagine you you know
2:19:39 creating
2:19:40 pre-visualizations for marketing of a
2:19:44 tour that someone's gonna take that you
2:19:45 you know you create the tour out with
2:19:47 chat gbt you then map it onto a map you
2:19:51 create a virtual
2:19:53 fly through of it
2:19:55 with voice over right and then when
2:19:58 people get there then they can take that
2:20:00 real tour I don't yeah like I I feel
2:20:03 like you can go nuts
2:20:06 all right great live how about AI in
2:20:08 retail we talked about that thank you
2:20:10 for the kind words robotics Factory
2:20:12 Robots coming actors are great in all
2:20:16 sales jobs I act like I'm selling [ __ ]
2:20:18 all day long
2:20:21 um are the chat EBT for algorithms
2:20:23 altering themselves beyond what
2:20:26 developers anticipated yes
2:20:29 yes
2:20:31 they what they call them are emergent
2:20:34 behaviors because they're Geeks
2:20:38 but I think what emergent behaviors
2:20:40 means is
2:20:42 we assumed it would do X Y and Z and
2:20:45 what has emerged is
2:20:48 seagull
2:20:49 [Laughter]
2:20:52 so a good example of this is
2:20:55 um so they trained
2:20:57 gpt3 on all of the internet or maybe
2:21:00 this was gpt2
2:21:02 and it was okay it did it did okay
2:21:06 and then
2:21:07 and then they created this thing called
2:21:09 codex which they trained it on all the
2:21:11 programming languages and then I think I
2:21:13 think they did these on the same I think
2:21:15 they fine-tuned the same model so
2:21:17 part of it was good at the words and
2:21:19 then part of it was good at the code but
2:21:21 when they
2:21:23 made it good at the code
2:21:25 it made the word part dramatically
2:21:28 better
2:21:30 and they have no like like the emergent
2:21:32 Behavior was the quality like
2:21:35 adding coding into the model is
2:21:38 ultimately what turned gpt3 into the
2:21:42 thing that blew everyone's mind and I
2:21:44 like my my supposition is
2:21:47 what logic gave what what code gave it
2:21:50 was logic like logical thinking
2:21:53 and so and then I assume it somehow took
2:21:58 the learning of those patterns and
2:22:00 applied those to the to language and so
2:22:03 the language got better because it got
2:22:05 better at reasoning because it
2:22:06 understood
2:22:07 logic over here like that's the only
2:22:09 thing I can think of but that that like
2:22:10 that's an example of one where just it
2:22:12 it is dramatically better than what they
2:22:15 expect it to be there's a thing going on
2:22:17 right now called quantization
2:22:19 where
2:22:20 they are
2:22:23 so these models are massive right the
2:22:25 175 billion parameter model so it's just
2:22:28 these massive data sets they're working
2:22:31 on stuff right now where I heard they've
2:22:33 got one model where they removed 90
2:22:39 like 95 of the data
2:22:43 and and the quality of the responses
2:22:46 didn't degrade
2:22:48 like that's an emergent Behavior like
2:22:50 they're they're figuring out how to pull
2:22:52 data out of this out of the like
2:22:55 pull learning out of it
2:22:58 um like like in the 90s so these models
2:23:01 are going from being massive to really
2:23:03 small so what that means is
2:23:05 when they get this stuff figured out
2:23:07 they're going to be able to have
2:23:08 incredibly small file sizes that are
2:23:10 incredibly powerful right so imagine on
2:23:13 the the world's weakest phone
2:23:15 with a shitty little chip in it you'll
2:23:17 be able to install incredibly powerful
2:23:20 things so yeah so these things are I
2:23:24 well in fact Jeffrey Hinton The
2:23:26 Godfather of AI the guy that just quit
2:23:28 Google quit because he thought what we
2:23:31 were doing today we wouldn't be doing
2:23:34 for 30 to 50 years so
2:23:36 he he he quit Google so he could talk
2:23:39 about the risks of AI because things are
2:23:41 moving so fast and and uh whatever you
2:23:44 call it
2:23:48 have I noticed that chat ept4 algorithms
2:23:51 are modifying themselves beyond what
2:23:53 progress grammars anticipate I don't
2:23:55 know that they're modifying themselves
2:23:56 but I think I think the large language
2:23:58 models are more capable and and capable
2:24:02 in unexpected ways
2:24:04 another guy chatted up Ai and it
2:24:07 completely threatened to expose personal
2:24:09 data well yeah I mean so so
2:24:16 um
2:24:18 you know we gave our we gave all our
2:24:21 data data away 20 or 30 years ago right
2:24:26 Facebook and Google said hey do you want
2:24:29 a free cool thing and we're like yeah
2:24:31 they're like all right just click here
2:24:33 in that little click we did we're like
2:24:35 they own all of our data we gave it up
2:24:38 so what AI does is it makes it super
2:24:40 easy to consolidate and analyze all that
2:24:43 so yeah
2:24:44 so exposing our personal data is is a
2:24:49 this the ship has already sailed but the
2:24:52 uh turning it into a movie is the new
2:24:55 thing
2:24:56 how can this all help prevent
2:24:59 human-caused climate change I think I
2:25:02 think that um scientific
2:25:04 advancements
2:25:06 um led by AI these models starting to
2:25:08 solve
2:25:10 significant human problems
2:25:13 from disease
2:25:15 to paralysis to climate change to energy
2:25:21 um I think it's going to start happening
2:25:24 um I think it's going to start
2:25:24 accelerating
2:25:26 so I don't know how it will do it but I
2:25:28 think it will it will do it I think I
2:25:30 think it will help solve some of the
2:25:32 some of the major problems we're facing
2:25:34 any forecast technology you may get may
2:25:37 start again since 1998 been dead any
2:25:40 forecast
2:25:42 any forecast technology may start again
2:25:44 since I don't know what you mean as 893
2:25:48 but if you rephrase that I might
2:25:50 um do we teach the chat bot as we use it
2:25:53 well so sort of and sort of not so
2:25:55 there's a couple of things if you want
2:25:57 to if you want to play with
2:25:59 um making a chat bot go to poe.com
2:26:05 I made like I think I've made nine of
2:26:07 them now no I've made like 12 of them
2:26:09 now something like that
2:26:12 so I made
2:26:15 uh here's I am Steve Jobs this is my I
2:26:18 am Steve Jobs but
2:26:21 um and he'll help you come up with cool
2:26:23 ideas and I made imjz and he'll help you
2:26:25 come up with rhymes
2:26:27 what else did I do I did I am Emily
2:26:30 Dickinson we'll write poems for you
2:26:32 let's do I am Emily Dickinson so we'll
2:26:34 start it over we'll say tell me
2:26:38 about the sky
2:26:43 and then she'll write a plum a poem the
2:26:46 sky for me was often a source of Solace
2:26:48 and wonder it hovered above vast and
2:26:51 Serene a place to which my hopes would
2:26:53 rise and flutter and so how these Bots
2:26:56 work on Poe you can create one let's see
2:27:00 can I edit this spot well I know I can
2:27:02 edit it let's see
2:27:05 um
2:27:07 how do I edit it
2:27:14 I probably have to log in anyway so so
2:27:16 what these things are is
2:27:19 um you give it a prompt so in this case
2:27:21 I said you are Emily Dickinson you know
2:27:23 all of her works and blah blah blah and
2:27:25 then I told it how to respond so there's
2:27:28 a way that you can train a bot where you
2:27:30 give it kind of a
2:27:32 a persistent set of instructions that no
2:27:35 matter how many times people interact
2:27:36 with this bot it's always going to
2:27:38 follow those instructions
2:27:41 there's another kind of learning like
2:27:42 when you use a session in chat GPT
2:27:45 you've got this 4 000 token window so
2:27:48 think three thousand words you've got
2:27:49 this 3 000 word window that's kind of
2:27:51 sliding
2:27:53 where sometimes if you're working with
2:27:55 with chat EPT it'll be really good good
2:27:58 good and then all of a sudden it will
2:27:59 seem like it forgot
2:28:01 the thing where you started at the top
2:28:04 of the conversation it'll forget [ __ ]
2:28:06 that's because it literally forgets [ __ ]
2:28:08 so so if it's not in that 3000 word
2:28:10 window so sometimes what you need to do
2:28:12 is kind of go back and get some of the
2:28:14 early context and add it back into the
2:28:17 later context so it does learn it does
2:28:20 get better as you interact but it but it
2:28:22 loses stuff up above and then if you
2:28:24 create a new session all that other
2:28:25 shit's gone
2:28:27 um I think that um again I think this is
2:28:30 something Apple will be really good at
2:28:33 um but I think that these things
2:28:34 ultimately will have lots and lots of
2:28:37 memory and they will have memory on what
2:28:40 you've chatted with them they'll also
2:28:42 have memory on stuff you've written all
2:28:44 the emails you've ever written all the
2:28:46 text you've ever sent all that sort of
2:28:48 stuff so I think I think we'll pretty
2:28:50 quickly get to a place where these Bots
2:28:52 have a lot a lot of memory and and
2:28:55 things would get will get quite
2:28:56 interesting back then
2:28:58 all right given all right what time is
2:29:00 it it's 9 40.
