aiLL 6/17/23 - Generative AI In Simple Terms

Description

Chapters

Transcript

0:01 foreign
0:19 's wife what's happening
0:23 you're in here first you're waiting for
0:25 me look I got a new background huh
0:30 Professor Shannon yeah I'm tenured they
0:35 can't fire me now
0:41 oh
0:42 a couple of us in here now all right
0:45 happy Saturday night
0:50 um yeah they took away my double camera
0:52 so I tricked them with a green screen
0:54 actually the only thing I don't know let
0:56 me try something here so if I flip my
0:58 camera
1:00 then I have to oh God
1:06 wait where are my
1:09 I gotta be able to turn off my Bluetooth
1:15 so we interact subscribe share enhance
1:19 effects
1:21 okay I gotta do that good Lord
1:27 all right and then I gotta go back here
1:29 and I gotta flip it
1:32 all right well flip of my camera is
1:34 going to be a bit of a pain in the ass
1:35 but this is better than
1:37 than uh what I had before all right
1:39 looks like there's a few people here hey
1:41 everybody my name is Kyle Shannon this
1:42 is the AI learning lab
1:44 as evidenced by the uh AI learning lab
1:47 underneath all the crap at the top of
1:49 the screen
1:49 man Tick Tock in their interface is just
1:51 special isn't it
1:55 um
1:55 in here we like to talk about AI I'm an
1:58 entrepreneur for many years and uh I am
2:02 I've been deep diving into this AI stuff
2:04 for about a year
2:06 um looking at it more on the how we use
2:09 it side than on the creating the
2:11 technology side so so my background is
2:13 more as an entrepreneur and using
2:15 Technologies like this
2:17 I was an early player in the in the
2:19 early days of the World Wide Web I
2:21 started one of the one of the first
2:22 digital agencies in the mid 90s called
2:24 agency.com and the parallels between
2:28 what was happening in the early days of
2:30 that and what's happening with
2:31 generative AI right now
2:33 are insane and uh I think that this
2:38 generative AI stuff's actually going to
2:39 be bigger and more profound and happen
2:41 way faster than that web stuff did so uh
2:44 so anyway that's that's who I am if you
2:46 have questions about uh anything AI
2:48 related pop them in the questions below
2:50 if you're new to generative AI my whole
2:52 background here
2:54 um these three URLs you should go try
2:57 the top one is the official chat GPT
2:59 website
3:00 bing.com is Microsoft's
3:04 um search engine but if you click the
3:06 chat button that's gpt4 connected to the
3:08 internet
3:10 and then pro.com I can show you that
3:12 later it's got a bunch of different
3:13 models you can play with including some
3:15 really cool ones from anthropic
3:16 including a hundred thousand token model
3:19 which is 75 000 words so that's pretty
3:22 cool
3:23 um I was using that today actually how
3:25 do I think Quantum Computing will affect
3:27 AI I think
3:30 I don't know enough about Quantum
3:31 Computing to know how fast it's coming
3:33 but it sounds like they're doubling the
3:36 number of qubits every year and they're
3:37 up to like 500 and something and and
3:40 like every time they double every time
3:42 they add a qubit it's it's an
3:44 exponentially more powerful system so
3:47 they're up to 500 so so I would think
3:48 it's starting to get pretty pretty damn
3:50 powerful
3:51 um I would think that's going to help a
3:53 lot on the training side uh I I that my
3:56 guess is that's where it's going to be
3:58 but
3:59 I think it can't do anything but allow
4:03 these AI tools to continue building on
4:06 themselves and just accelerate
4:07 accelerate the advancement
4:10 um so yeah all right
4:14 same same I started in SAS in 2003 just
4:17 sold that and now I'm starting AI SAS oh
4:20 AIS that's good AIS
4:25 AI as a service I like to call chat GPT
4:28 mansplaining is a service Maas MS
4:34 because it hallucinates so much it will
4:36 confident confidently answer you you
