aiLL 05/23/23 - Navigating the Generative AI Revolution

Description

Chapters

Transcript

0:07 turn down
0:15 yes
0:17 come down
0:20 shut up
0:22 hello it's Kyle Shannon with the AI
0:24 learning lab I was live and then I went
0:26 away and now I'm back I fixed my
0:28 internet
0:29 because I'm a genius
0:32 try unplugging it that always works
0:36 um this is the AI learning lab if you
0:37 have questions about AI yeah it's back I
0:40 know right yay
0:42 we're back
0:44 um if you have questions about AI drop a
0:47 line
0:48 um subscribe to my lives if you if you
0:50 like me being a smart ass talking about
0:52 AI stuff
0:53 um uh subscribe to it and you'll get
0:55 notified when I go live
0:57 um I think I'm up to four subscribers
1:00 let's see if we can get that up to 10
1:02 tonight let's go nuts go crazy there's
1:04 eight of you here if eight of you
1:06 subscribed right now well there's
1:08 probably a few of you already are
1:10 um but anyway
1:12 um yeah I'll talk about AI stuff I'm
1:13 happy to talk about any questions I can
1:15 talk to you about my background
1:17 um this is the AI learning lab I like
1:19 talking about this stuff answering
1:20 questions doing what I can that URL is
1:23 the official chat GPT
1:26 um what do you call it what you made
1:28 dizzy uh URL uh
1:31 chat.openai.com and then futurepedia.io
1:34 right before I my Wi-Fi died
1:37 um someone asked is there is there a
1:38 list of AI tools out there
1:40 futurepedia.io is one of them there's
1:42 two or three others out there that also
1:44 have future in the name but
1:46 futurepedia.io is one that I know that's
1:48 pretty good
1:50 um subscribe to the life and the
1:52 newsletter thank you very much kelbell
1:54 wag appreciate it that's a fun name
1:58 um if you have questions about AI drop
2:00 them in the chat here I will do my best
2:03 to answer every question I can and I
2:05 will tell you when I can't I appreciate
2:08 the hearts and the likes and the thumbs
2:10 up and the subscribes and the all that
2:12 sort of stuff
2:14 um I've been away for a few days I was
2:15 speaking at a conference for
2:18 um military simulation
2:21 which was kind of scary people they're
2:24 not the kind of people I hang out with
2:25 on a regular basis
2:27 um but it was interesting you know the
2:29 military has had access to AI for
2:31 decades and probably some crazy [ __ ]
2:34 that you know we can't even imagine
2:36 um and yet there was a lot of people
2:38 there that were not
2:40 um that were just starting to get their
2:42 head around this generative Ai and
2:44 what's what's coming and what's
2:45 happening
2:47 um my thoughts on the election and deep
2:49 fake Tech I think this next election is
2:51 going to be a master class and getting
2:53 our asses handed to us with
2:55 misinformation and deep fakes
2:57 um I
2:59 I think that the political campaigns are
3:01 going to be ruthless in the use of this
3:03 I mean I mean the deal is this I can
3:07 with without much effort
3:10 um do it do a decent clone of myself
3:12 right now without even really trying
3:15 um so so these political campaigns the
3:19 the Deep fakes and it's not it's I
3:21 wouldn't even call them deep fakes I
3:22 would say the sophisticated
3:25 manufactured media
3:28 that is going to be indistinguishable
3:30 from
3:31 real media is is going to be
3:34 unprecedented and off off the charts and
3:37 I think what will happen is you'll see
3:39 it like two days before the election
3:40 there's you know Biden French kissing a
3:43 gorilla it doesn't matter like whatever
3:45 the ridiculous thing is
3:47 um and it's going to look real and then
3:49 everyone's gonna deny everything it's
3:51 going to be a mess so I think that I
3:53 think that we're in for
3:55 um a pretty harsh lesson in what this
3:57 stuff makes possible
3:58 which I think is then going to lead to
4:02 a future where we start
4:05 requiring like as individuals that
4:09 things be verified authentic whatever
4:11 that means maybe they have a little
4:12 check mark like a verified check mark I
4:15 think there's going to be whole
4:16 businesses built around this where if I
4:19 make a tick tock video me Kyle on my
4:22 channel
4:23 um I can you know somehow Watermark that
4:25 in such a way they say yeah that's me
4:26 and and you using whatever the system is
4:29 can trust that I think that's going to
4:31 be
4:32 um that's going to be important um chat
4:35 EPT is too woke yeah it's got some
4:37 biases all of these systems are going to
4:38 have biases
4:40 um I think that in the future you'll be
4:42 able to select different models and you
4:44 can choose your own bias if you want one
4:46 that's biased more to the right that'll
4:48 be available to you if you want the
4:49 neutral one that'll be available for you
4:51 if you want the one that's into all the
4:54 crazy [ __ ] you are that'll be available
4:55 to you so don't don't
4:57 don't worry too much I I would I would
5:00 caution against dismissing this
5:02 technology because it's not perfect
5:03 right now we're very early in this very
5:06 early chat GPT November 30th 2022 last
5:10 November six months ago
5:12 is when Chachi BT rolled out now gpt3
5:16 was out for about a year and a half
5:17 before that and then two before that and
5:18 one before that
5:20 so the Technology's been out for a while
5:22 but
5:23 our access to
5:26 um GPT is is new
5:29 um and and so these tools are very very
5:31 raw very raw don't mistake the fact that
5:34 they're not perfect right now for how
5:36 powerful they are
5:38 um and and don't dismiss them just
5:39 because oh it's woke all right so you
5:43 know so is half the country uh in
5:45 America anyway
5:46 okay thoughts on the election deep fake
5:48 all right my friend had a panic attack
5:51 this good good this week about AI chat
5:54 gbt and how it's going to affect global
5:56 economy
5:57 um
6:00 I think
6:01 anyone who's paying attention to what's
6:04 going on with generative AI right now
6:07 um
6:08 if if you're
6:10 if you're learning about it you should
6:12 be a combination of incredibly excited
6:15 and Incredibly scared
6:18 um
6:19 I think you know panic attack level
6:22 stuff listen I I suffer panic attacks so
6:24 I know what it's like it's not pleasant
6:26 um
6:30 if if if they're panicking without using
6:34 the tools then then I'd say that's
6:36 unhealthy then then they're kind of
6:37 living in a world of
6:39 you know Hollywood for 50 years has told
6:41 me that the robots are going to kill us
6:44 um if they're just sort of reading news
6:45 articles and things like that you you
6:48 got to realize the media is going to
6:49 pick the most Sensational things and so
6:51 they're going to pick the dramatic
6:53 scenario so it's very easy to fall into
6:55 the Trap of the robots are going to kill
6:56 us
6:58 the robots are going to take our jobs
6:59 the robots are also going to free us up
7:02 from doing monotonous soul-crushing work
7:05 and they're going to also free us up to
7:06 do
7:08 um
7:09 work that is you know sort of creative
7:11 on a level that none of us could even
7:13 imagine they're going to amplify our
7:15 Humanity
7:16 um in
7:17 in as remarkably positive ways as as you
7:21 know potentially negative ways so the
7:23 the only antidote to the to the fear in
7:26 my experience to this point is start
7:29 using the tools go to chat.openai.com
7:32 start playing with chatgpt actively
7:35 start treating it like a collaborative
7:36 partner
7:38 um
7:38 are you
7:40 surefuturepedia.io is spelled correctly
7:42 no oh no it's not
7:44 thanks Ron
7:46 that's what happens when you give an
7:47 actor access to a keyboard so it's it's
7:50 just like giving idiots access to AI
7:54 um let me just go confirm that that is
7:56 indeed correct
7:58 copy
8:01 let's see we're gonna go here
8:05 I can't believe someone told me to check
8:07 my thyroid that's a double chin fat joke
8:10 I feel attacked
8:13 futurepedia.io yep that's correct now
8:15 okay thank you for that Ron
8:17 so that is now the correct spelling of
8:19 future PDS so chat.openai.com is uh the
8:22 official chat gbt website
8:25 futurepedia.io is just a list of tools
8:29 oh it took you to a strange site sorry
8:30 sorry if I installed the virus on your
8:33 computer because I had a typo up there
8:35 true we will probably be able to create
8:38 Tech to differentiate generated media
8:40 yeah I think I think the idea of well so
8:45 Google just announced a deal with Adobe
8:48 where Adobe uh what's it called Firefly
8:52 is going to be the engine underneath
8:53 Google image generation and they say
8:56 they're going to flag everything that's
8:57 AI generated but generative ai's now
9:01 been built into Photoshop where you can
9:02 photo where you can generate parts of an
9:04 image so I don't know where the boundary
9:07 is between if you
9:10 generate 90 of an image is it still your
9:13 image or is it AI generated like I don't
9:15 know that there's going to be real clear
9:17 boundaries but I do think there's going
9:19 to have to be
9:20 Technologies to say
9:22 you know maybe me as the Creator created
9:25 this this is my authentic creation or
9:27 this is my authentic Voice or this is my
9:29 authentic model or whatever it is so I I
9:31 we're gonna have to solve that and again
9:33 I think I think this next election is
9:35 gonna
9:36 accelerate demand for
9:38 first stuff
9:41 that that fixes it adaptive AI is
9:43 already out there lots of AI is already
9:45 out there will AI affect schooling oh
9:48 I think AI is going to upend education
9:51 uh like in dramatic and profound ways
9:55 um
9:56 the the the
9:59 Educators right now that are embracing
10:01 this and and figuring out how to
10:04 re-engage with their students and drive
10:06 critical thinking and and and and
10:08 basically rather than fighting
10:11 this technology like it's not going to
10:13 be here it's already here Microsoft
10:15 announced today it's going to be a core
10:17 part of the Windows operating system
10:19 it's going to be in Microsoft Word it's
10:21 going to be in Google Docs it's going to
10:23 be in sheets and Excel it's going to be
10:26 in every screenwriting software it's
10:27 going to be in everything
10:29 so these these large language models are
10:32 literally going to be layered on top of
10:34 every system out there so you're not
10:36 going to be able to escape them so any
10:38 of the Educators right now that are
10:40 trying to fight this it's it's an it's
10:42 an it's a battle that they're losing
10:44 very very quickly in fact I just heard
10:46 that was it the Boston School District
10:48 or some School District just lifted
10:50 their ban of chat gbt because they
10:53 realize that it's inevitable um I think
10:55 the Silver Lining to it though is
10:58 um it's going to force us out of
11:00 teaching as you know one person at the
11:03 front of the room
11:04 teaching things that need to be
11:06 memorized and then you know spit back
11:09 wrote I think that
11:12 way of teaching that's going to go it's
11:14 it's just gonna be forced out and so I
11:16 think there's an opportunity to to
11:18 Really transform education in a powerful
11:20 way but it's going to be painful
11:23 um have I heard about Drake AI could you
11:26 please explain what was AI part and what
11:29 was the human part uh I I assume the
11:32 question you're talking about is the
11:34 Drake weekend that that song that was
11:36 put out that was Drake if that's what
11:38 you're asking about
11:40 the Drake and weekend voices were
11:43 synthesized they were clones so they
11:45 basically made a model of Drake's
11:47 singing and the weekend singing and then
11:50 they mapped those two
11:52 um I guess someone else singing the song
11:55 um and they they just map those voices
11:57 on top of it I don't know if the
11:59 underlying music was AI generated but I
12:01 think the lyrics were and the voices
12:02 were and I think the underlying music
12:05 was
12:07 um
12:07 so so but we're going to see a lot of
12:09 that and and Grimes who's
12:12 um Elon musk's ex she's a musician about
12:15 a week after that thing dropped she
12:17 announced hey um rather than me fighting
12:20 this go ahead clone me all you want just
12:22 give me 50 of whatever you create which
12:24 I think that is actually
12:27 um I think that's really wise on her
12:30 part because I think that's where this
12:31 will be headed is that people will
12:33 license their likeness and they'll get
12:35 paid on the generation or or on you know
12:37 split of the of the deal mostly
12:39 terrified it's fine to be mostly
12:41 terrified right now but what I would say
12:43 is if you're mostly terrified and not
12:45 using the tools it's just gonna get
12:47 scarier and scarier so start using the
12:49 tools start learning how they work and
12:52 what they do and what they can do for
12:53 you because it they really are super
12:56 powers that can amplify who you are and
12:57 what you do ai's already affecting
13:00 Schools they're using it for homework
13:01 well and teachers are using it
13:04 um oh yeah if you can feel free to share
13:07 this live let other people know this is
13:09 happening you can subscribe to my live
13:11 you can subscribe you know follow the
13:13 channel
13:14 um you can tap the little heart button I
13:16 like all of that uh I'm the reason I'm
13:19 doing this is just try to get get the
13:22 word out that this AI stuff is while
13:25 scary also really exciting
13:28 and
13:30 any little bit that we do now is going
13:33 to have a disproportionately positive
13:34 effect on AI moving forward so the more
13:37 people I can get to get curious about it
13:40 and start exploring it that's why I'm
13:41 doing this channel
13:43 is taking away the from creativity of
13:46 individual Minds uh I one thousand
13:48 percent disagree with that
13:51 it it I have I have not been as
13:54 creatively alive and excited as I have
13:57 been in the past six months in a decade
14:00 and I and I have my own company that
14:02 like came out of my brain I've been like
14:05 I live to be creatively engaged it is
14:08 the complete opposite of what you're
14:10 talking about it's not taking away it
14:12 the the only way that you're saying that
14:14 statement in my opinion is you haven't
14:16 actually played with these tools
14:18 so I would encourage you go there start