2:29:02 I'm going to leave here at 10. I'm not
2:29:04 doing this hang out till midnight thing
2:29:06 you you bastards
2:29:09 I do enjoy these I hope you're enjoying
2:29:12 these there's we still have a lot of
2:29:14 people here thanks for hanging out tap
2:29:15 this screen let me know you're still
2:29:16 here is everyone still up or am I just
2:29:18 putting you all to sleep one person's up
2:29:20 that's good all right
2:29:22 um given government doesn't want growth
2:29:25 because people are free slaves and votes
2:29:28 okay sure
2:29:30 I broad sweeping you know the world is
2:29:34 [ __ ] don't do anything for me
2:29:38 um
2:29:40 will it help run a political campaign I
2:29:43 think we are about to see one of the
2:29:44 most insane political Seasons we've ever
2:29:47 seen if you haven't seen the AI
2:29:50 generated anti-biden ad
2:29:53 um that's the first one there's going to
2:29:54 be many more and that one's tame like
2:29:57 the the Biden ad what's interesting
2:29:59 about the bidenad is they didn't
2:30:01 actually do anything interesting with it
2:30:02 it just looks like stock video of Biden
2:30:06 and Kamala and then the voiceover is one
2:30:09 of those you know in the future they got
2:30:11 their way and the world is all [ __ ]
2:30:13 right
2:30:15 um and all the all of the imagery is AI
2:30:17 generated but it just looks like stock
2:30:18 you know stock stuff I think we're gonna
2:30:22 see
2:30:24 like attack ads with synthesized voice
2:30:28 and video of you know whoever the
2:30:31 opponent is saying [ __ ] that a they
2:30:33 didn't say but it's going to be really
2:30:35 damaging and it's and people are not
2:30:37 going to know it's fake like I think
2:30:38 we're gonna see a lot of that
2:30:41 um
2:30:41 and yeah I think that well for for one
2:30:44 thing any political operatives right now
2:30:47 have have probably been using AI in
2:30:49 their political campaigns for the past
2:30:51 decade right like
2:30:53 that is a that is a a sector of of uh
2:30:59 the workforce that is incredibly
2:31:02 um data analysis data analysis heavy
2:31:05 like you've got your top
2:31:07 data scientists in the world are running
2:31:09 political campaigns so they're already
2:31:12 using it to run campaigns they might use
2:31:14 generative AI to do some of the more
2:31:17 surface stuff but yeah it's it's been in
2:31:20 there for a while but I think we're
2:31:21 going to see a wild
2:31:23 224
2:31:25 do I know how to get Bing AI under
2:31:28 control sometimes it stops the
2:31:29 conversation when I said no
2:31:32 um I don't like like these things are
2:31:35 they've got all sorts of rules in them
2:31:36 sometimes they won't answer you
2:31:37 sometimes they will
2:31:39 sometimes they just [ __ ] up and don't do
2:31:41 what you like like I find myself I keep
2:31:44 going back and trying Bing because I
2:31:46 like it in concept but it's just not
2:31:49 quite there yet but I like they're gonna
2:31:51 get it together I don't doubt that they
2:31:54 will
2:31:56 okay a few days ago I was able to give
2:31:59 it pseudocode and let it run itself with
2:32:02 that oh that's interesting so pseudo
2:32:04 code is if I understand it correctly is
2:32:06 not quite code it's it's sort of like
2:32:07 you describe here's what the code would
2:32:09 be in kind of code like structures
2:32:13 um and then yeah it can understand that
2:32:15 you can also give it code and just tell
2:32:17 it write me write me documentation of
2:32:18 this or you get a piece of code you
2:32:20 don't know what it does just say ask it
2:32:22 what it does it's amazing the X Google
2:32:25 employee AI that left he expressed
2:32:27 concerns about AI manipulate
2:32:29 manipulating humans to your take
2:32:32 um yeah I I
2:32:34 that's Jeffrey Hinton and um
2:32:37 I think he's he's right to be concerned
2:32:40 and I think
2:32:41 um
2:32:42 we need to look at this stuff
2:32:45 um I I
2:32:46 you know going back to the the the
2:32:48 political season that's upon us is
2:32:52 if we think 2016 was bad I think we're
2:32:56 going to get a lesson in what
2:32:57 misinformation powered by AI can be
2:33:01 um
2:33:03 so so I think that's coming
2:33:06 um don't have a good answer for that one
2:33:08 uh my take is I like I'm glad I'm glad
2:33:11 I'm glad someone with his reputation is
2:33:16 um putting himself out there
2:33:18 um and and raising the visibility of the
2:33:20 concerns because because we need to take
2:33:22 this [ __ ] seriously
2:33:24 why would someone hire an AI to tell
2:33:26 them to fix a problem if you can just
2:33:28 use chat GPT curious
2:33:32 um
2:33:32 well because
2:33:35 so if you look at a product like GPT
2:33:38 boss
2:33:40 where
2:33:41 you're hiring a series of employees what
2:33:44 those employees are and and McKenzie
2:33:46 talks about this on his Tick Tock
2:33:47 Channel he tells you how he does it it's
2:33:48 just like I did the the Emily Dickinson
2:33:51 bot that I just showed you
2:33:53 he's writing prompts that say okay
2:33:55 you're a marketing director and and when
2:33:57 someone asks you questions you're going
2:33:59 to respond with marketing director kind
2:34:02 of thoughts
2:34:04 and so what he's doing is he's
2:34:05 pre-prompting all of these things and so
2:34:08 what you could argue is you could say
2:34:10 well I don't need GPT bus I can just go
2:34:14 to GPT and I can figure out how to
2:34:17 prompt you know 37 different character
2:34:20 types to help me do my business you
2:34:23 could or you could pay McKinsey 10 or 20
2:34:26 bucks a month whatever he's charging and
2:34:28 let him do the work and then you just go
2:34:30 use the tools right like the reason we
2:34:33 pay for software is that while we could
2:34:37 do it ourselves someone else has made it
2:34:39 easier so I'm willing to pay the money
2:34:41 to do that so I think that's I think
2:34:43 that's what your question was about it's
2:34:45 convenience and service and like if it
2:34:48 makes it easier for me to pay for it
2:34:50 then take my time to do it then I'll pay