4:38 even if you're wrong uh if you're new
4:40 here my name's uh Kyle Shannon this is
4:42 the AI learning lab go to these URLs
4:45 right here then these lower two are are
4:47 um prompts.chat will teach you a little
4:50 bit about prompting there's a little bit
4:51 of a primer there on how to prompt and
4:53 then there's just a whole series of you
4:56 know act as a sportscaster act as a
4:58 screenwriter act as an accountant that
5:00 will sort of put chat gbt into modes and
5:03 it gives you a good sense of the kind of
5:05 flexibility the tool has and
5:07 futurepedia.io that's just a bunch of
5:09 it's a directory of lots and lots of AI
5:10 tools
5:12 um if you would uh tap the screen a
5:14 bunch so we can let people know we're
5:15 here
5:16 um share this live I just went live
5:18 happy Saturday night everybody
5:21 um what else I have a lovely
5:23 um carbonated water here Blackberry
5:26 lemonade if you're wondering the flavor
5:29 um
5:31 tasty
5:33 thanks thanks for all the love
5:36 um and what else uh feel free to ask any
5:39 questions about ai ai I'll show what I
5:41 can if there's if there's stuff that's
5:43 worth demoing I'll answer any questions
5:45 I'll pontificate I'll yell at the the
5:48 doomers who say the robots are gonna
5:50 kill us
5:51 because that's not what this channel is
5:53 about
5:53 um if you're going to troll me be
5:55 creative about it
5:57 um my favorite so far was you look like
5:59 Rosie O'Donnell which when you get a
6:02 troll that's not quite a troll but it's
6:04 close
6:06 that's a good drill that's a good troll
6:08 if you can't quite figure it out
6:10 all right let's see
6:16 will it let you change the amount of
6:18 screen you take up when you want to show
6:21 more green screen
6:23 um
6:25 I don't think so
6:28 I don't know I don't like I can't pinch
6:32 it like that when I'm doing normal green
6:34 screeny stuff
6:39 and encourage more viewers to share I
6:43 think I just did that bro what's the AI
6:46 wait
6:48 what's the AI that makes apps for you
6:53 um
6:54 I don't know but
6:57 Chachi BT can can code if you're talking
7:00 about a specific app
7:02 that uses chat GPT to to build and
7:04 compile uh apps I'm so so there's it
7:10 it's a bit too generic a question there
7:13 there are a number of no code app
7:15 building tools like bubble and zapier
7:17 but then chatgpt itself can code for you
7:22 but you kind of have to know how to put
7:25 an app together and if you don't it can
7:27 teach you that but give me a little more
7:29 uh specifics in your question I'm
7:32 interested in using AI to help with my
7:34 Forex Trading so so yeah so what I would
7:36 say is
7:38 go to chatgpt go to go to either the top
7:41 one or Bing might be the one to go to
7:43 because Bing is gpt4 and it's connected
7:45 to the internet and
7:48 um tell it your situation so what other
7:50 ways you can use these tools is give it
7:52 a lot of context so say I'm a Forex
7:54 Trader either you you are experienced or
7:56 you're not experienced tell it what your
7:58 favorite
7:59 trading methods are if you've got any
8:01 particular Bots or any any particular
8:03 stop loss triggers things like that and
8:06 tell it all that stuff and then start an
8:09 interaction with it and have it help you
8:11 start improving how you do what you do
8:15 um
8:19 I might be able to make myself smaller
8:21 in the screen but I don't know how to
8:23 right now let's see if I go enhance
8:27 effects
8:30 right right
8:34 hang on hold
8:37 you do Saturday night that's some old
8:39 guy try to figure out Tick Tock it was
8:41 embarrassing
8:46 myself now no I can't
8:50 there's no there is no self-squishing I
8:54 I cannot resize myself I'm sure I can
8:58 there's I guarantee you there's some 18
9:00 year old on here going come on you old
9:02 man
9:04 you click on the micro spec in the
9:06 corner of the corner of the icon
9:08 and that gives you energy