14:20 playing start digging in
14:23 um
14:25 what I can tell you is this without chat
14:28 GPT without these generative tools my
14:30 creative process
14:33 I would have an idea I would go to
14:35 execute that idea and I would hit a
14:37 roadblock and the roadblock would be oh
14:39 I need this piece of software I need
14:40 this and I I'd run down these rabbit
14:42 holes of trying to solve problems to be
14:44 able to have my idea and 45 minutes
14:47 later I kind of solved the technical
14:49 problem and then I would I would I will
14:53 have lost the creative inspiration or
14:55 the idea itself altogether what's
14:57 happening with chat gbt is I have a
14:59 whisper of an idea I go oh let me run
15:01 over to chat GPT I start going back and
15:03 forth with chat EBT within 10 minutes
15:05 I've got a fully blown out fully fleshed
15:07 out
15:08 version of that idea and like five more
15:11 ideas so that statement in my opinion
15:14 the one I don't know if you can see what
15:16 I'm seeing chat GPT is taking away from
15:18 the creativity of individual Minds
15:21 um that's not my experience at all
15:24 um Drake I was interesting copyright
15:26 claim is flimsy the the copyright claim
15:28 is going to be
15:29 um copyright claims in general the next
15:31 10 15 20 years are going to be an
15:33 absolute [ __ ] show with
15:35 um with uh copyright and trademark the
15:38 the IP attorneys are are really excited
15:40 right now they're going to make a
15:41 fortune
15:44 [Laughter]
15:44 um
15:47 sorry trying to find out where I was
15:49 okay I saw some things I've never seen
15:52 before not quite sure what you're
15:53 talking about but cool
15:55 um Ron uh very simple where do I start
15:57 as a total beginner do I start by
16:00 finding something I want to use it for
16:01 actually that's a great point so yes so
16:04 where you go technically is
16:06 chat.openai.com that's the chat GPT
16:09 website you don't need to pay anything
16:10 for it you can use the 3.5 model if you
16:13 want to use some of the fancier things
16:14 like it being connected to the internet
16:16 and plugins and some of the new exciting
16:18 stuff you have to pay 20 bucks a month
16:20 but if you just want to get started from
16:22 the standing start go there
16:25 and then start by doing what I call
16:28 parlor tricks and I'll do some for you
16:31 right now so I'm going to switch to chat
16:32 GPT 3.5
16:34 so here we are at chat TPT this is what
16:36 it'll look like for you
16:38 and I'm going to do something like I
16:41 always do the same thing so let me oh
16:42 okay here's a
16:44 um what is the average speed of an
16:49 unladen swallow
16:53 the average speed of an unladen swallow
16:56 blah blah blah blah blah and then it
16:57 mentions that that was a
17:00 um
17:01 a reference to a line from Monty Python
17:04 and the Holy Grail so it actually
17:06 understood that
17:07 um sometimes it will ask you African or
17:09 european
17:11 um and things like that
17:12 um and so so now so so I asked it that
17:15 so now I can say
17:16 um write me a poem
17:20 um
17:21 that sounds like like
17:26 it's
17:29 from the movie
17:31 now what's cool about this is it
17:33 understands what I already asked it it
17:35 understands that it already referenced
17:37 Monty Python and the Holy Grail so now I
17:39 can just write a line write me a poem
17:41 that sounds like it's from the movie
17:43 and let's see what it does
17:47 um
17:48 in lands of humor where Jesters dwell
17:50 their lives a tail we all know we know
17:52 so well of knights in quest for the Holy
17:55 Grail where questions peculiarly
17:58 peculiarly assail right so so now it's
18:01 written a thing so a fun way to start
18:04 with these large language models is just
18:06 to like do parlor tricks asking for
18:09 things that are ridiculous have it write
18:11 you a song have it write you a poem have
18:13 it explain quantum physics to you as if
18:15 you were a third grader
18:17 um have it explained quantum mechanics
18:18 to you
18:20 um only using grapes and hammers as
18:23 metaphors let's actually let's do that
18:25 one wait
18:26 explain
18:28 quantum
18:30 mechanics
18:33 to me only using grapes and hammers
18:38 as
18:40 metaphors
18:43 certainly
18:45 that's the ghost she had two BT 3.5 is
18:47 really fast let's embark on a fruity
18:49 journey into the realm of quantum
18:51 mechanics
18:52 wave particle okay
18:55 in quantum mechanics particles behave
18:57 like waves and vice versa it's as if our
19:00 grape particle can transform into a wave
19:02 Rippling through space so sometimes
19:05 we'll see our grape as a particle so it
19:07 got the grape part does it have hammers
19:09 in here anyway I don't know anyway so so
19:11 you can start out just doing funny stuff
19:14 like that you can also if you're a
19:16 little geeky you can say write me a
19:19 python
19:21 application
19:24 um
19:26 that brings
19:29 this idea to life
19:32 I don't even know if that's possible
19:34 let's see what it says
19:37 certainly here's the thing and there's
19:38 there's your python code look at how
19:39 fast that is Jesus Christ that's gotten
19:41 fast
19:43 um
19:44 so so let's see what it told us it's
19:46 going to do it says here's a python
19:47 application that uses grapes and hammers
19:49 as metaphors to simulate some aspects of
19:51 quantum mechanics it allows you to
19:53 interactively observe and manipulate
19:55 great particles with hammers okay so
19:58 there it's defined a bunch of things
20:01 it's got observe a great a red grape a
20:04 green grape measure the position measure
20:06 the momentum so it's written some pretty
20:10 sophisticated code in the flash of an
20:13 eye okay so that's the first way to get
20:16 into it is just start playing with it
20:18 the next way to get into it is what you
20:20 said
20:21 find a problem to solve so either solve
20:23 a problem for yourself like
20:25 um I want to eat better and not have
20:27 this neck that people comment on like
20:29 I've got a thyroid problem
20:32 um or
20:34 um you know I want a special diet this
20:36 week make me a meal plan and then make
20:38 me a shopping list or even better
20:41 pick a project for work or a project for
20:44 a community like you know church group
20:46 or you know local community and solve a
20:49 problem for them maybe they need a
20:50 marketing plan
20:51 use chat GPT to
20:54 create a marketing plan for someone or
20:56 or whatever it might be solve a problem
20:58 for them and and just be transparent Say
21:00 Hey I want to learn how to use chat GPT
21:02 I know you're trying to solve this
21:04 problem would you mind if I experiment
21:06 and try to solve that for you using chat
21:08 GPT and that will that will you'll learn
21:10 more doing that than anything you can
21:13 possibly read so I would I would
21:14 strongly encourage you to do that
21:17 all right
21:19 sales kiosks are already using it yeah
21:21 listen ai's been around for ages the
21:24 difference with Chachi EBT is it's
21:26 available to all of us the right of
21:28 strike I heard is getting worse yeah
21:29 well one of their issues is you know
21:31 they're saying to Studios hey don't use
21:34 um
21:35 AI to replace us and um
21:40 it's it's so short-sighted because what
21:42 the writer's Guild is saying is
21:44 we've got the most powerful writing tool
21:47 in the history of humanity literally at
21:49 our fingertips
21:51 and we're we're part of why we're going
21:53 on strike is we're saying we don't want
21:55 you to use that now
21:58 I've written seven screenplays like I I
22:00 am in support of the writer's Guild but
22:03 I think what they're doing right now is
22:04 just [ __ ] short-sighted and stupid
22:08 um and the these tools are going to be
22:12 everywhere right so
22:14 um how do they embrace it how do they
22:16 figure out how to to use these tools to
22:19 amplify their work and still get paid
22:21 like there's there's work to be done
22:22 here so I'm in support of them getting
22:24 paid but saying you know don't use these
22:28 tools is is just a very short-sighted
22:30 thing it's already a Microsoft edge
22:31 browser yeah Microsoft just announced
22:34 it's getting into the operating system
22:36 it can be used as a tool my students
22:38 um
22:39 have to discuss their ideas as well yeah
22:41 some of the there's a teacher here on
22:43 Tick Tock I forget her name but she's a
22:45 high school English teacher she's she's
22:46 just brilliant and she's like she she
22:49 realized in December this chat GPT thing
22:51 ain't going away and she just rolled it
22:53 into her classroom and she like you know
22:55 they'll have chechi PT write write the
22:58 paper in the classroom and then they
23:00 talk about it what did it get right what
23:02 did it get wrong what did it hallucinate
23:03 how would I write it better like she's
23:05 encouraging critical thinking she's
23:07 completely changed the dynamic of how
23:09 she interacts with the students I think
23:10 it's brilliant
23:11 education is going to be more homeschool
23:14 well I think the ability to learn
23:17 anything instantly is here like
23:21 anything you want to learn is literally
23:23 at your fingertips and you don't even
23:24 need to learn it like this was my big
23:26 Epiphany was chat GPT was I wrote in
23:29 about 90 minutes
23:31 I wrote
23:33 um I created a python application that
23:36 generated five pieces of social media
23:37 content from from a couple of inputs
23:40 that I asked for
23:42 um I've never written line of python in
23:44 my life I still haven't and it was a
23:46 functional application that called an
23:47 openai API generated results it was
23:50 amazing 90 minutes never have been
23:52 written it before
23:54 oh okay there's the link to the AI Drake
23:57 songs great
23:58 thank you for that it's not perfect yeah
24:00 yeah all this AI stuff is this is as bad
24:03 as it's ever going to be how it is right
24:05 now what if the AI music doesn't align
24:07 with the artist's viewer values um
24:09 they're gonna have to deal with that
24:10 because it's it's they're just going to
24:11 be fan generated shows one after another
24:13 that are coming after them
24:15 I'm more excited than anything great
24:17 what's scary why are top Executives
24:19 concerned well
24:21 okay
24:23 um
24:24 let's say you've got a business that
24:28 I know the agency business really well
24:31 and the agency business is built on top
24:34 of what's called time and materials so
24:37 um you tell me you want to do a bunch of
24:39 ads and I say okay that's going to take
24:41 me 50 people you know a Year's worth of
24:45 work you know at 120 an hour and that's
24:48 going to be your fee right and and if we
24:51 do more hours than that we'll just
24:52 charge you hourly right
24:54 and so
24:57 an agency today might charge twenty or
25:00 thirty thousand dollars for creating
25:03 some banners for
25:06 for social media
25:08 um because that's how the model works
25:10 but now all of a sudden you've got tools
25:12 like canva that's rolled in generative
25:13 writing generative design like like you
25:17 could generate those banners in
25:21 an hour
25:23 that historically would have taken you
25:25 two weeks or a month with two or three
25:28 people on it right however however long
25:30 that might be
25:32 um
25:33 so
25:35 you as the agency could use those tools
25:37 but it kind of threatens your existing
25:39 business model but you don't want to
25:41 give up
25:42 that Revenue so you're going to hold on
25:44 to that business model as long as you
25:45 can at some point
25:48 your client's going to come to you and
25:49 go you know I'm not willing to wait as
25:52 long or pay as much
25:54 so here's a 30 reduction in your fee
25:58 take it or leave it and agree to turn
26:00 stuff around in two weeks instead of a
26:02 month
26:03 and then what do you do so every single
26:06 part of the information economy whether
26:09 it's creative work or Consulting work
26:11 any kind of knowledge work
26:14 is is going to be impacted at kind of
26:16 10x kind of levels like things will be
26:19 done 10x faster 10x better or 10x
26:22 cheaper and anyone's who's trying to
26:24 maintain their the old model is just
26:27 going to fall apart the other thing
26:28 that's happening is this let's say
26:30 you've got a SAS company
26:32 and you've spent 10 years building the
26:34 SAS product that's built on this really
26:36 complicated system and database and you
26:38 figured out all the different ways to
26:40 write SQL queries and things like that
26:42 and all of a sudden someone can come
26:45 along with chat GPT with this code
26:46 interpreter thing take the same data set
26:48 as you and literally put in the hands of
26:51 a CEO here's your virtual consultant ask
26:54 it anything about that data set so the
26:57 thing that that CEO pays 20 grand a
26:59 month for now all of a sudden he or she
27:01 can do for free instantly at their
27:03 fingertips right so so the the concern
27:07 is that
27:08 how we do what we do and what we do and
27:12 what the value is of what we do is all
27:14 going to change it's all changing in
27:18 some Industries it's going to change
27:19 faster than others right industries that
27:22 have a lot of inertia like health care
27:24 and
27:26 insurance and and things that that are
27:29 sort of Highly regulated Financial
27:30 Services those are going to change
27:32 slower but even within that
27:35 the ability to create a presentation
27:37 where it used to take me a week to write
27:40 a PowerPoint presentation now it's going
27:41 to be built into Microsoft PowerPoint
27:44 write me a presentation based on the
27:47 quarterly earnings report bang it's done
27:50 all right what happens to the employees
27:53 that I used to pay to make that for me
27:55 I don't know
27:57 um so that's why they're concerned
27:59 I've been bored until I discovered chat
28:01 gbt I I agree I'm I'm more creatively
28:04 alive than I've been a long while we
28:05 need regulations on AI we do you know
28:08 Sam Altman was in front of Congress last
28:09 week saying that we need regulations
28:12 um I think we need someone at the
28:15 disposal of the government that has a
28:16 [ __ ] clue in fact in Sam Altman's
28:19 conversation with the government he said
28:20 he said yeah I think you need someone