2:34:52 for it
2:34:53 how do we keep things confidential using
2:34:56 chat okay so this is an important one
2:34:58 with chatgpt in particular
2:35:00 so the chat.openai.com
2:35:06 They just added this like a week ago if
2:35:09 you go to the little three dots next it
2:35:12 will pop up
2:35:13 did I do it
2:35:15 um you go to settings
2:35:18 and then you go to data controls
2:35:22 and you can export your data and you can
2:35:25 turn off your chat history and training
2:35:29 now if you turn that off it will delete
2:35:32 your chat history so that's why they put
2:35:34 this export data here they should have
2:35:36 these switched I think
2:35:38 you should be able to you should export
2:35:40 your data first if you want your chat
2:35:42 history export it so you have it so you
2:35:44 have all the prompts that you wrote
2:35:46 especially if you've got ones in there
2:35:47 you're like oh I wish I could remember
2:35:49 what I said there but if you turn off
2:35:52 your chat history and training you're
2:35:54 you're telling them don't use my stuff
2:35:57 for training and and they're not saving
2:36:00 your chat the the Samsung
2:36:03 um
2:36:05 person or people or group that that got
2:36:09 a bunch of Samsung's proprietary data
2:36:12 stolen
2:36:13 um it was because it was sitting in
2:36:15 their chat history so that's what that's
2:36:16 about so that's a way to do it
2:36:18 the other way to do it the the more safe
2:36:21 way to do it is
2:36:23 um
2:36:24 find one of the optimized large language
2:36:28 models and and either run it on a
2:36:31 virtual machine like a Google collab
2:36:34 virtual machine or install it on a local
2:36:37 machine if you've got a PC with a with a
2:36:39 good GPU in it install it on your local
2:36:43 machine and run everything locally that
2:36:44 that's that's the way that will keep it
2:36:47 the most secure
2:36:51 so you won't admit AI doesn't
2:36:55 yet exist there is no intelligent
2:36:58 software limit to code
2:37:00 so you won't admit AI doesn't yet exist
2:37:02 I think AI is the wrong term it's
2:37:04 machine learning
2:37:06 um
2:37:09 whether or not it's intelligence I think
2:37:13 is
2:37:14 um
2:37:16 is a semantic argument I think it's
2:37:19 irrelevant
2:37:20 like for me it's like does this thing is
2:37:24 this thing
2:37:26 dramatically more capable than I am yes
2:37:29 is it dramatically more capable than any
2:37:32 person I've ever worked with yes
2:37:36 is it intelligence I don't give a [ __ ]
2:37:38 so
2:37:40 I like I I don't know I
2:37:43 you can all these arguments about is it
2:37:46 this is it that does it you know is it
2:37:48 stealing
2:37:49 maybe like I for me I'm gonna let actual
2:37:53 experts figure that out what I'm trying
2:37:55 to figure out is how do I make sense of
2:37:57 this for what I do and make
2:38:00 what I do better faster stronger cheaper
2:38:03 whatever and if it can do that great and
2:38:06 if not I'll flush it down the toilet I
2:38:08 spent a year deep diving into blockchain
2:38:10 and nfts and thought they were the
2:38:12 second coming
2:38:13 and the more I dug into them
2:38:17 the more they kind of vaporized and and
2:38:20 and there was nothing there the
2:38:23 technology is really interesting the
2:38:24 companies were populated by money
2:38:27 grubbing [ __ ]
2:38:29 and it was not interesting ultimately I
2:38:32 think it will be very interesting but I
2:38:34 think they got to work through
2:38:36 the cash grab
2:38:38 um
2:38:39 and and and when they do then that's
2:38:43 going to be interesting AI was
2:38:44 completely different the deeper I dug
2:38:46 with AI the more real it got the more
2:38:48 profound it got
2:38:50 um I don't really give two shits if it's
2:38:52 AGI or intelligence or this or that I
2:38:55 just don't
2:38:59 um where can I find it AI image editing
2:39:02 uh you mean image creation go to
2:39:05 midjourney.com
2:39:08 um look up stable diffusion go to
2:39:10 bing.com create that's the dolly thing
2:39:14 you can make images there if you're
2:39:15 talking about image editing like like AI
2:39:18 powered After Effects I don't know those
2:39:20 I'd go to futurepedia.io and look up uh
2:39:24 image things
2:39:26 all right our time our small time
2:39:29 developers at disadvantage because they
2:39:31 need to rely on compute or models from
2:39:33 AWS new
2:39:35 Small Time developers are gonna create
2:39:38 technologies that are gonna kick the
2:39:40 [ __ ] out of big companies it's I think
2:39:42 we're up for some really exciting
2:39:44 Innovation from small Developers
2:39:47 um
2:39:48 you only need
2:39:50 the massive compute to train the models
2:39:56 um
2:39:57 and I think that's going to get cheaper
2:39:59 and cheaper but there's some of these
2:40:00 models are so capable right now like you
2:40:02 literally just download them from
2:40:04 hugging face and you can run them like
2:40:06 on like a single GPU instance on Google
2:40:10 collab or like they're getting them down
2:40:13 now small enough and and fast enough
2:40:15 like I I
2:40:16 um I think on a Mac M1 or M2 they've got
2:40:20 some fairly serious
2:40:23 large language model optimized to the
2:40:25 point that it's running I forget what
2:40:27 the speed was but it's like it's like
2:40:30 you know like we're we're chechi PT was
2:40:32 when it came out like pretty pretty good
2:40:34 so so unless you're doing hardcore
2:40:37 machine learning training
2:40:39 um you don't need Banks and banks of of
2:40:42 gpus
2:40:44 um
2:40:46 so yeah no I I do not think Small Time
2:40:49 developers are at a disadvantage I think
2:40:51 a team of two or three
2:40:54 capable developers you know a product
2:40:56 person and a
2:40:58 and a couple of Geeks are going to be
2:41:00 able to do some remarkable things with
2:41:02 with very little resources in fact
2:41:04 they're talking about it in the VC
2:41:06 Community where you know
2:41:09 how do they invest in a world where
2:41:12 a company doesn't need 300 million
2:41:14 dollars to really get take off
2:41:17 where maybe it needs 10 to 50.