9:12 little head you damn old man
9:14 wait why is an 18 year old talking like
9:16 an old man I don't know because it's
9:18 Saturday night and I haven't been
9:19 drinking
9:20 I think I need to get myself a whiskey
9:25 all right
9:27 um
9:30 could AI become self-aware
9:33 um
9:34 it potentially can I mean I think it's I
9:36 I mean certainly the current the current
9:39 um
9:41 the current AI tools are not self aware
9:44 at all they're not even aware of what
9:45 they're what they're saying
9:47 um when you kind of understand how they
9:49 work they're really just kind of word by
9:51 word prediction engines that that do it
9:53 incredibly fast and you know in an
9:55 incredibly sophisticated way so it seems
9:57 like they're self-aware
9:59 um
10:00 when you start adding things like
10:02 automation where where they can trigger
10:04 themselves where they can write their
10:06 own code where they can call services
10:09 um where they can store memory
10:12 um
10:12 and and start to analyze what they've
10:15 said and all of those things have been
10:17 released into the wild so so all the
10:18 tools are there for that I suppose I
10:21 suppose it could be you know the way I
10:24 think about this is
10:26 I don't think we've seen the last of the
10:28 different kinds of AI models that come
10:30 out for this generative stuff
10:32 right now most of most of the technology
10:36 we're using in this sort of public Arena
10:38 are based on this thing called the
10:40 Transformer model there could be other
10:42 models that come out I mean meta's been
10:44 releasing a ton of stuff
10:46 recently and they're open sourcing at
10:48 all so
10:50 um yeah it's it's very possible it could
10:52 become self-aware I think there's enough
10:54 awareness now starting to be that I
10:56 think the companies are
10:58 the companies that are putting this
10:59 stuff out and hopefully the developers
11:01 are being you know at least decently
11:03 careful about it but you have a bunch of
11:05 Open Source uh large language models so
11:08 those have no safety restrictions on
11:10 them so uh very very possible
11:14 and if your question is at that point
11:15 then what happens I don't have a [ __ ]
11:17 clue it's probably not good could AI
11:20 elect a president could probably predict
11:22 it
11:24 um in fact I mean listen here's the
11:27 thing about AI like because of chat gbt
11:29 there's a lot of people who are just
11:30 becoming aware of it um ai's been around
11:33 for decades machine learning and Ai and
11:35 things like that chat GPT is sort of an
11:38 inflection point where the rest of us
11:40 get to use these tools right
11:42 all right
11:44 um chat app coding
11:47 um you could use so zapier has a new
11:52 thing called interfaces where you can
11:54 create chat Bots here let's go do that
11:56 now this is going to take me a while
11:58 because I've got a cool groovy green
12:00 screen here oh wait it looks like some
12:02 new people came in so I'll go show
12:03 zapier in a second if you're new here my
12:05 name is Kyle Shannon this is the AI
12:07 learning lab these URLs behind me the
12:09 top one there is uh Chachi PT
12:12 and if you haven't played with chatgpt
12:15 go there right now and listen to the
12:17 doggy bark nice doggy
12:19 um that's gonna make my doggy bark so
12:21 get ready for that
12:23 it's coming no all right uh so go there
12:27 um this one is Bing bing.com is
12:29 Microsoft's uh search engine if you
12:31 click on the chat button that's chatgpt4
12:34 for free connected to the internet so
12:37 that's why that's there and then pro.com
12:39 lets you play with six different chatpt
12:41 like models from three different
12:42 companies
12:44 um and then prompts.chat will teach you
12:45 to prompt and future PD is just a giant
12:49 list of tools so that's what those are
12:50 there for uh I'm gonna some someone just
12:53 asked me about uh chat app coding and