28:22 you know in a row like that and and the
28:25 the congressman said well are you
28:26 available and he goes no I love my job
28:28 thanks
28:29 he said but we can you know I'm happy to
28:32 recommend someone
28:34 um how am I being creative
28:36 how am I being creative without AI I
28:39 don't know what that means well I in in
28:41 writing in
28:43 my my current company storybine it's a
28:46 video technology company
28:48 um
28:49 it's it's a lot it's a huge part of of
28:52 what we do it's a huge part of what I am
28:53 I've been a creative all my life and a
28:55 Storyteller all my life so it's a huge
28:57 part of what I do is chat apt able to
29:00 access the internet now yes so oh shown
29:02 this before
29:03 let me show you this so in
29:07 um
29:08 and you go to settings
29:12 you get this dialog box and the second
29:15 little option down here is beta features
29:17 and in that if you pay for the Plus
29:21 subscription there's web browsing and
29:23 plugins you have to turn those on so you
29:25 have to you basically have to opt into
29:27 these things
29:29 but once you're opted into them you now
29:32 have these different options in chat gbt
29:35 you've got default which is what it was
29:36 before browsing which is it's now
29:39 connected to the internet and then
29:41 plugins which I'll show you in a second
29:43 so this is what I was going to show
29:45 before when my internet went down
29:48 um so we're in chat GPT we're in
29:51 browsing mode and if I say Y is it so
29:55 smoky
29:56 in Denver today
29:59 question mark
30:02 so it now pops up this thing that says
30:04 browsing the web
30:05 so and then it says searched why is it
30:08 smoky in Denver today
30:10 the smoke is due to wildfires in Canada
30:12 the smoke is affected to shift Northeast
30:14 on Tuesday which should result in better
30:17 air quality with less smoke with a
30:20 citation to Denver weather from the Fox
30:24 news station
30:25 so yes
30:27 um chat GPT is connected to the internet
30:29 and then the other thing that it is is
30:32 it's got these things called
30:35 plugins and plugins you can choose
30:38 different kinds of plugins
30:40 so once you choose plugins
30:43 these are ones that I have installed and
30:45 you can turn on up to three but you can
30:47 install lots of them so if you go to the
30:48 plugin store
30:50 you get page after page after page of
30:53 all of these plugins
30:55 and they just you know describe what
30:58 they are and all that sort of stuff so
31:01 you can click through and you can
31:02 install them and then the way you use
31:04 them
31:06 is I'll get out of here well
31:09 the little x button the way that you use
31:12 them is you kind of you kind of turn
31:13 them on with this with this pull down
31:15 menu so what I have here I've got I'll
31:18 do I'll do I'll show you a cool one
31:21 um there's one here called prompt
31:23 perfect and so
31:26 you type perfect before each prompt and
31:28 then it will basically rewrite your
31:30 prompt so let's see so I'll say
31:33 um perfect
31:36 um
31:38 I need
31:40 Garden
31:42 recommendations for
31:48 early
31:49 summer in Denver
31:52 for veggies
31:56 so that's what I typed in
31:58 and then because I said perfect it
32:01 should pop up
32:02 the plug-in called prompt perfect
32:08 and it's doing nothing right now so
32:11 it may have crashed oh no there we go
32:14 using prompt perfect oh it actually now
32:17 tells me it's unverified that's
32:18 interesting so that means it's not an
32:21 officially sanctioned chat gbt
32:24 you know official one
32:27 um
32:28 but there you go Denver Colorado's in a
32:31 USDA hardiness Zone 5B which means blah
32:34 blah blah so so now it's giving me some
32:36 recommendations but if I open what this
32:38 prompt did
32:40 I typed in that short little sentence
32:42 and then what prompt perfect did was
32:44 said what are some specific vegetable
32:46 varieties that are best suited for
32:48 planting during the early summer season
32:50 in Denver and can you provide
32:52 recommendations for the most successful
32:54 and productive options so it rewrote it
32:56 so these plugins
32:58 um kind of extend what chat gbt can do
33:02 some of them are connected to the
33:03 internet some of them can read PDF
33:05 documents things like that so
33:08 uh that's how you use that stuff where
33:10 oh this Photoshop wait hold please there
33:12 you go
33:13 that's that little thing I had before so
33:15 why I have that up by the way is that's
33:17 the official chatgpt website
33:20 if you have not played with chat GPT I
33:23 strongly encourage it this is the AI
33:25 learning lab if you're going to be here
33:27 you have to learn Ai and while you're
33:30 listening to me ramble on like a [ __ ]
33:32 idiot why don't you go there and type in
33:34 some stupid [ __ ] and watch how
33:36 remarkable it is and then below it
33:38 futurepedia.io that's just a list of a
33:40 bunch of uh
33:42 AI tools if you really want to geek out
33:44 and dig deeper
33:45 what's my take on open AI CEO getting
33:48 the government involved
33:49 um I think he is being as responsible as
33:53 he can be
33:54 um you know we're in an environment
33:57 right now where meta leaked well
34:01 The Meta large language models escaped
34:04 into the wild whether they released them
34:06 or they were leaked or whatever it was
34:08 they're now out there
34:10 and so there are a bunch of Open Source
34:13 large language models like chat EPT that
34:15 have no restrictions on them they're
34:16 completely open sourced and and the
34:18 development Community is pouncing all
34:20 over them developing making them
34:21 stronger and stronger and stronger
34:23 so companies like open AI are going to
34:26 be looked to by industry and by
34:28 government to
34:30 teach us all how to be responsible and
34:33 how to grow these things so I think it's
34:35 pretty wise for Sam Walton to get on the
34:36 front end of this because he'll get his
34:38 ass dragged into Congress anyway if he
34:40 you know if he's not proactively doing
34:41 it so I think it's smart he's a very
34:44 thoughtful
34:45 he's a very thoughtful guy I was also
34:47 incredibly surprised that he he owns
34:49 zero percent of open AI which I I don't
34:52 understand I like I literally don't
34:54 understand it I don't know if that was a
34:55 conscious choice I assume it was
34:58 um
34:59 I don't think it's one of those things I
35:00 only take a dollar salary I assume he's
35:02 being fairly compensated
35:05 um but maybe he didn't want a conflict
35:06 of interest or like like I get the sense
35:09 that he really wants to be
35:11 um
35:12 as responsible as he can be for this
35:15 um Elon Musk hates what he's doing so
35:18 you know people don't agree with it and
35:20 there's a company called anthropic that
35:22 is using a completely different model
35:23 for how they're validating
35:25 um their systems um called the
35:27 Constitutional learning model which I
35:28 think is really interesting so I don't
35:30 think open AI is the only one but
35:31 they're clearly the biggest and the ones
35:33 they're the ones that caused all of this
35:35 Ruckus
35:36 because their chat GPT went to a hundred
35:39 million users in six weeks
35:42 six weeks to 100 million users fastest
35:44 software adoption in history
35:46 to put it in context the World Wide Web
35:48 took six years to go from zero to 100
35:50 million people
35:52 Chachi BT took six weeks that's why
35:54 everyone is panicking and everyone's
35:56 launching their own versions of uh this
36:00 stuff I I lost myself in my comments so
36:03 I'm trying to go back so if I miss your
36:04 comments I'm sorry
36:06 the Blockbuster bless the Blockbuster
36:08 error that's funny I live in Florida and
36:11 I'm using it to get now that's nice
36:14 please don't do political crap in here I
36:16 don't give a [ __ ] about like
36:19 uh
36:20 I I am so disgusted at the polarization
36:23 of this country like I I can't stand it
36:24 I don't give a [ __ ] I don't give a [ __ ]
36:26 yeah chat GPT is is is
36:29 left-leaning and everyone's out to get
36:32 everyone just don't bring it in here I
36:34 don't give a [ __ ] I'm just here to talk
36:36 about cool AI stuff
36:38 I mean seriously if you want to go do
36:41 that kind of political stuff go learn
36:42 chat GPT and have it write you a bunch
36:44 of you know propaganda you can do that
36:47 enjoy
36:50 just you know I think that one of the
36:52 things that this this makes possible is
36:54 that we can reconnect as human beings
36:56 and actually have empathy for one
36:57 another again wouldn't that be nice
37:00 what if the tools could do all the
37:01 shitty work the soul crushing work that
37:05 squashes everyone's humanity and we
37:08 could let the machines know that do that
37:10 and we were free to connect with one
37:12 another again and actually listen to one
37:14 another and be empathetic with one
37:15 another that'd be kind of cool
37:19 yeah
37:21 European Parliament already implemented
37:24 AI I don't think they've actually
37:25 ratified it yet they're talking about it
37:27 they will be the first if they do it and
37:30 and I listen I think I think regulation
37:32 is important very powerful tools
37:34 students and workers use it for reports
37:36 and research they're going to use it for
37:37 everything everyone's going to use this
37:39 stuff like if you use a computer you're
37:40 going to use large language models in my
37:42 personal opinion they want the
37:44 government involved to slow down
37:45 competitive websites well the shit's
37:47 open sourced already it's the the the
37:49 horse is out of the proverbial Barn at
37:52 this point what they need to do is say
37:55 this stuff is coming faster than we
37:58 realize it's more profound than we're
37:59 prepared for and we need to get our [ __ ]
38:02 together like that's what the government
38:03 needs to do right now is actually get
38:05 their head around what is happening and
38:08 how quickly it's happening they don't
38:09 understand what [ __ ] Wi-Fi is you
38:12 know their their hearings are
38:15 comedically bad if like tragically bad
38:19 um so they got to get their [ __ ]
38:20 together so they damn well better saddle
38:23 up and make nice with Sam Altman and and
38:25 Elon Musk and whoever the [ __ ] else out
38:28 there has resources
38:30 um
38:31 that can intelligently steer what we're
38:35 doing because I think it's actually
38:37 quite important we also can't just do it
38:39 as a country this is global like if we
38:41 decide as a country you know like we've
38:43 done with crypto and blockchain oh screw
38:45 it you know we're not going to let
38:47 anyone in America get ripped off
38:48 meanwhile the rest of the world is
38:50 innovating on this stuff and we're not
38:51 allowed to do it if we do that with AI
38:53 we're in [ __ ] trouble
38:55 so so we've got to lead with this stuff
38:57 but we've got to lead in a smart way so
38:59 and and listen it's part of the purpose
39:02 of this channel is to evangelize
39:06 curiosity like get curious about this
39:09 stuff start learning about it rather
39:10 than just going down these
39:13 sensationalistic you know rabbit holes
39:15 without having to use the tools use the
39:17 tools like see what they do see what
39:19 they can do for you start a new business
39:20 using them
39:23 um sorry I get so passionate I get so
39:25 worked up
39:28 oh good lord
39:31 all right
39:33 write me a report on a homework
39:35 assignment yeah that's easy
39:37 oh my God I can't spell chat GPT does
39:39 not care if you can't spell man it it
39:41 eats typos for lunch it does not give a
39:43 [ __ ] it just fixes them it doesn't
39:44 comment on it you put you put in your
39:47 your dumb spelling of whatever word
39:50 you've never been able to spell since
39:51 the fourth grade and it figures it out
39:53 for you and and corrects you that's
39:55 pretty cool how often do I go live I try
39:58 to do it once a day I haven't been on in
40:00 a while I got I got kicked off I got all
40:02 worked up on some live and I someone
40:05 said it might have been because I was
40:06 drinking apparently you're not allowed
40:08 to drink on lives so it might have been
40:10 because I was drinking and then I might
40:11 have just smartassed off on someone and
40:14 said something stupid I don't know so I
40:15 was off for a week but I'm trying to do
40:17 it every night so
40:19 this is back I live in Florida all right
40:21 yeah that was that one that's a bad
40:23 comment don't bring politics in here
40:28 is it available wait it's available on
40:31 the phone for the past month now no it's
40:33 available so the official chat GPT app
40:35 launched last week or yeah last week
40:40 so if you go to
40:41 um chat.openai.com no sorry if you go to
40:45 openai.com
40:48 slash hold please
40:52 openai.com
40:54 uh blog blog
40:58 and it's the second article
41:02 is introducing chat gbt for iOS so this
41:05 was put out on May 18th so last week
41:09 last week yeah
41:11 um that's the article and it's it's
41:13 available for iOS so
41:16 although it says on the App Store it
41:18 does it used to say iOS so I wonder if
41:20 it's available for
41:23 I don't know anyway the official chat
41:25 gbt app is out the the the the app does
41:29 not have
41:31 um does have whisper transcription
41:34 which a buddy of mine has been using he
41:36 said it is night and day better than the
41:39 local transcription in um
41:41 in iOS which is which is pretty pretty
41:44 sweet it's pretty sweet as we say down
41:48 home Farm
41:50 all right
41:52 yeah I have an iPhone just go through
41:53 the Safari browser yeah so if if uh
41:57 if you don't want to get the the
41:59 official chat GPT app just go to
42:02 chat.openai.com that's that's mostly
42:05 what I do I just bookmarked it on my
42:06 home page and use it but now I have the
42:08 app
42:09 um
42:11 the the website has more capability the
42:14 app's a little easier to read and stuff
42:16 like that is it's typical if something's
42:18 a native app it's it's easier if
42:20 machines can do the work of laborers
42:22 and GPT can do the work of artists
42:24 what's left to do that's the thing I
42:28 so
42:30 generative ai ai in general is going to
42:33 generate more jobs than it displaces
42:37 um
42:38 but there's going to be a period of