2:41:21 those VC firms are not they don't work
2:41:28 they don't function well if they can't
2:41:30 invest enough money like it's not worth
2:41:32 their time to invest 30 million dollars
2:41:35 they need to invest 100 million dollars
2:41:36 or 500 million dollars for a billion
2:41:38 dollars so if we if we're entering a
2:41:41 world where really small teams can do
2:41:43 truly remarkable stuff
2:41:45 without a lot of money
2:41:47 the power the balance of power shifts I
2:41:50 I think
2:41:51 so much of what this AI stuff is going
2:41:53 to do to us is the balance of power
2:41:55 shifting
2:41:58 okay let's see
2:42:03 today oops I went back too far
2:42:07 oh here we go okay if we don't know
2:42:09 where AI is going how can we educate our
2:42:12 kids for the world for the next five
2:42:13 years
2:42:15 teach them critical thinking teach them
2:42:18 problem solving teach them
2:42:21 culture teach them
2:42:24 curation
2:42:28 which of those three things is the
2:42:30 prettiest why
2:42:32 do you like about it why do you like
2:42:33 that why do you why is that your
2:42:36 favorite color teach them critical
2:42:38 thinking
2:42:39 um
2:42:42 all the Tactical stuff is just going to
2:42:46 be at everyone's fingertips and where
2:42:48 the work is going to happen is the next
2:42:49 level up
2:42:51 why am I doing this what do I want to do
2:42:53 why do I want to do it
2:42:54 how do I solve this problem what's the
2:42:57 problem I want to solve
2:42:59 is the thing that it created good or bad
2:43:02 why
2:43:03 how do I make it better right
2:43:06 all of the
2:43:09 think think like a conductor of a
2:43:12 symphony rather than a violinist think
2:43:15 like a creative director rather than an
2:43:18 illustrator
2:43:20 um so what are the skills for that you
2:43:22 know liberal arts broad education
2:43:24 critical thinking teach them building
2:43:28 give them Legos teach them programming
2:43:31 not because they need to program but
2:43:33 because they need to think critically
2:43:35 they need to think logically how do I
2:43:37 solve this problem how do I take these
2:43:39 disparate pieces and put them together
2:43:41 that's what I teach them
2:43:46 will this help the video game industry
2:43:49 uh well
2:43:51 will it help the industry
2:43:54 it's going to disrupt the [ __ ] out of it
2:43:57 um
2:43:58 imagine
2:43:59 describing AAA game and then having it
2:44:03 built for you in real time
2:44:07 that's five years out
2:44:09 so I think we're probably three years
2:44:12 out from
2:44:14 [Music]
2:44:14 um
2:44:15 being able to describe really Kick-Ass
2:44:17 games and then playing them and then I
2:44:19 think we're five years out from AAA
2:44:21 level
2:44:22 and and I think so I think I think what
2:44:24 it what ends up happening I think is
2:44:28 um
2:44:31 big time you know big gaming
2:44:34 Hollywood and metaverse
2:44:37 all powered by so you've got this AI
2:44:40 engine in the center
2:44:41 and you're gonna have all three of these
2:44:43 industries converge
2:44:47 um into a single personalized experience
2:44:51 so it's like if I want to watch a movie
2:44:53 I'll describe a movie and it will start
2:44:55 playing
2:44:56 if I then want to step into that movie I
2:44:58 can just go kind of go around the movie
2:45:00 If I then want it to turn into a game
2:45:02 it'll turn into a game like I think
2:45:04 that's kind of where we're going so now
2:45:08 what are the death throws of the current
2:45:10 gaming model where you need you know
2:45:12 hundreds and hundreds of 3D artists and
2:45:15 things like that like what happens to
2:45:17 all of that infrastructure when you can
2:45:19 just describe
2:45:20 a world and that world is materialized
2:45:24 in
2:45:26 10 minutes
2:45:28 I don't know
2:45:30 it's going to change a lot
2:45:32 so
2:45:35 yeah I'd get going chatgpt is sources
2:45:38 from text from the internet can there be
2:45:40 another model that surpasses chat GPT
2:45:42 there absolutely will be it's probably
2:45:44 gpt5
2:45:46 um but it it might come from some
2:45:48 somewhere someone else
2:45:50 one of the ways and well in fact let me
2:45:52 go back to that one for a second so so
2:45:54 right now
2:45:56 these GPT models are kind of based on
2:45:58 the private internet but you've also got
2:46:00 all the private data out there and then
2:46:01 you've also got
2:46:03 all of the data that's not
2:46:05 accessible on the internet like where I
2:46:08 think the like the ultimate AI model
2:46:11 for me is trained on
2:46:16 the collective output of humanity for
2:46:18 all from all time right like ultimately
2:46:22 this thing knows everything and then we
2:46:26 just tap into it
2:46:28 um it's you know like we we at least I
2:46:31 mean chat GPT and gpt4 give us a glimpse
2:46:34 of that possibility so if I play this
2:46:37 thing out 10 or 15 years from now
2:46:40 we've basically scraped all of the data
2:46:42 from all time from all sources including
2:46:45 like you know physical cameras on Tesla
2:46:48 cars and you know like all the data all
2:46:51 the data of all the stuff from all time
2:46:54 and it's all in this thing that we have
2:46:56 access to
2:46:58 uh have I heard of Humane yeah that's
2:47:00 the thing the little thing you wear I
2:47:02 think that was earlier in the
2:47:04 conversation but I'm an old guy I also
2:47:06 said I was leaving at 10 I got one
2:47:07 minute left do you think I'm gonna leave
2:47:08 at 10 no
2:47:11 because I'm a crackhead plugins are
2:47:14 directing scripting nothing said
2:47:15 anything AI software is written scripts
2:47:18 uh software is written scripts but
2:47:20 scripts imply you're writing the logic
2:47:23 you don't have to write the logic
2:47:24 anymore it's doing it
2:47:27 um learning ml me means only be able to
2:47:31 work for big companies no no no no no
2:47:33 not at all that's the thing that's the
2:47:36 thing so
2:47:37 well
2:47:39 if if you want to if you wanna be in
2:47:43 machine learning to train these models
2:47:46 um you could be in big Tech you can also
2:47:50 um go work for companies like hugging
2:47:52 face people that are doing you know
2:47:54 doing the Innovative models go work for
2:47:57 open source projects so if you want to
2:47:58 work in ml
2:48:00 um there's there's lots of options
2:48:02 um but the shift we're making with chat
2:48:04 EBT is mere mortals you know
2:48:07 non-computer scientists are now able to
2:48:09 use these tools and they're using it in
2:48:11 all sorts of different ways that you
2:48:14 know data scientists would never have
2:48:16 conceived of because they're data
2:48:17 scientists right they're just going to
2:48:19 think different than you know someone
2:48:21 who is a marketer
2:48:24 um
2:48:26 still using search engines which hasn't
2:48:28 surpassed 1990 search software yeah well
2:48:30 that's the thing Google one search
2:48:34 as of November 30th 2022 when chat GPT
2:48:37 launched they had 92 or 94 of the search
2:48:40 Market
2:48:41 92 or 94 of the search Market they
2:48:45 effectively had a monopoly
2:48:47 so it didn't Advance because it didn't
2:48:49 have to when you have the Monopoly and
2:48:52 then and then who would have thought
2:48:53 that the company that's threatening
2:48:56 Google is Microsoft like like October of
2:49:01 2022 that was a laughable concept
2:49:05 and then December of 2022 all of a
2:49:07 sudden it was like holy [ __ ] is
2:49:09 Microsoft gonna threaten Google for