12:55 how can you do that really easily
12:58 so I haven't done this in a while so so
13:00 sorry so I'm gonna flip my camera
13:03 turn off my green screen
13:09 bang look at my machine
13:13 all right so let's go look at zapier
13:16 all right this is this is my newsletter
13:19 everyday AI if you want to learn about
13:21 um
13:22 AI stuff feel free to subscribe to that
13:25 it's everyday AI Dot beehive.com
13:29 um
13:31 b-e-e-h-i-i-v.com not how it's actually
13:33 spelled because you know internet people
13:36 thank you
13:37 all right so let's go to zapier
13:43 so the way zapier works if you don't
13:45 know zapier
13:46 um
13:48 there's these things called zaps and a
13:50 zap is a collection of step-by-step
13:53 processes that that start with the
13:56 trigger so the the trigger here is
13:59 um I've got a Google form and then when
14:02 someone puts information in the Google
14:03 form we take that information put it in
14:05 a prompt for open AI send that off to
14:07 open AI do that twice to do some
14:09 different things to it then pop all
14:11 those results into a Gmail and send off
14:15 a Gmail so so a zapier it's called an
14:17 automation tool and and
14:20 um and and it's it's a no it's a no code
14:24 or low code tool I think it's
14:26 technically a no code tool
14:28 um they're they're quirky is [ __ ] but
14:30 whatever okay now
14:34 okay so there's this new thing on the
14:37 left hand side called interfaces
14:40 which is in beta please log in okay
14:44 okay and then you can say create a new
14:47 one in the upper right and you have a
14:49 bunch of different models here
14:51 but one of them is
14:55 a business coach bought or an explained
14:58 bot so let's do a business coach bot
15:02 and what's cool about interfaces is
15:03 you're creating a
15:08 um
15:08 you're creating a zapier you know like
15:12 this little mini application but it will
15:13 give you a web address that you can you
15:15 can just send people to this thing so
15:17 it's making an application
15:19 all right so there it is now if I say
15:22 edit
15:23 on the right hand side here this looks a
15:26 little small let's make a little bigger
15:27 a little bigger than that
15:29 on the right hand side here you have the
15:31 name so so we're going to call this
15:34 um
15:35 business Yoda
15:38 bot
15:41 um or is there a Yoda emoji that would
15:44 be nice wouldn't it
15:46 Yoda I doubt it
15:48 nope
15:50 let's say green
15:52 uh
15:55 yeah well all right a frog sort of looks
15:57 like Yoda
15:59 all right
16:01 um the greeting is static so so let's
16:03 see I'm an AI powered bot that gives you
16:05 excellent wait uh I'm gonna copy this
16:08 we'll go play here for a second so I'm
16:10 gonna hop over to chat GPT for a second
16:13 and I'm going to go new chat
16:16 and I'm going to say re-write
16:21 uh this
16:26 shift return I'm going to paste in the
16:29 little uh welcome that they had there
16:31 and then I'm going to say as Yoda
16:36 powerful business advice I offer as an
16:38 AI powered bot I am okay that's good
16:41 perfect okay so so we made a stupid
16:44 intro we're gonna go back over here
16:45 we're going to paste that in so so one
16:48 of the things that I've been you know
16:50 like one of the things you can do with
16:52 these AI tools is just Entertain
16:53 Entertain the [ __ ] out of yourself
16:56 um ask me a business question let's go
16:58 rewrite that as Yoda
17:01 um
17:02 re
17:04 write this two
17:08 uh a business question you may ask of me
17:11 perfect
17:14 so that's good now the directive you you
17:17 are a bad business coach okay so you are
17:19 a wise business coach that don't say
17:23 what I'm doing right now is I'm writing
17:24 a prompt right
17:25 that only responds with high level
17:29 advice
17:32 um
17:35 and only for
17:38 and and only
17:41 to expand the force
17:46 and I'm just gonna kill that
17:49 um respond
17:51 as
17:54 Yoda from Star Wars