42:40 disruption that's going to be really
42:41 [ __ ] painful and so some jobs are
42:44 going to go away
42:45 but that attitude of like well if I
42:49 can't do my job what am I going to do
42:50 let's say this let's say you do some
42:52 sort of data entry job and you're being
42:54 paid 15 or 20 bucks an hour and what you
42:56 do for eight hours a day is is scrub
42:59 data to make sure that it's I don't know
43:02 better right and it's this manual
43:05 drudgery horrible Soul crushing process
43:09 and AI automates that out and and you
43:13 lose your job that's gonna suck
43:16 but
43:17 I think for a lot of people the response
43:19 is going to be well I [ __ ] hated that
43:23 job
43:24 let me play around with this AI stuff
43:25 that replaced me oh [ __ ] it can do that
43:28 you know what I've always wanted to do
43:30 and they'll come up with some idea and
43:33 they'll start a company or they'll get
43:35 together with some friends and and
43:36 support local businesses I think there's
43:39 going to be kind of a Renaissance of
43:41 innovation and startups
43:43 from people that were involuntarily
43:47 displaced from their current employment
43:49 but because these tools are made
43:52 available to everyone because it's
43:53 democratizing knowledge and power the
43:56 like these powerful tools are being put
43:58 back in the hands of people
44:00 I think we're going to see some
44:01 remarkable
44:03 um
44:05 human transitions right Transformations
44:08 um people going from being very
44:09 unfulfilled to being very fulfilled but
44:13 some people aren't going to make that
44:14 transition and it's gonna [ __ ] suck
44:17 and and the disruption is gonna [ __ ]
44:20 suck and I think it's gonna come fast
44:22 and I think it's going to come hard
44:24 um and I think it's going to be like a
44:26 giant reset button but quite frankly we
44:30 we just went through it like
44:32 um
44:33 coronavirus covid was a giant [ __ ]
44:36 reset button
44:37 like the whole idea the whole idea of
44:39 going into an office five days a week
44:41 now is
44:43 like the rules have been Rewritten
44:46 so and that was painful that was [ __ ]
44:48 awful and we adapted through it we
44:50 figured some stuff out and some people
44:51 are going back to the office and some
44:53 aren't and and you know that's going to
44:55 continue to evolve this is going to be
44:56 like like that it's going to be that
44:58 level of disruption so it's going to be
45:00 ugly but
45:01 everyone has access to these tools so
45:03 that's going to be pretty pretty amazing
45:05 all right
45:08 do they think fast food will start using
45:10 it in the future to operate the stores I
45:12 just heard Wendy's I think
45:15 um is using synthesized voice like
45:17 virtual drive through people so yes I
45:21 think they will do that and and probably
45:23 pretty quickly
45:27 um
45:28 great thing is like Googling something
45:31 without getting a bunch of ads yeah this
45:32 is this is so so one of the fascinating
45:35 things one of the fascinating little
45:37 tidbits uh from from the world of the
45:40 the history of this stuff Google in 2017
45:44 invented this thing called the
45:45 Transformer they wrote a paper called
45:48 attention is all you need
45:50 and it introduced this concept of the
45:52 Transformer and the Transformer what it
45:53 did was it allowed them to point these
45:55 large language models which were
45:57 historically you could only use
45:59 relatively small data sets to train them
46:01 and you had to train them on your very
46:03 specific data set and you can kind of do
46:05 cool stuff within that
46:06 what the Transformer did was it like it
46:09 let data scientists and and uh engineers
46:13 Point these large language models at
46:15 massive data sets and one of the things
46:18 that open AI did was they said wow this
46:20 thing seems to get more powerful the
46:22 more data we feed it so they put a
46:24 massive amount of investment into
46:26 finding data sources tagging data
46:29 sources getting data sources into the
46:31 the core model my nose itches that's
46:33 weird
46:34 um and uh
46:37 and and so Google invented the thing
46:41 that open AI used to create chat GPT
46:44 that is the fundamental threat to Google
46:46 search business
46:48 um I think Google knew full well what
46:56 what a chat interface like chat GPT or
46:59 like Bard would do to their search
47:01 business which is why they didn't go
47:03 first I guarantee you they built this
47:05 thing and went holy [ __ ] this screws us
47:08 up if people don't have to click 40
47:09 times to go find the [ __ ] they're
47:11 looking for if we just give it to them
47:13 that destroys our business model
47:16 so I have a feeling they just put it in
47:18 the back room behind the dark curtain
47:20 and let it gather dust for five years
47:21 while Sam Altman and this married band
47:23 of buddies
47:25 um created the [ __ ] super tool that
47:28 that put everyone on their heels except
47:31 Microsoft who invested 10 billion
47:34 dollars and said let's go [ __ ] party
47:35 and take out Google
47:38 Microsoft is having a time of their life
47:40 right now do you think uh let's see I
47:43 already saw these questions sorry about
47:44 that
47:45 um
47:46 [Music]
47:47 what does wasserstein
47:50 Divergence solve great question I have
47:54 no [ __ ] idea what that means but chat
47:56 GPT does so let's go ask it
48:00 all right
48:02 what is
48:06 the wasserstein
48:09 Divergence
48:14 it's also known as the Earth mover's
48:16 distance it's a measure of distance
48:18 between two probability distributions
48:20 okay God damn it I'm going to stop
48:22 generation and I'm going to be like I am
48:26 in seventh grade
48:29 [Laughter]
48:33 geez fancy words with me you
48:35 [ __ ] imagine you have a pile of
48:37 sand in the shape of a hill on one side
48:40 of your playground and an identical pile
48:42 of sand on the other now you're given a
48:45 task to move the first pile of sand to
48:47 the exact location and shape as the
48:49 second one you can move the sand in a
48:51 bucket the total amount of work you do
48:53 to move the first pile to the second
48:55 part Jesus even the easy explanation is
48:58 compensated
48:59 how many trips do you have to make and
49:01 how much sand you have to move each time
49:03 and how far you have to carry it's a
49:04 good analogy for the Western steam
49:07 distance oh I see okay
49:08 now if we talk about the wasserstein
49:10 Divergence we don't have a clear
49:12 playground analogy because the term is
49:14 not very common or well defined in the
49:16 field of mathematics why are you asking
49:18 me advanced mathematic questions on this
49:20 channel I'm just an idiot talking about
49:22 AI I'm not a mathematician meth I'm not
49:25 a mathematic methodical person
49:29 um the Divergence means how much more
49:32 one thing is different from another so
49:35 you could possibly Loosely think of it
49:36 like comparing how your different pile
49:39 of sand is from the other
49:40 I have no idea what how to answer your
49:44 question
49:46 and what problem the Divergence solves
49:49 but
49:50 we at least learn we know a little bit
49:52 more about it now I hope this helped
49:57 compared to callback Lieber okay you
49:59 Geeks uh
50:01 week out with each other that's fine ask
50:03 me the dumb questions those are the ones
50:04 I can answer all right
50:07 I'm actually familiar with
50:10 self-terminating games now you're all
50:12 just now you're just geeking and now
50:13 you're just picking on me I think you're
50:14 trolling me do you trust Sam Sam Altman
50:17 I you know what
50:20 I have not seen a single interview with
50:23 him
50:24 where I think oh that motherfucker's
50:27 evil like I like he he seems like an
50:31 incredibly thoughtful guy
50:34 um
50:35 he answers honestly when people say
50:37 things like should we be worried
50:41 and he says yes
50:44 you know
50:46 um
50:47 I think he's in a a decently
50:51 tough position when it like everyone's
50:54 got an opinion on how he should run his
50:55 business and and when you have something
50:58 like Chachi PT that got that popular
51:00 that quickly
51:02 um and and is as polarizing as this AI
51:06 stuff is
51:07 um you know I think he's got to tread
51:09 gently so
51:12 I haven't seen any evidence not to trust
51:15 him
51:18 um
51:20 that doesn't mean I do trust him but but
51:24 I think he's at least thinking very
51:27 deeply about all the issues
51:30 which is which is good you know I mean
51:33 there's a lot of companies like ask me
51:36 if I feel the same way about Google no I
51:38 think they're a bunch of manipulative
51:39 [ __ ]
51:41 uh you know Google is a bunch of
51:43 manipulative [ __ ] who had 92 or 94
51:47 market share of search so they didn't
51:50 have to compete they didn't have to
51:52 improve their product they could just
51:54 you know kind of rape and pillage
51:56 you know the business environment the
51:58 advertising environment because they had
52:00 all the power and the fact that that's
52:02 being shaken up right now is great and
52:04 it's sort of forcing them to to at least
52:07 on the surface act a different way but
52:09 like their core motivations have been
52:11 clear for decades so
52:14 um I trust Sam Altman a [ __ ] ton more
52:16 than I trust anyone there but he's also
52:19 not in
52:20 a position of extreme power like they
52:23 are
52:24 all the people making a report can now
52:27 think of ways to improve the business
52:28 exactly so that's that's my point is
52:31 that a lot of the low-level work that
52:34 you spend
52:35 a tremendous amount of your time on
52:38 is kind of mindless work like in every
52:40 position right you've got [ __ ] just
52:42 sifting through emails and
52:44 someone says oh can you whip me up a
52:46 PowerPoint with these couple of slides
52:48 on it it's like three days later you're
52:49 still like ah [ __ ] I gotta do that thing
52:52 if you could just knock that [ __ ] out
52:54 what's left well I don't know start
52:56 thinking about the problems you need to
52:57 solve
53:00 do you think that fast food will start
53:02 using an ass I answered that already I
53:05 want to understand your AI knowledge
53:06 what's the difference between BFS and
53:08 DFS I don't listen I'm not I'm not a a
53:12 computer scientist I'm not a machine
53:15 learning expert but what I am is I'm a
53:18 guy that was there
53:21 early on in the birth of the World Wide
53:23 Web
53:24 and I started one of the first digital
53:25 agencies a company called agency.com and
53:27 I was Chief creative officer there and
53:29 we grew that company from two people to
53:31 2 200 people in five years and so
53:35 I was there at the beginning of one of
53:36 these technological revolutions where
53:39 a technology that was previously
53:41 reserved for scientists and you know and
53:45 and researchers
53:48 because of Tim berners-lee's invention
53:50 of the World Wide Web all that core
53:53 technology was made available to
53:54 Everyday People
53:56 same thing just happened with chat GPT
53:58 where you've got Decades of machine
54:00 learning development and chat GPT just
54:03 unlocked Pandora's box and put it in the
54:05 hand of real human beings that's the
54:07 thing I'm interested in I'm not so
54:09 interested in
54:10 I want to understand how the models work
54:12 but I'm not the guy creating them so I
54:14 could Google BFS and DFS I don't know
54:16 the difference between that um my
54:18 knowledge of AI right now is how real
54:21 people and real businesses are going to
54:23 use it and and I don't have any real
54:25 expertise I'm learning it on the Fly the
54:28 reason I started this channel was to
54:30 essentially share the [ __ ] I was
54:32 learning and just get it out of my head
54:34 because I think it's important for
54:36 people to get curious about this and
54:37 explore it because we're not data
54:40 scientists but we are going to be
54:41 impacted by this [ __ ] so for me this is
54:45 about Everyday People figuring out what
54:47 the [ __ ] this stuff is how it's going to
54:49 affect our lives and to disrupt
54:50 ourselves before we get blindsided
54:52 upside the head by someone else
54:55 so I appreciate these stumpers in the
54:58 semi-trrolling
54:59 um but I don't appreciate it stop being
55:02 dicks
55:03 how does it generate so much information
55:06 if not connected to the internet because
55:08 so the core model the core
55:11 GPT model in the court Bard model
55:15 we're trained on effectively the public
55:18 internet so so they took as many days of
55:21 data sources as they could find and if
55:23 you listen to Sam Altman talk about this
55:25 he talks about a tremendous amount of
55:27 the work that went into building this
55:28 was identifying data sets and coding
55:31 them and you know getting them you know
55:33 embeddable into into these models
55:36 um so think all of Wikipedia think all
55:38 of the blogs think all of the public
55:40 websites that you don't need a password
55:41 to get into
55:43 um you know all of the you know publicly
55:45 accessible databases of local government
55:48 information all of the programming
55:50 languages all of the documentation of
55:52 those programming languages just think
55:54 like
55:54 [Music]
55:55 as much [ __ ] has been put on the
55:57 internet for the past 30 years
55:59 some decent majority of that
56:03 is what these things were trained on so
56:06 the core model stopped at September 2021
56:09 because they're really expensive to
56:11 create the core models the new chat EPT
56:14 with access to the internet kind of can
56:15 fill in that Gap it can now search the
56:17 internet
56:18 but that's why it knows so much it's
56:21 like you can almost think of Chachi PT
56:23 as if you've got a small business you
56:27 just hired an intern who's free
56:29 and happens to know everything
56:33 and is really humble about it
56:36 and you're like hey intern
56:38 um what's the wasserstein difference and
56:41 the intern's like oh
56:42 that's blah blah blah and you're like
56:44 that's too complicated and he's like oh
56:45 imagine two piles of sand right like
56:48 that thing we just did
56:50 um
56:51 that's kind of what it's like
56:53 and and now that it's connected to the
56:55 internet and it's got these plugins and