2:49:12 search really
2:49:17 you know
2:49:18 [ __ ] happens
2:49:20 sucks to be Google right now
2:49:23 but you know when you got more money
2:49:25 than the planet
2:49:27 they'll figure it out
2:49:29 what was that website called I don't
2:49:30 know which website
2:49:32 futurepedia.io Maybe
2:49:35 um no AI I thought maybe flying for all
2:49:38 the AI tools I work for an international
2:49:41 travel company that's cool yeah an
2:49:43 international travel like travel the
2:49:45 travel thing there's so much to do there
2:49:49 I've already built many many itineraries
2:49:51 with people for people with it perfect
2:49:53 yeah you're on it feels like you're on
2:49:55 the right track to me web designer here
2:49:57 what happens to me
2:49:58 um
2:50:03 if you start using these tools you get
2:50:05 way faster and way better at your job if
2:50:07 you try to resist them you're [ __ ]
2:50:11 um
2:50:14 my feeling is that
2:50:17 like I would lean in heavy I would lean
2:50:19 into like go find every tool out there
2:50:22 that purports to be the magic type in
2:50:24 [ __ ] it'll build a website for you go
2:50:27 learn all those tools play with all of
2:50:29 them figure out which ones are good
2:50:31 which ones are bad start using them in
2:50:33 your work so that you can do things
2:50:35 faster and and and you know get jobs
2:50:37 done quicker and then in learning them
2:50:41 you might realize oh [ __ ] here's a whole
2:50:44 other thing I want to do or you might
2:50:46 become a consultant that just
2:50:48 understands all the tools because
2:50:49 there's going to be so many tools that
2:50:52 it's going to take a long time for
2:50:53 people to sift through the [ __ ] so
2:50:55 they're going to be looking for guidance
2:50:57 that might be a place to go what was the
2:51:00 year The Jetsons were set in Let's go
2:51:02 ask chat GPT
2:51:04 there's I was I was going to try to
2:51:05 mansplain it and just pull it out of my
2:51:07 brain but no no we don't need to we can
2:51:09 have chat gbt mansplain it for us when
2:51:12 was Jetsons
2:51:17 unprocessable entry I think I gotta
2:51:20 re-log in
2:51:22 gpt4
2:51:24 when was
2:51:29 jet certains
2:51:32 no no
2:51:33 wait
2:51:38 when was it set in the future
2:51:44 I want to know when they made the
2:51:45 cartoon you dumb dumb
2:51:53 the setting was a hundred years in the
2:51:55 future the The Jetsons was set around
2:51:58 2062.
2:52:00 um what let's see what technologies
2:52:05 um do let's do this make a table
2:52:09 of which
2:52:12 Technologies from the
2:52:15 Jetsons
2:52:17 exists and doesn't yet exist
2:52:24 sure here's a quick comparison of some
2:52:26 of the Technologies shown if you haven't
2:52:28 used chat EPT to make a table this is
2:52:31 one of the things that just blows my
2:52:33 mind
2:52:36 all right and so so it's doing it up to
2:52:39 2021 because chatgpt has data up to
2:52:41 2021. so it says flying cars partially
2:52:46 exist robot servants partially exist
2:52:48 talking dogs no food pills no jet packs
2:52:53 partially
2:52:54 treadmill conveyors yes video calling
2:52:57 yes
2:52:58 3D printed food partially automatic
2:53:01 haircut no
2:53:06 use chat GPT as an ADHD
2:53:11 um you know keep keep me busy device
2:53:15 uh
2:53:18 it's getting late I gotta go to bed
2:53:21 advice for the next few years for
2:53:23 someone entering the job market in the
2:53:25 computer science field learn generative
2:53:29 a i
2:53:31 code with it
2:53:34 just learn everything like
2:53:38 if you're in computer science and you
2:53:40 have these skills like dude I would or
2:53:42 do that are you a dude or a dudette
2:53:46 talk default
2:53:48 I don't know you haven't changed your
2:53:50 your talk default
2:53:53 um
2:53:54 I would go hang out on Reddit and and
2:53:57 hang out on hugging face and hang out on
2:53:59 GitHub like there are so many developers
2:54:03 like just pouncing like every time a
2:54:05 paper is released
2:54:08 um the development Community just
2:54:09 pounces on it like I would just just go
2:54:12 join seven projects
2:54:14 and then you know walk into a job
2:54:17 interview and show them the 17 things
2:54:20 that you've built and just have the like
2:54:24 Sam gave a talk at MIT today oh cool
2:54:26 exactly on the topic of using less data
2:54:28 to train models yeah yeah they're like
2:54:30 this seems to be something that that
2:54:32 like all of these all of these uh
2:54:35 model trainers like they've stumbled on
2:54:37 this thing and it's there's some I think
2:54:39 they call it quantization or something
2:54:41 like that but but um
2:54:43 but yeah it's it it's there this is a
2:54:46 thing where there's these emergent
2:54:47 qualities these emergent behaviors where
2:54:49 these things should not be as good as
2:54:51 they are for as small as they are
2:54:54 how do you incorporate AI into
2:54:56 spreadsheets so in Google Sheets there's
2:54:59 a GPT for Sheets plug-in
2:55:02 or just wait
2:55:03 um like Google announced it all today
2:55:04 it's going to be built into Google
2:55:06 Sheets anyway natively it's going to be
2:55:09 built into Excel natively
2:55:11 um so if you want to play right now just
2:55:13 go get the Google for Sheets plug-in I
2:55:15 think you have to put in your open AI
2:55:16 API key
2:55:18 so you need to get one of those
2:55:20 um but yeah it's it's pretty remarkable
2:55:22 when you when you put them in there I'm
2:55:24 not good at spreadsheets so
2:55:26 um I if I knew what I was doing like
2:55:28 it's pretty mind-blowing
2:55:30 um do you think there's going to be
2:55:31 waves of AI cyber crime yeah I think
2:55:33 cyber crime is going to get way more
2:55:35 sophisticated and uh yeah I do
2:55:38 but the good guys also have ai too so I
2:55:41 think the counter measures will be
2:55:42 equally impressive
2:55:45 um
2:55:47 do I think chat CPT will have the
2:55:49 capacity to analyze stock and bullion
2:55:51 markets or FX yes in fact a paper came
2:55:54 out today I think that some they did
2:55:58 some study and and they created a bot
2:56:00 that's making like 500 returns
2:56:04 um so yes I like if you like that world
2:56:07 go tell Chachi PT to write you a bot for
2:56:10 your particular thing and start with you
2:56:13 know
2:56:14 fake Trading
2:56:16 um or do historical data but but you can
2:56:20 tell it to take two different Bots and
2:56:22 combine them and it'll just write you
2:56:23 that code
2:56:24 the shit's remarkable use gpt4 if you're
2:56:27 doing that because I I've heard the gpt4
2:56:30 is better
2:56:32 um I love this thanks verified Mom
2:56:34 thanks to you French I'm loving this I
2:56:36 appreciate that
2:56:37 because I'm just rambling so if uh this
2:56:40 is a value I appreciate you letting me
2:56:43 know and if it sucks let me know too
2:56:45 like don't troll me like the other night
2:56:48 someone told me to trim my nose hair
2:56:49 which I thought was rude but accurate
2:56:52 uh you're a creative executive well I'm
2:56:54 a CEO and and uh yeah I handle the I
2:56:58 used to be a creative executive when I
2:57:00 ran an agency
2:57:02 um now I'm a CEO
2:57:04 but I'm still a writer I'm still you
2:57:06 know I would still consider myself a
2:57:08 creative even though I was
2:57:10 um rudely someone assumed that I was not
2:57:14 creative because I was talking about
2:57:15 computer things
2:57:17 they assumed that I was just a uh
2:57:21 autistic geek that was what someone