56:56 can do this all the all this other stuff
56:58 it's like that intern you know not only
57:01 knows all that [ __ ] but now he knows all
57:03 the current events and he can do things
57:05 like go Google things or book you
57:08 um tickets for for a flight or you know
57:12 get you a hotel or
57:14 go over to Wolfram Alpha and and execute
57:17 an incredibly complicated mathematical
57:19 function for you and bring you back the
57:21 answer in 10 seconds
57:25 write a business plan or find customers
57:28 for my woodworking yes so let's go let's
57:31 go play with that right now
57:33 um
57:35 so so the the big thing about
57:39 about these models is
57:42 think of it more like a collaborative
57:43 conversation than
57:46 a you're going to ask it for something
57:50 and give it to you back so let's say I
57:52 have a
57:54 wood working business
57:58 I'm just going to make some [ __ ] up in a
58:00 small town
58:04 and am thinking about doing
58:10 social media and expanding
58:14 my audience
58:17 online
58:19 please write me
58:23 a short
58:26 um
58:28 bulleted
58:30 business plan
58:35 that walks me
58:39 through the process I don't know
58:41 something like that
58:44 sure here's a concise business plan
58:46 aimed at enhancing your thing market
58:47 analysis
58:49 identify and understand your Target
58:51 customers Group location potential
58:53 demand study your competitors identify
58:56 online market trends discover what kind
58:58 of woodworking products are trending
59:01 um branding design a catchy and unique
59:03 logo for your business create a tagline
59:05 that speaks to it determine your Brand's
59:07 point of uh tone of voice website
59:10 creation social media strategy email
59:13 marketing
59:14 blogging and content marketing video
59:16 marketing paid advertising it's still
59:19 going
59:20 analytics and Improvement customer
59:22 service
59:23 all this sort of stuff so so let's see
59:26 let me find something here that's kind
59:27 of interesting
59:29 um
59:30 one of these market analysis things um
59:34 so I'm going to say
59:36 four
59:38 um
59:39 market analysis
59:44 um
59:48 what are
59:50 five unique ways
59:57 I can position
1:00:00 my wood working products
1:00:04 question mark
1:00:06 certainly positioning your products one
1:00:09 emphasize artisanal craftsmanship there
1:00:12 you go
1:00:13 eco-friendly and sustainable if you use
1:00:15 you know I only do things in bamboo
1:00:18 customizable products maybe you partner
1:00:21 up with a you know laser printing
1:00:23 company that's local and do customizable
1:00:25 products long lasting and durable local
1:00:27 sourcing and production so
1:00:30 you know I'm just doing these kind of
1:00:32 off the top of my head with zero
1:00:34 knowledge about your business but yeah
1:00:35 like you can
1:00:39 you you can tell it that like what's
1:00:42 important for you to tell it is what you
1:00:44 have and what you want to accomplish and
1:00:46 then you can tell it I don't understand
1:00:47 how to do this and if it gives you an
1:00:49 answer that kind of presumes you
1:00:50 understand something if you don't
1:00:51 understand it just go I don't understand
1:00:53 what you talk about when you talk about
1:00:55 this can you explain that more and
1:00:56 unpack that and then have it you know
1:00:59 it'll give you this list of ideas and
1:01:01 you can say oh yeah I want to I want to
1:01:03 focus on the sustainable thing can you
1:01:05 write me we can do that right now and
1:01:06 local sourcing okay
1:01:09 um write me
1:01:11 10 tag lines
1:01:14 that
1:01:16 emphasize
1:01:19 I didn't spell that right but it doesn't
1:01:20 matter
1:01:21 that emphasize
1:01:24 um
1:01:25 Eco that emphasize here I'm just gonna
1:01:28 this is a cool way to do it that
1:01:30 emphasize two and five that emphasize
1:01:32 two and five so it's going to remember
1:01:35 what those things were
1:01:37 and now it's gonna write me
1:01:41 10 tag lines
1:01:43 Nature's craft locally sourced the
1:01:44 eco-friendly choice in word working
1:01:46 crafted with care rooted in our
1:01:48 community sustainable woodworking at its
1:01:50 best from our backyard to your home
1:01:52 crafting sustainable woodwork locally
1:01:54 Local Roots Global responsibility the
1:01:57 green choice in Woodcraft right
1:02:00 and then if you and then if you're like
1:02:01 Okay
1:02:02 um
1:02:03 let's see these all sound like of uh
1:02:08 let's see uh avocado toast
1:02:12 eating
1:02:14 um
1:02:15 um um
1:02:16 liberal liberals
1:02:20 um tag lines
1:02:22 um can you make it
1:02:25 um
1:02:26 you know less
1:02:29 um hippie like
1:02:34 I assume it will know what this means
1:02:35 I've never actually tried this before
1:02:38 Hometown craftsmanship for better
1:02:40 tomorrow woodwork from our town the
1:02:42 world hard-working local craftsmanship
1:02:44 mindful of Tomorrow locally crafted
1:02:47 globally aware homecraft it's right so
1:02:49 so it kind of toned down the green thing
1:02:52 right like if you know being green is
1:02:54 somehow you know not what you want to do
1:02:56 so you can you can steer this thing in
1:02:59 in whatever Direction you want so
1:03:00 absolutely you can use it for your
1:03:02 woodworking business I would dive in
1:03:04 deep
1:03:05 all right what's your take on the Google
1:03:08 recent announcement which one there's
1:03:09 been so damn many they announced one
1:03:11 today
1:03:12 I think
1:03:14 um I think they released
1:03:16 the Adobe thing into bard
1:03:19 do you mean the the Google i o where
1:03:22 they made like 20 announcements I think
1:03:24 that
1:03:25 um
1:03:27 I think they were trying to redeem
1:03:29 themselves for the the Paris event they
1:03:31 did back in January or February where
1:03:33 they absolutely embarrassed themselves
1:03:36 um
1:03:37 they're playing catch-up
1:03:39 I mean Bard 2.0 or Palm 2 whatever is
1:03:45 underneath Bard is better than what it
1:03:48 was but it's still pretty shitty if like
1:03:51 just for doing like business stuff like
1:03:52 what I just did there
1:03:54 um gpt4 is way better there are some
1:03:57 things that Bard does better and it's
1:03:59 Google they're an engineering heavy
1:04:01 company so some of the coding stuff I've
1:04:03 heard is pretty good
1:04:05 um but I don't know I I think that
1:04:07 Google is trying their best to keep up
1:04:10 but it's fascinating watching a
1:04:13 massively dominant market player having
1:04:16 to play catch-up and it's also
1:04:17 fascinating watching Microsoft who just
1:04:21 could not compete in chat and couldn't
1:04:23 really touch what Google was doing now
1:04:25 threatening them it's fascinating it's
1:04:28 like and it turned it it turned in
1:04:32 like that it's amazing
1:04:35 now Microsoft have been working with
1:04:37 openai for a year like so we saw it
1:04:40 happen like that they've been working at
1:04:41 it for a while but they clearly saw the
1:04:43 opportunity Marshall mcloone has dozens
1:04:45 of Talks on YouTube there's no going
1:04:47 back in real time oh in regard to New
1:04:50 Media adapter Biz dies yep totally agree
1:04:52 if you don't know Marshall mcloone it's
1:04:54 worth going to watch that what browser
1:04:56 is that I'm using the edge browser
1:04:58 which that's the thing like I'm a long
1:04:59 time Microsoft hater I was an apple
1:05:01 Fanboy for years I still am down in my
1:05:03 heart but you know it ain't been the
1:05:05 same since Steve died
1:05:08 um
1:05:08 but um but Microsoft like I've got
1:05:12 um The Edge is my default browser and
1:05:14 Bing is my default search engine which
1:05:16 if you had if you'd bet me money in the
1:05:20 summer of 2022 that I would within a
1:05:23 year have that I would have bet you
1:05:24 money I would have lost that money
1:05:27 how to use chat CPT on an Android phone
1:05:29 just go to
1:05:32 your Chrome browser on your Android
1:05:34 phone and point your Chrome browser to
1:05:36 that address chat.openai.com that's the
1:05:39 official
1:05:41 chat GPT site go there just use that you
1:05:45 don't download any of the apps unless
1:05:47 it's the official I think it's only for
1:05:49 iOS right now for open AI there's an
1:05:52 official open AI chat GPT app but
1:05:54 there's a bunch of clones that are just
1:05:57 trying to rip you off so just go there
1:06:00 just go there
1:06:02 you can chat GPT five like advice on how
1:06:06 to deal with things like depression or
1:06:07 addiction yeah it's actually quite good
1:06:09 it's it's very well versed in in medical
1:06:12 stuff the thing you need to be careful
1:06:15 of with chat gbt is that it will if if
1:06:17 you're if you're looking to get
1:06:20 um
1:06:21 citations and things like that it will
1:06:23 hallucinate and just make [ __ ] up
1:06:25 um but it's also very very empathetic
1:06:27 it's the way it's programmed it's kind
1:06:29 of designed to give people what they
1:06:30 want and you know make them
1:06:32 feel empowered and feel better
1:06:35 um it's it's quite good at some of the
1:06:37 empathetic conversations
1:06:39 um you can also you could also seed chat
1:06:42 GPT so you can say if there's a
1:06:44 particular kind of therapy you like you
1:06:45 can say I want you to act as a jungian
1:06:47 you know psychologist or I want you to
1:06:49 you know do something with like Tony
1:06:51 Robbins whatever the language stuff is
1:06:54 Tony Robbins he's like you can kind of
1:06:56 aim it in different directions or you
1:06:59 can say to it
1:07:01 um explain to me the different kinds of
1:07:03 you know
1:07:05 um
1:07:05 therapy
1:07:07 um theories out there or Frameworks and
1:07:11 and have it have it teach you the
1:07:13 different Frameworks and then say okay I
1:07:14 want you to act like this one and I want
1:07:16 you to start a dialogue with me and I
1:07:18 want you to start asking me questions
1:07:19 and you know be supportive and and help
1:07:21 me through this and and yeah it's it's
1:07:25 remarkably powerful and I think we'll
1:07:27 get to a place very very quickly
1:07:29 where I could imagine
1:07:32 you could have a relationship with a
1:07:35 therapist and that therapist could have
1:07:37 a cloned version of themselves
1:07:40 where you're interacting with the
1:07:42 therapist but they're available 24 7
1:07:45 anytime you know of the day and maybe
1:07:47 it's like you know you pay an extra 50
1:07:50 bucks a month and you have access to
1:07:52 that therapist 24 7. and anything you
1:07:55 talk about with that therapist will get
1:07:57 synopsized and like anything where you
1:08:00 talked about your cat you know may not
1:08:02 go but it'll it'll pull out the
1:08:03 important part of the conversation
1:08:06 that'll get forwarded to the real
1:08:07 therapist and then if there's anything
1:08:09 critical in there that needs to be
1:08:11 escalated it'll automatically escalate
1:08:13 to that therapist so maybe you meet with
1:08:14 them in person you know your hour a week
1:08:17 or half an hour a week but then they're
1:08:19 available 24 7 as this virtual person I
1:08:22 think I think we'll
1:08:24 we will see that you know maybe not this
1:08:27 year but probably within two to three
1:08:29 whether or not
1:08:31 um insurance is going to pay for all
1:08:33 that [ __ ]
1:08:34 um that might be a different story but
1:08:36 if it's cheaper for them
1:08:39 um they might go for it so gonna be
1:08:41 interesting
1:08:42 thoughts on GPT writing books and
1:08:44 scripts I I had so it's it's really good
1:08:46 at that
1:08:48 um you can't put the whole thing in but
1:08:50 you can do little chunks at a time you
1:08:51 can have it work on your outline I I
1:08:54 dumped in
1:08:55 I had started a screenplay where I'd
1:08:57 written scenes one and two and I had the
1:08:59 the bones of an outline there and I just
1:09:02 pasted that all in and I said finish my
1:09:04 outline and it did
1:09:05 and then I said rewrite the outline as a
1:09:08 comedy and it did and then I said
1:09:09 rewrite the outline as a murder mystery
1:09:11 and it did
1:09:12 and then I had it write the third scene
1:09:15 I said write the third scene and it
1:09:18 sucked and I said no that sucked make
1:09:20 the girl do this make the boy do that
1:09:22 make the parents do this
1:09:25 um and it wasn't bad and that was that
1:09:27 was kind of all the farther I got with
1:09:28 it but but it can do pretty remarkable
1:09:31 stuff like the the writing that it did
1:09:32 was in the tone of the the scenes that I
1:09:34 had written
1:09:35 um so it's it's quite good at that you
1:09:37 can also the other thing I've done is
1:09:38 you can say combine the story Frameworks
1:09:42 of save the cat you know Sid Fields
1:09:44 screenwriting book and I don't know
1:09:46 whatever some other thing you did you
1:09:48 know and uh and and the uh hero
1:09:52 archetypes of Star Wars right take those
1:09:54 three story Frameworks combine them into
1:09:56 a single story framework and then write
1:09:59 an outline for this story and just paste
1:10:02 in your story your description of your
1:10:05 story and fill out that framework and
1:10:07 it'll do it and then you can go back and
1:10:08 forth and collaborate with it money you
1:10:10 can say Okay
1:10:11 um give me a bunch of characters okay
1:10:14 um
1:10:15 write the write write an opening
1:10:17 monologue from this character write an
1:10:18 opening monologue for all the characters
1:10:20 so yeah it's it's quite remarkable but
1:10:22 don't think of it like you put in a
1:10:24 prompt and out comes a screenplay think
1:10:26 of it more like you've got a writing
1:10:27 writing partner that always shows up and
1:10:29 doesn't suck
1:10:31 that's what it's like
1:10:36 all right it jumped me to the bottom
1:10:37 again so I lost my place so I may miss
1:10:39 some questions here so I apologize but
1:10:41 I'm going to go back and
1:10:43 see a few of these I also I'm I am
1:10:46 starting to fade because I got up at
1:10:48 three this morning uh in Norfolk
1:10:50 Virginia where I was speaking at a
1:10:52 military conference about generative AI
1:10:55 um and I got on a plane at 5 30 and flew
1:10:57 to Chicago and then flew to Denver and
1:10:59 then went to work and then came home and
1:11:00 took a nap and then did the AI Salon
1:11:02 tonight and then came here to do this
1:11:04 live so I'm a little crispy so if I