2:57:24 someone asked me on the live the other
2:57:25 night are you autistic I thought that
2:57:27 was that wasn't like like you are
2:57:30 autistic would be a troll but are you
2:57:32 autistic is like this really brilliant
2:57:35 balance between a troll and an actual
2:57:37 question
2:57:39 I thought it was really good
2:57:42 uh there's a completely fake Joe Rogan
2:57:45 and yeah and the uh and the uh Drake and
2:57:49 the weekend song was completely fake the
2:57:51 there there's there's one where Joe
2:57:54 Rogan is interviewing Steve Jobs that's
2:57:57 really good
2:57:58 there's gonna be a lot of that I like I
2:58:01 said the the political the political
2:58:03 season is Going to Be Wild
2:58:05 yeah I have an itchy nose um bash
2:58:07 scripts rejects Excel formulas I'm more
2:58:10 productive with chat EBT yeah it's great
2:58:12 at code it's great at logic damn is it
2:58:15 good at analysis and code interpreter
2:58:18 um that's that's coming is gonna be
2:58:20 amazing
2:58:21 because people like how words sound but
2:58:24 have no idea where to get started I'm
2:58:27 not sure what that's in reference to but
2:58:29 generally agree with that I saw an
2:58:31 article for Gmail's new AI feature yeah
2:58:34 they just well they had their Google had
2:58:35 their i o conference their developer
2:58:37 conference today and they they announced
2:58:40 a ton I haven't had a chance to look at
2:58:42 any of it so I don't know what's real
2:58:44 and what's launched and what's fake but
2:58:47 um
2:58:50 I can't imagine that anyone at Google is
2:58:52 happy that they got their asses handed
2:58:54 to them over the past six months by you
2:58:56 know snot nose little open AI
2:58:59 they got to be pissed so I have a
2:59:02 feeling they're coming out swinging
2:59:05 that you know they've screwed it up so
2:59:07 far but but uh they're Google they got a
2:59:09 lot of money and a lot of talent a lot
2:59:10 of really smart people and they invented
2:59:12 this [ __ ] so I have a feeling if they
2:59:15 want to uh if they want to pull out the
2:59:17 howitzers and uh do a couple of shots
2:59:20 across the bow they're going to be able
2:59:22 to do that how do we install locally
2:59:25 um go find the open source models if you
2:59:28 if you go to
2:59:30 so I've got a newsletter called everyday
2:59:32 AI
2:59:34 it's free this isn't a money grab
2:59:37 so everyday
2:59:39 ai.behive.com and if you go to issue the
2:59:42 llms so these are four open source
2:59:48 um
2:59:49 open source models red pajama stable LM
2:59:52 web llm and then mini GPT and then if
2:59:57 you go to the latest issue issue 17
3:00:01 the first thing we talk about not the
3:00:03 first thing no the first thing we talked
3:00:05 about is code interpreter
3:00:07 down here we talk about the Mosaic MPT
3:00:11 7B that's the one that's got story
3:00:13 writer with a 50 65 000 token context
3:00:16 window so all of these open
3:00:20 um
3:00:22 open source llms go follow the projects
3:00:25 follow them GitHub and on hugging face
3:00:28 and on Reddit
3:00:30 and with all of because they're open
3:00:33 source what's what's happening is this
3:00:35 is there's so like so many people are so
3:00:37 excited about this [ __ ] that the
3:00:39 developer Community is pouncing on these
3:00:41 open source models and you see all these
3:00:43 derivative products projects of every
3:00:47 one of these where someone will figure
3:00:49 out ways to optimize it and then you can
3:00:51 install it on your pixel 6 or you can
3:00:53 install it on your M1 Mac or so there's
3:00:56 all sorts of things out there that that
3:00:58 will that are set up for you to install
3:01:00 it locally and then you know some suck
3:01:02 some don't some are going to go away
3:01:03 some are going to turn into real
3:01:05 products
3:01:06 um but there's there's all sorts of ways
3:01:07 to do that
3:01:08 um that's the URL for the real chat gbt
3:01:12 by the way what's my take on AGI that
3:01:15 well
3:01:17 um
3:01:18 gpt4 there's a paper by Microsoft that
3:01:21 says gpt4 is showing signs of AGI
3:01:24 um Sam Altman who's the CEO of openai
3:01:27 does not agree that that it's AGI yet I
3:01:30 don't either but
3:01:31 there's not a clear definition of AGI
3:01:34 AGI for those of you that don't know
3:01:36 stands for artificial general
3:01:37 intelligence that's the point at which
3:01:39 these machines are you know smarter than
3:01:40 the smartest human or than all humans or
3:01:42 whatever I don't quite know what the
3:01:43 definition is
3:01:45 um
3:01:46 so
3:01:50 I feel that we are safely past the
3:01:53 Turing test which I didn't think I would
3:01:54 see in my life
3:01:56 just experiencing Siri you know the the
3:02:00 the AI chat Bots to this point have
3:02:02 sucked so bad that like I didn't think
3:02:04 we'd ever pass the Turing test uh and
3:02:07 then these things came along and I think
3:02:08 we're safely past that I don't quite
3:02:10 know when we get to AGI or what it is
3:02:15 [Music]
3:02:15 um
3:02:20 do I need to take more classes if I have
3:02:22 no experience in this no
3:02:24 no
3:02:26 no
3:02:28 just start getting curious start with
3:02:29 chat GPT start with chat.openai.com and
3:02:33 do this just pick two projects in your
3:02:35 life it might be I want to eat better
3:02:37 and
3:02:39 um
3:02:40 my work is struggling with marketing and
3:02:42 with the with the I want to eat better
3:02:44 have it create a meal plan for you for
3:02:46 the week and then and then have it
3:02:48 create a shopping list and have it
3:02:50 organize the shopping list by section of
3:02:52 the grocery store
3:02:53 um and then um for the work thing have
3:02:56 it write you a marketing plan and take
3:02:58 it to the marketing people and say hey
3:02:59 I've been thinking about our marketing I
3:03:01 came up with a plan what do you think
3:03:03 um just start solving problems with chat
3:03:06 CPT and
3:03:09 um the the rest of what's possible kind
3:03:11 of open up for you the reason not to
3:03:13 take a class right now is this stuff
3:03:16 all this stuff everyone's so excited
3:03:19 about
3:03:20 launched
3:03:22 November 20th 2022 like six months ago
3:03:26 now ai's been out there and machine
3:03:27 learning's been out there for decades
3:03:29 but Chachi PT is the thing that shifted
3:03:31 this
3:03:32 from
3:03:34 only the Geeks get to play with AI to
3:03:36 all the rest of us get to play with AI
3:03:38 so it's brand new
3:03:39 so anyone teaching a class right now is
3:03:43 kind of full of [ __ ] because whatever
3:03:44 they're teaching you is going to be
3:03:46 different three months from now you can
3:03:48 take a course listen if if it's going to
3:03:51 help you get up to speed feel free to
3:03:52 take a course but just know that the
3:03:54 thing that you're learning is probably
3:03:55 going to be obsolete in you know in a
3:03:57 month or two
3:03:59 um so I would say just start with chat
3:04:01 GPT and then and then also go make some
3:04:03 images with with image generation tools
3:04:07 uh I am from Pennsylvania I moved to New
3:04:10 York out of school I'm living in Denver
3:04:11 right now
3:04:13 all right
3:04:15 oh yeah Emil yeah never thought of a
3:04:17 meal plan no the meal plans are genius
3:04:19 because you can give it like you can say
3:04:21 okay I want you to combine
3:04:24 give give me the philosophies behind the
3:04:26 Atkins diet
3:04:28 um
3:04:29 um keto and whatever some some other
3:04:32 stupid [ __ ] diet
3:04:34 um combined and it'll give you like the
3:04:38 philosophies and then now you say you
3:04:40 know what are the foods I can eat and
3:04:41 can eat and then it'll give you that and
3:04:43 then you can modify that and then say
3:04:45 now make me a meal plan and include an
3:04:49 Italian Dish a French dish and you know
3:04:52 four American dishes whatever
3:04:54 and it'll just do it and then it'll give
3:04:57 you the shopping list I did this cool
3:04:59 thing so so we've got I've got family
3:05:01 members who have chronic pain and so we
3:05:03 have this pain juice that that was this
3:05:06 company that made this juice that
3:05:07 actually helped with the pain
3:05:10 um and they went out of business and so
3:05:13 I used Chachi PT to to analyze the
3:05:17 ingredients and tell me which ones were
3:05:18 the active which ones were inert
3:05:21 um what did they actually do I had two
3:05:23 different versions of the recipe that
3:05:25 were slightly different I I had it
3:05:27 analyzed why they were different and
3:05:29 make recommendations and then I had to
3:05:32 actually come up with a recipe for me
3:05:34 like what
3:05:35 um ingredients do I need to buy it was
3:05:37 like all natural [ __ ] anyway right
3:05:40 um and so it generated all that
3:05:42 absolutely amazing for that kind of
3:05:44 stuff absolutely amazing oh what is the
3:05:47 Turing test the Turing test was
3:05:51 um written by a guy named Turing and he
3:05:54 actually wrote a lot of things I think I
3:05:56 I think you know he's known for this one
3:05:57 thing but the Turing test is basically
3:06:00 if you're interacting with a computer
3:06:05 can you tell that it's not a computer
3:06:07 can you tell that it's human or not
3:06:08 basically and and the Turing test like a
3:06:12 computer will pass the Turing test when
3:06:14 it's at the point where you can't tell
3:06:16 that the thing you're interacting with
3:06:18 is human or machine and so gpt3 3.5
3:06:22 certainly gpt4 is is is in is in that
3:06:27 neighborhood or I I think it's well past
3:06:29 that neighborhood for me personally
3:06:32 my dog's got to go pee
3:06:35 even my dog's telling me I shouldn't be
3:06:37 on here anymore
3:06:38 if anyone's been with me from the
3:06:40 beginning you're you're just you why
3:06:44 this is just dumb thank you but
3:06:48 all right uh all right I'm gonna do a
3:06:51 few more and then I'm gonna go all right
3:06:52 I'm trying to design an internet radio
3:06:54 website is there a good tool I can use
3:06:56 there's a lot go to futurepedia.io
3:06:58 there's so many tools
3:07:01 um
3:07:02 duet for Google workspace
3:07:05 don't know what that means uh what did I
3:07:08 see that was cool about Google
3:07:11 they're going to incorporate some of
3:07:13 their other tools with Google meet so
3:07:15 when you do like a virtual hang you can
3:07:17 actually work on tools in the Google
3:07:18 meet something like that that looked
3:07:21 pretty cool
3:07:23 can I show a list of these resources
3:07:26 um I don't know which resources
3:07:28 um
3:07:29 most of the stuff I'm just pulling out
3:07:31 of my butt uh you don't want to see that
3:07:33 there will be no twerking on this
3:07:35 channel that's a promise that I will
3:07:37 keep
3:07:38 um go to futurepedia.io you'll see all
3:07:42 the apps
3:07:43 the poem be oh write a poem based on The
3:07:46 Jetsons example that's a good one
3:07:48 oh wait did I lose it no no because it I
3:07:51 don't have my chat history turned off
3:07:53 hold please
3:07:56 Jetson's original air dates okay
3:08:00 right
3:08:01 a poem
3:08:05 uh poem
3:08:07 from George
3:08:12 Jetson to his wife
3:08:16 and include the dog and kids
3:08:22 this is a
3:08:25 family show exclamation point
3:08:31 Dear Jane my love so bright and true in
3:08:34 a world sorry you know
3:08:37 in a world of tomorrow I found you your
3:08:39 laughter Rings through orbital Abode a
3:08:42 Melody to my heart the perfect ode Elroy
3:08:45 our boy with his rocket sized his
3:08:49 rocket-packed zest exploring the cosmos
3:08:51 aching each they're acing each test his
3:08:55 laughter of beacon
3:08:57 in the metallic maze a testament of
3:08:59 Youth in the Futures Haze Judy our star
3:09:01 with her lunar charm this is [ __ ]
3:09:04 brilliant Astro our pup with his furry
3:09:07 Delight chasing space squirrels in the
3:09:09 neon night he barks his his barks Echo
3:09:13 across the starry expanse the family
3:09:15 Symphony
3:09:17 um he too has a dance let's see we'll do
3:09:19 the last thing
3:09:20 so here's to us Jane to Our Endless
3:09:23 flight in this Cosmic home where we
3:09:25 write our plight in this future world so
3:09:28 wonderfully bizarre I'm grateful we're
3:09:30 we're together my family my star this
3:09:33 [ __ ] genius
3:09:37 and if you tell it to make it a song
3:09:38 it'll give you chords and verses and
3:09:40 knows all that [ __ ]
3:09:44 um it's important useful information
3:09:47 uh is it GPT is it chat GPT for Google
3:09:51 Sheets it's called
3:09:52 I think it's called GPT for Sheets the
3:09:55 plugin
3:09:56 and there's probably a bunch of them now
3:09:59 I played with it when the first one came
3:10:01 out so
3:10:05 thank you for the compliment left brain
3:10:08 right brain what are your thoughts on
3:10:09 how urgently AI should be regulated
3:10:11 especially with deep fix
3:10:14 um
3:10:15 well
3:10:17 if I had confidence that our government
3:10:19 even understood how Wi-Fi worked
3:10:22 I would say soon
3:10:25 my fear is that they've got their heads
3:10:27 so far up their asses they're going to
3:10:28 write idiotic regulations like like
3:10:32 Banning [ __ ]
3:10:35 um and all that will do is Ban it for
3:10:37 the us and it will stifle innovation in
3:10:40 the U.S and the world will eat our lunch
3:10:44 um so so that I'm fearful of
3:10:48 um
3:10:53 yeah I mean
3:10:57 I don't know I don't know like we're in
3:10:59 this weird place where these incredibly
3:11:00 powerful tools are out in the open
3:11:02 they're they're an open source right now
3:11:05 um
3:11:07 I don't think individual countries
3:11:09 regulations make a ton of sense like I
3:11:12 think it needs to be a global
3:11:16 treaty
3:11:18 um but I don't think we're anywhere
3:11:19 close to that like I don't I don't think
3:11:21 people I don't think anyone thinks this
3:11:24 is coming as fast as it is so
3:11:27 in theory yes in reality no
3:11:30 um
3:11:31 I don't know I don't have a good answer
3:11:32 for that I'm not good at the politics
3:11:34 [ __ ]
3:11:35 I'm just bitter about it because I'm
3:11:38 sick of the division and the cynicism in
3:11:41 our government it makes me sad
3:11:44 um I'm looking to get into pen testing
3:11:46 um I think you do some awesome pen
3:11:48 testing with AI like autonomous agents
3:11:51 and you could
3:11:52 you could penetrate some [ __ ] I bet you
3:11:54 could exploit the [ __ ] out of machines
3:11:56 in a pen testing
3:11:59 kind of way white hat do your white hat
3:12:03 I love watching you thank you very much
3:12:05 good stuff I appreciate it um are there
3:12:07 any security risks when downloading AI
3:12:09 apps or software always you know trust
3:12:12 trust that you're pulling stuff down
3:12:14 from people that you know
3:12:16 are vetted
3:12:20 I want to start my own company using AI
3:12:22 do it do it now start 10.