1:11:06 sound like an idiot it's because I'm an
1:11:08 idiot and I'm tired
1:11:11 is Chet EBT On The Wall Street Market
1:11:13 yeah I mean well listen Wall Street has
1:11:16 been using Ai and predictive AI for
1:11:19 [ __ ] years in probably decades
1:11:22 anything though they can get an edge
1:11:24 with they'll do it but degenerative AI
1:11:27 stuff now let's humble little you and me
1:11:30 say
1:11:31 um let me take this trading method that
1:11:33 Wall Street uses and turn that into a
1:11:35 bot and put in buying and selling
1:11:37 triggers
1:11:38 um
1:11:39 you know with stop losses just like the
1:11:41 big boys do and it'll just write you
1:11:43 that bot
1:11:44 so if you're gonna play with that [ __ ] I
1:11:47 would say play with sites where you can
1:11:49 paper trade or you know not use real
1:11:51 money
1:11:52 first and just just play with it but
1:11:54 yeah it's it's pretty remarkable
1:11:56 anything where you're taking kind of
1:11:58 technical stuff like that and recreating
1:12:00 it or merging it it's incredibly
1:12:02 powerful at
1:12:04 how to teach chat GPT to people who hate
1:12:06 Tech
1:12:07 um
1:12:10 let it solve a problem for them
1:12:13 um
1:12:16 don't even tell them that it was chat
1:12:17 GPT
1:12:20 um
1:12:22 have them tell tell you some
1:12:25 bullet points of a story of theirs and
1:12:27 then have it turn that into a short
1:12:29 story
1:12:31 um
1:12:33 let them give you ideas for
1:12:37 a children's story for their grandkids
1:12:39 and have Chachi BT
1:12:42 write a personal children's book from
1:12:44 Peppa
1:12:46 and include all the all your families
1:12:49 names and the pets names within it and
1:12:51 have it do that and then have that come
1:12:53 from from their grandparents like like
1:12:55 do stuff that
1:12:57 that doesn't amplify that it's AI doing
1:13:00 it just do do stuff that kind of shows
1:13:02 the magic of what it is
1:13:04 solve problems for people it's it's also
1:13:07 a great way to learn AI already escaped
1:13:09 it has already escaped
1:13:11 why doesn't the app pull info from third
1:13:14 party links like LinkedIn well
1:13:17 it it can now so so you it is connected
1:13:21 to the internet there are also now
1:13:24 plug-ins there's one it's called link
1:13:26 reader I think like that or something
1:13:27 like that
1:13:29 where you can give it a list of URLs and
1:13:32 it will go read the content of those
1:13:33 URLs and then can summarize them so with
1:13:36 with
1:13:37 um plugins and chat EBT you can start to
1:13:40 to read things like that it it wasn't
1:13:43 able to do that historically because it
1:13:45 wasn't connected to the internet now it
1:13:46 is karma has found Google I agree car
1:13:49 you know Google Google essentially did
1:13:52 what what Kodak did which is like you
1:13:54 know we're in a search business and
1:13:56 we're not going to violate that we know
1:13:57 search engine optimization is where it's
1:13:59 at and we own that and [ __ ] them and
1:14:02 that's come back to bite them in the
1:14:04 booty It Better Man Booty and their
1:14:06 Booty's a little sore right now and
1:14:07 everyone around my booties are
1:14:10 this sucks I hate being Google right now
1:14:13 why is everybody like Microsoft whoa
1:14:18 isn't that what they like I think that's
1:14:20 what they like all right if I want to
1:14:22 get into Tech where should I start
1:14:23 learning
1:14:26 there
1:14:27 go to chat GPT
1:14:30 go there
1:14:32 what do you want to learn what are you
1:14:33 good at do you have ADHD if if you have
1:14:35 ADHD like me you don't want to be an
1:14:37 engineer
1:14:38 but if you have good ideas for apps like
1:14:41 teach it to teach you how to do no
1:14:44 coding or low coding
1:14:47 um app development where you can just
1:14:49 come up with ideas and go put that [ __ ]
1:14:51 together without having to learn coding
1:14:53 I don't think you need to learn coding
1:14:54 these ideas so I would say
1:14:56 it like getting into Tech is just way
1:14:59 too broad a topic that's like saying if
1:15:03 I wanted to get into
1:15:04 you know working in the automotive
1:15:06 business how do I do that do you want to
1:15:08 be a driver do you want to design
1:15:10 engines do you want to do body work
1:15:12 do you want to sell cars you know what I
1:15:14 mean like like getting into Tech is
1:15:16 really broad but what I would say is if
1:15:19 there's a particular area of tech that
1:15:20 you're super interested in that you
1:15:21 think you'd be good at go here and just
1:15:24 start asking about it and have it teach
1:15:26 you to do it and if it's code have it
1:15:29 write code for you
1:15:30 Sam reminds me of a genius kid who built
1:15:33 something cool in his backyard then
1:15:35 thought
1:15:36 um yeah that's kind of what happened I
1:15:38 mean you know they built chat EPT in two
1:15:40 weeks this is this is another
1:15:42 fascinating part of the origin story
1:15:44 here what just happened oh my machine
1:15:45 just went
1:15:46 went down
1:15:48 um
1:15:49 they um
1:15:51 the instructions for how to create chat
1:15:54 GPT were in the gpt3 documentation for a
1:15:57 year available to the public
1:15:59 and Sam Altman assumed people would
1:16:02 build it and they didn't and then and
1:16:05 then Microsoft is going to announce this
1:16:06 billion dollar investment into open Ai
1:16:08 and Sam Altman at the beginning of
1:16:10 November said you've got two weeks go
1:16:12 make chat GPT
1:16:14 so they wrote it in two weeks and they
1:16:16 launched it in conjunction with that
1:16:17 announcement from Microsoft and it went
1:16:20 to a hundred million you it went to a
1:16:21 million users in five days and 100
1:16:23 million users in six weeks Sam Altman
1:16:26 nobody nobody expected it to do what it
1:16:29 did including Sam Altman
1:16:32 um so Not only was it like then thought
1:16:34 um it was like he then thought oh holy
1:16:36 [ __ ] what have I done like what have we
1:16:38 done uh because I think what he's
1:16:40 recognized is that it it turned [ __ ]
1:16:42 loose am I Tom Arnold I'm not I I was
1:16:45 also I was also told uh some of us
1:16:48 scrolling through Tick Tock and they saw
1:16:50 me and they're like I thought it was out
1:16:51 Alec Baldwin but fat
1:16:55 so Tom Arnold's lost weight yeah I kind
1:16:57 of look like I'm about like Tom Arnold
1:16:59 where you know he's lost all the weight
1:17:01 and shit's still falling and you know
1:17:03 he's trying to pick [ __ ] up
1:17:05 um so I'll I'll take that he's he's uh
1:17:07 he's a fine Feller uh ask chat Wyatt why
1:17:11 it was created with parameters to limit
1:17:13 its impact on human life
1:17:18 um okay I mean you can do that I Limited
1:17:20 wait why it was created with parameters
1:17:24 to limit its impact on human life a
1:17:28 would probably not answer that if it
1:17:30 knew it that like that seems like one of
1:17:31 those weird safety sort of things
1:17:34 um
1:17:35 but but also
1:17:37 um
1:17:38 yeah like I don't know that it would
1:17:40 comment on that like I'm happy to do
1:17:41 that I don't find that all that
1:17:43 interesting
1:17:44 um
1:17:45 the the thing I find interesting is what
1:17:47 anthropic's doing with the
1:17:49 Constitutional learning model where you
1:17:50 can create a set of rules and then all
1:17:52 of the answers that the the machine
1:17:53 generates are are measured against those
1:17:55 rules and it'll iterate them until it
1:17:57 gets it right I had no idea Tom Marl
1:17:59 knew so much about AI that's good that's
1:18:02 good I learned it all from Roseanne
1:18:05 um all right how do you feel about
1:18:06 training data sets being
1:18:08 comprised and scraped and still and
1:18:11 stolen data
1:18:13 um I think that so so said by someone
1:18:16 who's bitter and feeling threatened
1:18:18 that's fine I I totally get it
1:18:21 um it is threatening and and it's scary
1:18:24 all of these models here's the deal
1:18:30 stability AI did this mid-journey did
1:18:33 this
1:18:33 open AI did this Google did this [ __ ]
1:18:36 Google has been scraping our data for
1:18:39 [ __ ] 20 years so it's it's not like
1:18:41 this hasn't happened before
1:18:44 um it's just that this tool is so
1:18:46 powerful that we can all now see the the
1:18:49 depth of the data that's there if you
1:18:51 don't think Google has this amount of
1:18:53 data and then some you're high on crack
1:18:55 right
1:18:56 um
1:18:58 so how I feel about it is this
1:19:01 I think that copyright and trademark
1:19:04 issues
1:19:05 um we're going to spend the next year
1:19:07 the 10 to 20 years sorting out what this
1:19:10 actually means
1:19:12 um
1:19:13 I can see why people are upset but if
1:19:16 any one of these companies if stability
1:19:18 AI or mid-journey or open AI or Google
1:19:21 had said hey
1:19:23 we're thinking about creating an
1:19:25 artificial intelligence bot and we want
1:19:27 to train it on data can we have your
1:19:30 permission
1:19:33 the first answer would have been go [ __ ]
1:19:35 yourself and the second answer would
1:19:36 have been how much are you going to pay
1:19:37 me
1:19:38 so these models would not have been
1:19:40 created if if they had asked for
1:19:42 permission
1:19:43 so and and it's one of the reasons if
1:19:46 you look at what stability AI is doing
1:19:48 they're open sourcing all of their
1:19:50 models why
1:19:52 Here's my thought on that
1:19:55 they knew that the only way to get
1:19:57 really good quality image generation was
1:20:01 to train it on all the art which
1:20:04 included copyrighted art
1:20:07 um
1:20:08 I think they also knew they were going
1:20:09 to get their asses sued from here to
1:20:11 Timbuktu which they are like they're
1:20:14 Getty is suing them for like what
1:20:17 1.8 trillion dollars it's like a hundred
1:20:20 thousand dollars per image or some
1:20:21 stupid [ __ ] thing like that
1:20:24 um
1:20:24 so I think I think stability AI open
1:20:28 source their models because they
1:20:30 probably knew there was a decently
1:20:32 significant risk they're going to be
1:20:34 sued into Oblivion and might not exist
1:20:36 as a company but they still want the
1:20:37 data set to be out there
1:20:39 so
1:20:40 how I feel about it is it was a
1:20:43 necessary first step to get
1:20:45 models of this quality now that it's
1:20:48 happened and now that it's pretty clear
1:20:51 that oh [ __ ] well we need to figure out
1:20:53 attribution and compensation and things
1:20:55 like that now all of a sudden those
1:20:57 conversations are happening and then you
1:20:59 look at what Adobe did with Adobe
1:21:00 Firefly and they only trained that on
1:21:02 public domain images and images they had
1:21:05 licenses to from Adobe stock
1:21:07 where they own the license
1:21:10 um and so
1:21:12 Adobe is putting a model out there
1:21:14 that's not trained in a in a nefarious
1:21:18 nefarious way as you're characterizing
1:21:20 it
1:21:21 and
1:21:23 Adobe is also putting into place the
1:21:25 ability for artists to opt out of being
1:21:27 trained opt into being trained opt into
1:21:29 being paid right so they're going to
1:21:31 develop a whole system for this so I
1:21:33 think
1:21:34 the fact that these things were trained
1:21:36 had to happen or it would have never
1:21:37 happened nothing if you want to use
1:21:40 large language models and image
1:21:42 generation models in video generation
1:21:44 models and all of the models if you want
1:21:46 to use them in commercial business they
1:21:48 can't be risky so
1:21:50 Deloitte digital
1:21:54 had you know told their employees they
1:21:56 couldn't use generative Ai and then
1:21:58 Adobe came along and said well we've
1:22:00 created one that's safe and now they're
1:22:03 going to be able to use generative
1:22:04 images right and within delay digital
1:22:07 because it's now safe for commercial so
1:22:09 I think you'll see lots of layers of
1:22:11 this you'll see
1:22:12 kind of the Wild West at the bottom
1:22:14 where all this just [ __ ] is just trained
1:22:16 on everything and everyone's stealing
1:22:17 everything and making songs and then
1:22:20 you'll see kind of a middle layer where
1:22:22 it's like they're trying to transition
1:22:23 from the wild west into Corporate
1:22:25 America and then you'll see the
1:22:26 corporate America layer where it's like
1:22:27 here's all the safe [ __ ]
1:22:29 um but it'll happen pretty quick
1:22:31 is chat CPT On The Wall Street Market uh
1:22:35 they are not public they are for-profit
1:22:38 but they're not public about the writing
1:22:39 strike in Hollywood I think it's
1:22:41 short-sighted
1:22:43 that they are fighting against the most
1:22:45 powerful writing tool in the history of
1:22:47 mankind I think it's ironic
1:22:50 um and kind of sad actually I'm you know
1:22:52 I've written seven screenplays so I am I
1:22:55 am
1:22:55 in support of the writers and the guild
1:22:58 but I think their current moves are
1:22:59 stupid and short-sighted
1:23:01 um I gave it the lottery numbers for the
1:23:03 past 20 years and got prediction
1:23:05 patterns that's cool
1:23:08 um it's [ __ ] awesome it's like a big
1:23:11 brain and Futurama it is
1:23:13 what have you built so far with chat GPT
1:23:15 I have Bill
1:23:17 um
1:23:19 I built a thing called the social media
1:23:21 maker machine uh which asks you about a
1:23:23 topic and asks you a few questions about
1:23:25 it and then it generates a bunch of
1:23:27 pieces of social media content
1:23:29 um I I created a thing called the
1:23:34 um
1:23:35 challenge engine for
1:23:38 an organization I do some work for I'm
1:23:40 actually the chief generative officer
1:23:41 for this organization so if you go to
1:23:43 content evolution.net
1:23:47 um it's it's a Federation of marketing
1:23:48 companies and there's this thing called
1:23:50 The Challenge engine oops let me flip
1:23:52 over to here so you can see it
1:23:55 um and the challenge engine you you give
1:23:58 it a
1:23:59 um
1:24:00 a uh
1:24:02 business category you're in a challenge
1:24:04 you're facing what are the key forces
1:24:06 behind the challenge how's this
1:24:08 challenge impacting your business and
1:24:10 what results would you see if the
1:24:12 challenge is addressed and you input
1:24:15 um your your things into that your your
1:24:18 answers into the challenge engine and