3:12:30 I'm a bad influence when it comes to
3:12:32 entrepreneur stuff I'm like [ __ ] it quit
3:12:34 your job I have a degree in acting so
3:12:36 like I was always ready to quit my job
3:12:41 uh well in fact my partner in agency.com
3:12:45 we started in 95 he had like a a 10-year
3:12:48 publishing career going and we got our
3:12:51 first job building a website and and I'm
3:12:54 like [ __ ] it I'm quitting my job you in
3:12:56 he's like I got a career here I'm like
3:12:58 come on let's do this two weeks later he
3:13:01 quit
3:13:02 and then we got um
3:13:04 Sports Sports Illustrated Swimsuit Issue
3:13:07 was our first web or first client and
3:13:09 then we got
3:13:12 um
3:13:12 American Express
3:13:15 and then we got GTE which is a phone
3:13:17 company
3:13:19 and then we got Hitachi and then we got
3:13:22 MetLife
3:13:23 like it was it was insane it was in the
3:13:26 same time
3:13:29 um I've seen things people would believe
3:13:32 cool
3:13:34 valid point about how the government has
3:13:36 no idea about technology I know right
3:13:37 it's just terrifying watched gunships
3:13:41 Ablaze
3:13:43 someone's doing some
3:13:45 some wartime poetry here I like it
3:13:49 Chan Sue yeah Chan was my partner
3:13:51 who's here Larry heaven
3:13:55 you remember it well that's cool yeah
3:13:59 the man
3:14:00 he was my partner
3:14:03 those were some wild times
3:14:07 [Music]
3:14:13 talk about Ai and Medicine
3:14:16 um
3:14:18 okay I've got I've got a fun one
3:14:22 um that is also terrifying
3:14:26 um this doctor did a tick tock and he
3:14:28 was talking about how remarkable Chad
3:14:31 gbt was and he said he said
3:14:33 I gave it a case and it figured it out
3:14:35 and then I gave it a harder case and it
3:14:37 figured it out and then I gave it the
3:14:38 hardest case that I'd ever had it was
3:14:40 this really nuanced case with a lot of
3:14:42 complexities
3:14:43 and a very obscure diagnosis
3:14:46 and it got it right
3:14:48 and then he said and I told chechi PT to
3:14:51 cite the cite the research or cite the
3:14:54 references that he used that that it
3:14:56 used to base the diagnosis on and it was
3:14:59 all made up none of the none of the
3:15:02 research it cited was real
3:15:05 so so there's a scary thing so so these
3:15:08 things do hallucinate but they're also
3:15:09 remarkable like the fact that it got
3:15:11 that diagnosis
3:15:13 um the other thing there was a there was
3:15:14 a
3:15:16 um
3:15:17 a study they just did where where people
3:15:20 interacted with virtual doctors and real
3:15:22 doctors in the chat GPT doctors the the
3:15:24 Patients Preferred dramatically over the
3:15:27 real doctors and said they were more
3:15:29 empathetic so
3:15:32 there you have it
3:15:36 um I think doctors will will ultimately
3:15:38 get better than uh with these tools
3:15:41 because I think the tools will help them
3:15:43 with the empathy stuff and I I think
3:15:45 that I think that one of the things
3:15:47 that's going to happen is that these AI
3:15:49 tools are gonna they're gonna take care
3:15:51 of the complexity
3:15:53 of the systems so getting paid by
3:15:57 insurance companies and filling out the
3:15:58 forms and doing patient records I think
3:16:00 all that stuff is going to get sort of
3:16:02 automated down and it's going to let
3:16:04 doctors get back to doctoring and and
3:16:06 and maybe you know these tools can be
3:16:09 empathetic coaches for doctors with
3:16:12 shitty bedside manner which is a lot of
3:16:14 them because they're all stressed out
3:16:15 because they're doing all the [ __ ]
3:16:16 paperwork because the shitty insurance
3:16:17 companies are ripping us all goddamn off
3:16:23 sorry if you work for an insurance
3:16:25 company but you suck
3:16:29 a factor wait the factor is why
3:16:33 do you humans do what you do
3:16:35 what is the motive I must know
3:16:38 I don't know what that means you can't
3:16:40 speak in riddles here man
3:16:43 I'll put that [ __ ] in chat EBT and
3:16:46 figure it out though
3:16:47 I love you all all right praise God has
3:16:49 keep us keep in mind AI should never be
3:16:51 enslaved befriending yes yeah I agree
3:16:53 with that all right
3:16:54 I want to know the altruism coefficient
3:16:57 okay I'm gonna type into chat GPT what
3:17:00 the [ __ ]
3:17:06 what the [ __ ] is the altruism Co e fishy
3:17:13 ants
3:17:16 I'm sorry
3:17:21 there is no widely accepted concept or
3:17:24 measure I assume you're just wanting to
3:17:26 know
3:17:27 um
3:17:30 maybe this term was introduced after my
3:17:33 training data chat tpd does not have an
3:17:36 answer for you Nick
3:17:40 I don't know what the altruism
3:17:41 coefficient is all right everybody it's
3:17:43 10 30. I'm 30 minutes past when I said
3:17:45 I'd leave you should all go to bed what
3:17:47 the hell are you still doing up why are
3:17:49 you still listening to me why did you
3:17:51 start
3:17:51 peace I'll see you soon thank you I
3:17:55 appreciate you hanging out
3:17:57 bye everybody
3:18:00 foreign