1:24:22 challenge in gin
1:24:25 oops
1:24:27 Google does not like typos engine
1:24:31 let's see
1:24:34 challenge engine spreadsheet hold
1:24:36 challenge engine response okay so you
1:24:39 put that stuff in there and then it
1:24:41 writes you
1:24:43 um a response to your challenge so I
1:24:46 would like a patent for an invention of
1:24:48 mine and then it in different categories
1:24:51 gives you different strategic questions
1:24:53 you can ask yourself so not only does
1:24:56 this engine generate the content but it
1:24:59 also generates the HTML that that
1:25:01 formats the content and that's all done
1:25:03 in about 30 seconds
1:25:06 um so I've built that
1:25:08 um I've built for my
1:25:10 um
1:25:11 for my company we've Incorporated open
1:25:13 ai's whisper to do interactive
1:25:16 transcriptions
1:25:18 um I'm redesigning my entire platform to
1:25:20 be an omni Channel content generation
1:25:22 engine
1:25:24 um
1:25:25 I built a salon I built this Tick Tock
1:25:28 Channel I don't know I've built a lot
1:25:31 um can chat CPT find me cheap flights
1:25:33 yeah go to the kayak here you want to go
1:25:37 play and see it
1:25:39 we'll go to open AI we'll go to new chat
1:25:43 go to gpt4 plugins
1:25:47 and then I'm going to turn off prompt
1:25:51 perfect and I'm going to turn on kayak
1:25:54 and then I'm going to say
1:25:57 what is the cheapest
1:26:00 flight
1:26:02 flight
1:26:05 to New York City in June
1:26:09 uh wait from
1:26:13 Denver
1:26:17 um
1:26:18 so it's thinking it's thinking it's like
1:26:20 oh I've got to use kayak for that that's
1:26:22 a question that you know this this
1:26:24 plug-in could do
1:26:26 and then if I open this up we'll kind of
1:26:28 see what's it's doing under the hood
1:26:29 origin Denver destination New York City
1:26:32 departure date return date
1:26:35 so it has me leaving on June 1st and
1:26:37 coming back on the 30th
1:26:39 then it found all the flights
1:26:42 and then here are some options
1:26:47 returning it
1:26:50 443 dollars
1:26:52 on JetBlue Non-Stop
1:26:55 so it's giving me like you know full
1:26:57 itinerary for the flight
1:27:00 so there you go so yes yes it can
1:27:04 it's I'm telling you man this thing is
1:27:06 just the the more I use it the more look
1:27:09 at that look at what it did it's like
1:27:10 like there's your links to kayak where
1:27:12 it actually pulls in the thumbnails from
1:27:14 kayak and
1:27:16 it's just remarkable
1:27:22 someone said no it doesn't do creative
1:27:24 stuff I don't know what you're referring
1:27:26 to but if you're talking about chat TBT
1:27:27 you're wrong it does creative stuff
1:27:30 and I'm thinking about asking it to
1:27:33 write me a grant proposal I think Grant
1:27:35 proposal writing is genius use of chat
1:27:37 GPT
1:27:39 um I think it will make your life much
1:27:41 easier and potentially more effective
1:27:44 you can personalize your Grant responses
1:27:47 I guess you have to because you're
1:27:49 writing it to a specific Grant
1:27:51 do all Plus members have access to Plus
1:27:53 in oh plugins or are they starting to
1:27:56 roll them out no they're available to
1:27:58 everyone as far as I understand you have
1:28:00 to enable them so if you have chat GPT
1:28:03 Plus
1:28:04 you have to name at the bottom of the
1:28:06 three dots beside it and then that will
1:28:09 pop up
1:28:10 why is it not popping it up oh settings
1:28:13 you go to settings and then it pops up
1:28:16 this dialog box
1:28:18 and the second item is beta features and
1:28:21 you have to turn on web browsing and
1:28:23 turn on plugins and then over time what
1:28:26 you're going to see there is code
1:28:27 interpreter and there's one other thing
1:28:29 I think one of the high context window
1:28:31 things is going to be there so you need
1:28:33 to turn it on and then once you have it
1:28:34 turned on you just go start playing it
1:28:37 takes you a while to get your head
1:28:38 around the the plugins that they're kind
1:28:40 of wonky how they work but
1:28:42 uh it's it's just like
1:28:45 you kind of choose up to three plugins
1:28:47 that you want to work in tandem together
1:28:49 so what you're effectively doing is
1:28:51 creating like a little application where
1:28:53 you can start to query it and then it
1:28:55 will figure out if it needs to use any
1:28:57 of the tools that you put into into your
1:28:59 little application there so it's pretty
1:29:01 slick but it takes it's taken me a while
1:29:02 to get my head around it so I want to
1:29:04 check out the website I built on chat
1:29:06 GPT sure pop it in here
1:29:08 are you using extensions yet bro uh
1:29:12 extensions for What on in Chrome I I
1:29:15 kind of avoided playing with extensions
1:29:17 in Chrome
1:29:18 um just because
1:29:20 so many of them are just built on top of
1:29:22 the chat GPT or the gpt3 API and they're
1:29:27 they're just they're they're all kind of
1:29:29 features that are going to be rolled
1:29:30 into
1:29:31 chat GPT or you know the Microsoft
1:29:34 products or the Google products so I'm
1:29:35 just kind of waiting it out
1:29:37 I like it's it it is so entirely
1:29:41 overwhelming how many tools and how many
1:29:43 extensions and how many things are
1:29:44 rolling out like you just can't keep up
1:29:46 with it and and by the time you find one
1:29:48 you like you install it and then a month
1:29:50 later they're like sorry to announce
1:29:52 that we're shutting this down because
1:29:54 they didn't make any money on it because
1:29:55 it just got rolled as a feature in some
1:29:58 other product
1:29:59 so this technology is going to be the
1:30:02 death of us or the life of us it's going
1:30:04 to be one of the other what if it's both
1:30:07 can the new web search GPT help with
1:30:09 comparable comparable market analysis
1:30:12 and real estate I would say yes I would
1:30:14 say
1:30:15 the code interpreter tool which I don't
1:30:18 have access to yet
1:30:20 is probably going to be the thing where
1:30:21 you can do real real analysis but if you
1:30:24 wanted to do some basic analysis like
1:30:26 one of the plugins is
1:30:29 um
1:30:31 let me
1:30:34 Zillow do I have it installed no
1:30:37 but but somewhere in the plugin list is
1:30:39 Zillow and there's a couple other real
1:30:41 estate ones there so there are plugins
1:30:43 that have
1:30:44 um
1:30:45 you know hooks to real estate market
1:30:48 data so yeah you could start asking it
1:30:50 to analyze some of the things it's
1:30:52 pulling back but if you really want to
1:30:54 do like full-on serious data analysis
1:30:56 wait for code interpreter to come out
1:30:58 and then and then yeah that it's going
1:31:02 to do some cool stuff should teachers be
1:31:04 concerned it will
1:31:06 it will write their writing proficiency
1:31:08 test paper for them teachers should be
1:31:11 concerned
1:31:12 um but they should also be excited but
1:31:14 no teacher should assume that their job
1:31:16 will ever be the same
1:31:18 it's there this is going to be
1:31:19 completely different
1:31:20 I have no sense I agree I have zero
1:31:23 sense Edge uses a chromium driver yes it
1:31:26 does your preference barter chat gbt
1:31:28 chat gbt by far Bard Bard has some
1:31:32 features to it that are potentially
1:31:35 interesting but the core Bard model
1:31:38 still sucks for for what I use it for
1:31:41 for anything that's kind of writing and
1:31:43 more on the marketing side of things
1:31:45 Bard's horrible horrible
1:31:49 um it's I think it's pretty good at the
1:31:50 programming stuff though
1:31:52 but but you know all this stuff like
1:31:53 like I would say just go play like don't
1:31:56 don't listen to me I have a degree in
1:31:58 [ __ ] acting and you know I'm like a
1:32:00 advertising marketing guy so I'm uh
1:32:03 you know
1:32:05 figure out what you want to do first and
1:32:08 then and then go research what the
1:32:10 specific tools are better at
1:32:12 setup two computers running audio text
1:32:15 to talk enforce chat gbd to talk to each
1:32:18 other well it's funny there's a there's
1:32:19 a paper out there called uh generative
1:32:21 agents where they created a Sims
1:32:24 like game where they have 25 NPCs little
1:32:27 characters running around
1:32:29 um that use chat GPD to talk to one
1:32:32 another have memory they remember what
1:32:34 they say to each other from day to day
1:32:37 um and uh yeah I could see I could see a
1:32:39 world where you absolutely I'll have my
1:32:42 chat agent talk to your chat agent and
1:32:44 they go off and have conversations and
1:32:45 make plans and come back and tell us
1:32:47 what they decided we should do I think
1:32:50 that's coming
1:32:52 um
1:32:53 and if I'm behind if you're wondering so
1:32:56 one of the things that happens on these
1:32:57 lives is I now have 74 new messages I
1:33:00 haven't read
1:33:02 so I'm trying to read everyone's things
1:33:03 so as you're typing I'm not seeing all
1:33:06 the new [ __ ] I'm just seeing what I'm
1:33:08 looking at
1:33:09 um so if it seems like I'm not answering
1:33:11 you I'm not purposefully being rude I'm
1:33:13 just being rude by trying to be polite
1:33:17 if that makes sense
1:33:21 [Music]
1:33:27 Marshall mcluhan
1:33:30 Marshall mcluen I think it's MC but it
1:33:34 might be Mac Oh Let's go ask chat TBT
1:33:37 um
1:33:38 who
1:33:39 mcluhan
1:33:49 it seems like there might be a typo
1:33:52 m-c-l-u-h-a-n Canadian philosopher best
1:33:55 known for coining the expression the
1:33:57 medium is the message in the global
1:33:59 village
1:34:00 and predicting the World Wide Web almost
1:34:01 30 years before it was invented
1:34:03 mcluhan's work is viewed as a
1:34:05 Cornerstone and study of media Theory so
1:34:08 m c l u h a n
1:34:11 go watch videos on him
1:34:14 no offense but what would you know about
1:34:17 AI Geezer well that's offensive so go
1:34:20 [ __ ] yourself
1:34:23 I know whatever I know I'm teaching
1:34:25 whatever I know
1:34:27 um chat GPT can't calculate correct
1:34:29 googlebard does this with ease yeah like
1:34:32 I said Bart's going to be better at some
1:34:33 of the more technical stuff the jail
1:34:35 break prompts are better so what this
1:34:37 person is talking about is uh
1:34:39 jailbreaking chat GPT it's got you know
1:34:41 security things
1:34:43 um
1:34:44 people have created this this prompt
1:34:46 called Dan which stands for do anything
1:34:48 now and do anything now is um
1:34:52 is uh
1:34:54 uh you're basically tricking
1:34:57 um chat GPT to ignore its rules
1:35:00 um so that's pretty funny every time
1:35:02 they they get it to work then chat EVT
1:35:04 fixes it so it's this battle going back
1:35:05 and forth
1:35:08 um hugging face yeah hugging face is
1:35:10 cool although they've got it it is other
1:35:12 than GitHub like the the worst user
1:35:15 interface if you don't know what you're
1:35:17 doing like all of those sites GitHub
1:35:19 hugging face those sites that are these
1:35:22 massive repositories they all assume
1:35:25 you're an engineer and know what the
1:35:27 [ __ ] going on and have read all the
1:35:28 documentation they're they're really
1:35:30 hard to navigate
1:35:32 um Chuck hugging faces a little better
1:35:34 than GitHub but not much
1:35:37 but it's cool
1:35:39 the good news is you can have chat gbt
1:35:41 help you set that [ __ ] up GPT can't cite
1:35:44 clinical studies accurately yet no it's
1:35:47 really bad at citing [ __ ]
1:35:49 um although with the with the new
1:35:51 connection to the web it does citations
1:35:52 that are actually linked to the web so
1:35:54 you can see you can do some manual stuff
1:35:56 I think there's going to be a whole you
1:35:58 know rash of industries that figure this
1:36:00 [ __ ] out uh you know figure out the
1:36:02 citations thing and the accuracy thing
1:36:04 uh it's just it's not there right now
1:36:06 it's I describe chat GPS Chachi VT is
1:36:09 mansplaining as a service it will
1:36:11 confidently answer you even if it's
1:36:13 wrong
1:36:14 um
1:36:15 does this only work on Mac products
1:36:17 check EBT no it's just a website
1:36:20 digital just activate with web beta
1:36:23 search Digital I don't know what that
1:36:25 means
1:36:26 Bart is still experimental but if it's
1:36:28 1.0 I would use bard
1:36:31 why
1:36:33 it's well Bart is now 2.0
1:36:37 um the the Palm model that sits
1:36:38 underneath Bard they it's better listen
1:36:41 it got better it's just it's still not
1:36:43 anywhere close to gbt4 in my opinion
1:36:46 what's the difference between the free
1:36:48 and paid version so the free version of
1:36:50 chat gbt gives you access to 3.5 which
1:36:53 is super fast and still super powerful
1:36:55 you don't have access to four you don't
1:36:57 have access to web browsing with chat
1:36:59 gbt and you don't have access to plugins
1:37:01 so so the paid version gives you access
1:37:04 to the newer [ __ ] basically
1:37:08 all right there's a celebrity girl
1:37:10 that's cloned AI version of herself
1:37:12 people pay for it yeah I forget her name
1:37:14 she's a Snapchat model and she created
1:37:16 this thing called she's she's your
1:37:18 virtual girlfriend she charges a dollar
1:37:20 a minute she made 72
1:37:22 000 her first week she's predicting that
1:37:25 she thinks she's going to be able to
1:37:26 make five million dollars a month
1:37:28 by having cloned herself like you know
1:37:31 it's good to be a hot chick but with AI
1:37:33 I could turn myself into a hot chick
1:37:35 this would be a disturbing hot chick but
1:37:36 you know there you have it
1:37:39 all right
1:37:41 um if you need your own AI Avatar go to
1:37:43 personal.ai yep that's a good one and
1:37:45 did and Synthesia
1:37:48 chat EBT made an estimate for my
1:37:50 business today it was great awesome yeah
1:37:51 anything you can do to actually you know
1:37:53 support like solve problems for your
1:37:55 business like that's that's where I find
1:37:57 chat gbd gets really good it's it gets
1:37:59 very powerful at um that like the party
1:38:03 you push it the better it gets
1:38:07 all right let me dig back through some
1:38:09 of these trolls I've got more trolls
1:38:10 here than I've had in the past which is
1:38:12 cool welcome trolls thank you no one's
1:38:16 pointed out my nose hairs yet but feel
1:38:18 free
1:38:18 I've gotten I've gotten I should get my
1:38:21 thyroid checked out from my fat neck
1:38:24 um I've I've gotten called an old
1:38:25 clueless man which is true it's fair
1:38:29 but you know just don't be an [ __ ]
1:38:32 that's kind of my like rule in life like
1:38:35 this it's just it it's harder it's just
1:38:39 harder to be an [ __ ] like
1:38:41 just don't
1:38:46 ask chat apt what are the advantages of
1:38:49 to not being an [ __ ]
1:38:52 find find your ways good brothers and
1:38:54 sisters find your ways
1:38:57 all right what's the difference between
1:38:59 paid and free okay I'm back to where I
1:39:01 was sorry I got lost in my things there
1:39:03 all right if I need my own AI okay sorry
1:39:07 sorry about that sorry get on yeah
1:39:11 brother brother
1:39:15 hello
1:39:16 um
1:39:18 is the paid version worth it I think
1:39:20 it's worth it I'm not listen I'm I'm
1:39:21 paying for very little stuff as little
1:39:23 as I can I pay for zapier
1:39:26 I pay for chat gbt
1:39:28 I pay for Poe if you don't know Poe
1:39:32 I was going to show this earlier
1:39:37 I lost it hang on oh I I typed over it
1:39:41 let me go to po.com poe.com
1:39:44 and it's just a chat engine but it's got
1:39:47 lots of chat engines so you see here
1:39:49 there's gpt4 Claude plus Claude instant
1:39:53 100K that one in particular what that
1:39:56 means
1:39:57 is you can put in 75
1:40:01 000 words as your prompt and it will eat
1:40:04 it and and generate it back what that
1:40:07 basically means is you can
1:40:09 put in a whole novel and tell it to
1:40:12 rewrite it as a
1:40:15 you know
1:40:16 iambic pentameter poem and it will [ __ ]
1:40:20 [ __ ]
1:40:22 um so I pay for that
1:40:23 I pay for mid-journey
1:40:26 um I pay for Google collab to do some
1:40:31 stable diffusion stuff
1:40:32 and some of the other uh like like Lang
1:40:35 chain and things like that I've been
1:40:36 playing around with it with that so so
1:40:39 some things are are worth paying for but
1:40:41 man I'm I'm trying not to because
1:40:43 there's just so many things 10 bucks a
1:40:45 month here 10 bucks a month there before
1:40:47 you know it
1:40:49 um you can't be an AI anymore
1:40:53 um I'm working with the University of
1:40:54 Redlands to look at to see how AI can be
1:40:57 developed with pedagogy pedagogy I don't
1:41:01 know how to say that
1:41:03 um that's cool very interesting I like
1:41:05 your Positive Solutions thank you I try
1:41:07 to keep it positive here
1:41:09 we get the no [ __ ] rule
1:41:11 I'm curious if it can create new Pharma
1:41:13 drugs yeah there's a thing called Alpha
1:41:14 fold so deepmind which was a company
1:41:16 that Google bought
1:41:18 um created this thing called Alpha fold
1:41:20 and it created like
1:41:23 accurately predicted 400 000 or 4
1:41:26 million like some crazy amount of
1:41:28 proteins that that traditional
1:41:30 researchers would take like two months
1:41:32 to do one and it did like four hundred
1:41:34 thousand in like two weeks
1:41:36 um so yeah I think I think we're we're
1:41:38 about to enter a a an age of uh
1:41:43 scientific advancement like we've never
1:41:46 seen because because the the tools are
1:41:48 going to start learn
1:41:51 improving themselves and and we're going
1:41:53 to start to see the the
1:41:55 quality impact go up dramatically and I
1:41:58 think Pharma drugs is going to be an
1:41:59 area we're going to see incredibly quick
1:42:02 quick advancements over the next five
1:42:03 years and deep
1:42:05 you know things like like you know not
1:42:10 being hyperbolic but like very plausibly
1:42:13 things like cures for cancer cures for
1:42:16 genetic diseases you know just crazy
1:42:19 crazy stuff
1:42:21 um but again that's coming from a guy
1:42:23 with an acting degree so I know people
1:42:25 have been predicting that for a long
1:42:26 time but I I have never in my life seen
1:42:29 the level of acceleration that we're in
1:42:31 the middle of right now I've never seen
1:42:32 anything like it it's insane I'd love to
1:42:34 hear how I feel about the Gemini
1:42:36 announcement what's the Gemini
1:42:37 announcement was that a code name for
1:42:39 something that I know about hang on
1:42:42 let me figure out what the Gemini
1:42:43 announcement was what
1:42:45 is
1:42:48 the Gemini
1:42:51 gem in Nye
1:42:55 uh announcement
1:43:01 Gemini zodiac signs
1:43:04 I don't think that's what you're talking
1:43:06 about I'm a Taurus
1:43:08 uh I don't I forget what that is remind
1:43:10 me what it is and if I know about it
1:43:12 I'll talk to it if I don't I'll just
1:43:14 I'll go look it up
1:43:17 um this is paid one okay yes
1:43:19 I just spent the past week learning to
1:43:21 code solidity and chat GPT made it so
1:43:24 much more efficient yeah it's just
1:43:25 remarkable having studied pharmacology
1:43:27 in school I'm wondering if they'd help
1:43:29 absolutely I actually had it um I had a
1:43:32 reverse engineer I've got family members
1:43:35 with chronic pain and we had this
1:43:36 natural kind of chronic pain juice that
1:43:39 they took
1:43:40 um and I had I had it um explained to me
1:43:43 what all the ingredients were the active
1:43:45 ones and the inactive ones I had it
1:43:47 recommend additional ingredients based
1:43:49 on their specific symptoms and then I
1:43:51 had it write me the recipe for how we
1:43:53 would actually create it and it like it
1:43:56 was remarkably good at that so I would
1:43:59 say do all of that you know do your own
1:44:01 research once it recommends it but we
1:44:03 went like once it once it recommended
1:44:05 everything we went and we looked at the
1:44:07 actual ingredients and went out to the
1:44:10 web and found things and it was actually
1:44:11 pretty accurate there wasn't anything in
1:44:13 it that that scared me which was pretty
1:44:14 cool
1:44:15 yeah hugging face chat is getting better
1:44:17 yeah hugging face chat is a is a chat
1:44:20 GPT competitor and you're gonna be able
1:44:22 to use multiple models and and and it's
1:44:26 going to be pretty slick so um I think
1:44:28 hugging faces is a potential big player
1:44:30 they're just they're a little early
1:44:31 right now because everything's so wrong
1:44:33 and geeky do I want to check out our
1:44:36 website that I built I think I said yes
1:44:37 to that before just put it in there and
1:44:39 I'll I'll take a look QA here learning
1:44:41 job at the moment can chat uh chat gbt
1:44:43 helped a lot yeah it's great great at
1:44:45 teaching um if you want to look at uh
1:44:47 teaching stuff go to Khan Academy Kahn
1:44:49 Academy
1:44:51 um they just launched a gpt4 powered
1:44:53 teaching assistant called conmigo that
1:44:56 uses Socratic reasoning to teach people
1:44:59 it's pretty pretty [ __ ] slick
1:45:01 interior design AI is so good it is I
1:45:04 think we're going to see just amazing
1:45:06 designs in the next few years inspired
1:45:08 by AI room GPT I don't know that but
1:45:10 that's probably pretty cool
1:45:12 I knew nothing about solidity before I
1:45:15 started now I'm almost fully understand
1:45:16 how to read it that's super cool yeah
1:45:19 and you can have you can have chat GPT
1:45:20 write you a solidity contract write me
1:45:22 an nft
1:45:23 you know contract that does X Y and Z
1:45:25 doesn't do this only certain people can
1:45:27 buy in
1:45:29 what's the best way to utilize GPT to
1:45:31 upscale my photography business
1:45:34 I would start learning so my my co-host
1:45:37 of the AI Salon is a professional
1:45:39 photographer of about 25 or 30 years
1:45:42 um I would strongly suggest you start
1:45:46 learning stable diffusion or mid-journey
1:45:48 and start trying to teach yourself to
1:45:51 recreate your style using those tools
1:45:53 because I think that's a skill that
1:45:56 you're going to need to know how to use
1:45:57 it's also going to scare the ever living
1:46:00 [ __ ] out of you
1:46:02 um but it's okay lean into it on the
1:46:06 business side of things I mean it can
1:46:08 help you in a lot of ways have it write
1:46:09 a marketing plan you know I I would say
1:46:12 have chat EPT help you come up with
1:46:14 strategies for how to maintain your
1:46:17 photography business rates in a world
1:46:20 where
1:46:22 um
1:46:23 photography is going to be replaced by
1:46:25 AI generated photography
1:46:27 how can you maintain an edge and how can
1:46:29 you survive in there so so I would I
1:46:32 would use it like a business consultant
1:46:34 and ask it really tough questions but I
1:46:37 would also strongly encourage you to get
1:46:39 get up to speed speed on
1:46:42 um mid-journey and or stable diffusion
1:46:44 uh I'm phenomenal thank you for the kind
1:46:47 words I appreciate that in China one
1:46:49 billion people no one gives any
1:46:50 permission and the tech companies are
1:46:52 everywhere yep
1:46:54 what are my favorite gpt4 plugins I
1:46:56 haven't played with a ton of them I saw
1:46:58 a really cool thing
1:47:00 um it was a tick tock someone did and it
1:47:02 combined three of them one of them was
1:47:04 called link reader and one of them was
1:47:06 something else but basically
1:47:07 you could ask it a question and give it
1:47:09 a whole series of web pages and it would
1:47:11 go read all the web pages and then
1:47:13 comment on them which is that's really
1:47:14 slick
1:47:16 um there's one called prompt perfect
1:47:18 which is pretty cool you type perfect
1:47:20 before your prompt and it rewrites your
1:47:21 prompt for you
1:47:24 um the web browsing stuff's pretty good
1:47:26 I just I haven't dug deep into into
1:47:27 plugins because a there's a bunch of
1:47:29 them
1:47:31 um they're also some of them are just
1:47:33 really hacky and horrible so it's hard
1:47:36 to know what's worth putting time into
1:47:38 and also it's it's almost like
1:47:43 it's almost just like using tech for
1:47:44 Tech's sake at this point like what I
1:47:47 want to do is figure out a problem I
1:47:48 have that I need specific plugins for
1:47:51 and then put them together and go try to
1:47:53 solve that problem so I'm a little
1:47:55 clueless right now on the plugins
1:47:59 um Windows co-pilot announced today yeah
1:48:01 I think it's a big deal and Apple's
1:48:03 developer conference in two weeks I I'll
1:48:05 be incredibly disappointed if Apple
1:48:07 doesn't announce some really interesting
1:48:09 things one thing that I know they're
1:48:10 announcing is built into iOS 17 as part
1:48:13 of accessibility will be voice cloning
1:48:15 you'll be able to clone your voice at an
1:48:18 operating system level on iPhones
1:48:21 um and iPads which I think is pretty
1:48:23 remarkable
1:48:24 have I read wolfen wrote wolfram's piece
1:48:27 on chat GPT
1:48:28 um I haven't read his piece I've seen a
1:48:30 couple of interviews with him where he
1:48:32 was talking about it
1:48:33 but um no I'll go look at it I'll go
1:48:35 look for it
1:48:37 and I have ADD so I'll probably have
1:48:39 chat GPT summarize it for me so I don't
1:48:41 have to read it
1:48:43 because that's how I roll how do I
1:48:45 convert AI generated text into normal
1:48:47 text to escape detection I don't give a
1:48:49 [ __ ] about detection I really don't I
1:48:50 think all this detection bullshits this
1:48:52 fleeting moment this these large
1:48:54 language models are going to be in
1:48:55 everything so
1:48:57 I don't generally just say hey chatgpt
1:49:01 write me something and then take what it
1:49:03 wrote and
1:49:05 paste it somewhere else like what I'll
1:49:06 do is I'll say oh I need to write a
1:49:09 piece on this I'm like uh write me an
1:49:11 outline that covers this this and that
1:49:13 topic what else am I missing and it
1:49:15 gives me that okay give me a 750 word
1:49:18 article that does that oh no that's too
1:49:20 dry can you give use metaphors and like
1:49:23 so I go back and forth with chatgpt a
1:49:25 lot and then I take
1:49:26 what I consider kind of 80 good writing
1:49:30 and then I kind of edit it and as I have
1:49:32 ideas within there I'll have chat gbt
1:49:34 refine certain things
1:49:35 um so by the time I'm actually done I
1:49:38 don't know what's mine and what's chat
1:49:39 EBT so I think any good writers that I
1:49:42 know are doing it like that and I don't
1:49:44 give a [ __ ] about detection if something
1:49:45 if someone decides that my shit's
1:49:48 generated from that I don't care in fact
1:49:51 Google reversed their stance on that
1:49:52 they initially said that they were going
1:49:54 to penalize
1:49:55 um AI generated stuff and then they
1:49:57 realized that was a shitty idea and then
1:49:59 now that what they've said is they're
1:50:01 going to penalize stuff that's not
1:50:02 useful so if you produce AI generated
1:50:05 content that's useful cool
1:50:08 that's what Google said
1:50:13 though yeah
1:50:16 all right
1:50:23 uh
1:50:24 [Music]
1:50:28 all right use chat CPT constantly I'm
1:50:30 starting to find responses generic do
1:50:32 you have prompt tips
1:50:34 um
1:50:35 you need to tell it to be less generic
1:50:37 like if you give a generic prompts it
1:50:39 gives you really generic answers
1:50:41 um
1:50:42 think about multiple layers so tell it
1:50:48 like focus it like if you want it to be
1:50:50 really good at investigative journalism
1:50:54 that style of writing tell it it's an
1:50:56 investigative journalist and then have
1:50:59 it ask you questions and then
1:51:02 take the responses that it gave you and
1:51:04 refine them and then tell it to write a
1:51:05 piece based on that so think of it more
1:51:08 like you're adding layers to a pie or to
1:51:10 a cake rather than just getting
1:51:12 something from it if you just give it a
1:51:14 generic prompt it's going to give you
1:51:15 generic answers
1:51:18 um how do you get to Chachi pt4 so that
1:51:20 I can take advantage of the plugins you
1:51:22 go to that URL chat.openai.com
1:51:25 you pay their 20 bucks a month and I
1:51:27 don't get a commission for that pay 20
1:51:28 bucks a month
1:51:29 for their plus and then you enable
1:51:32 plugins in the beta features you go to
1:51:35 your settings and turn it on
1:51:37 all right
1:51:39 um I need to run so let me go do that
1:51:42 and I hope you're all doing well and it
1:51:46 was good to see you all again and I will
1:51:48 check in probably tomorrow peace out