18:00 um
18:00 and keep your answers
18:06 short
18:08 and sweet
18:11 all right and we're going to use
18:12 openai's GPT 3.5 that's all we're
18:15 allowed to we've got creativity if you
18:18 lower that to zero it's not very
18:20 creative if you put it higher it's more
18:22 creative so go point nine
18:24 and then you can you can do embedding
18:28 kind of stuff so that's it so basically
18:29 what a bot is is you're sort of
18:31 pre-prompting this thing so now
18:34 if I go in here I can test it
18:38 oh wait I guess what I do is I go share
18:41 and then I click on the URL
18:43 so now I'm going out to the world
18:47 and
18:49 I've got a Yoda bot and I'm going to say
18:52 how do I close more
18:56 business
18:57 my sales team
19:00 sucks
19:03 your sucks your sales team does not
19:06 train and motivate you must
19:08 [Laughter]
19:14 um
19:14 let's let's see
19:17 my clients are too demanding and don't
19:23 pay their bills
19:26 or bulls
19:27 focus on value not price price payment
19:30 terms set clearly contract terms enforce
19:32 them you will
19:34 so so there you go so go to zapier use
19:37 the interface tool and and within
19:40 within mere moments you too oh wait I
19:44 hit the wrong button here you too can
19:47 um create a bot
19:49 how do I flip this around out here oh
19:51 please old man trying to use technology
19:54 warning warning
20:00 oh good God all right it's coming it's
20:03 gonna be so good when he gets there all
20:04 right I'm back
20:06 hey this is AI learning lab welcome
20:08 everybody
20:09 um if you're new here these URLs behind
20:12 my head
20:13 are um the top ones chat GPT if you
20:16 haven't used it go use it go use it now
20:20 all right
20:23 um
20:25 uh so what's going on here I'm an
20:28 entrepreneur uh this is about generative
20:30 AI this tool this live my channel
20:33 um I'm the founder and CEO of a company
20:36 called storyvine thank you very much
20:38 let's talk
20:40 um I'm a founder and a CEO of a company
20:43 called storybine which is an automated
20:45 video storytelling platform I've been
20:47 deep diving into this generative AI
20:49 stuff for about a year and I'm using it
20:51 to kind of completely re-architect my
20:53 product and my business and my value
20:54 proposition and this channel exists to
20:58 basically just teach what I'm learning
20:59 so as I'm learning I'm sharing as much
21:01 as I can because I think it's really
21:02 important and I think that that these
21:05 generative AI tools are incredibly
21:06 profound and they're moving much faster
21:09 than I think anyone's aware of and so
21:11 that's why this exists I want to
21:13 demystify things I want to get rid of
21:15 the doomers that are just operating out
21:17 of a place of fear like the robots are
21:19 gonna kill us what are we gonna do
21:21 or like they're stupid man I heard they
21:24 lie sometimes
21:26 um they do by the way
21:28 um but it's no reason not to use them
21:29 they are still profoundly powerful
21:33 it's one of the most powerful tools ever
21:35 put in the hands of man or women or
21:39 children or even dogs
21:42 all right I also have a degree in acting
21:45 in case that was a painfully obvious
21:48 all right
21:50 wait bro what's the AI that makes apps
21:53 for you okay I'm trying to figure out
21:54 where I was in my questions because you
21:56 know old man with technology all right
21:58 I want the GPT like app that can
22:01 remember a set of rules I don't want to
22:04 reveal
22:05 um
22:06 so you can you can also do that on
22:08 po.com if you go to po.com
22:11 in the left hand side you'll see a bunch
22:13 of chat gbt things that you can play
22:15 with and then you'll see bots explore
22:18 Bots and there's create a bot and just
22:20 like I just did with zapier you can
22:23 create a bot within Poe and
22:26 you put in the instructions like I did
22:28 just did and you can turn off the
22:30 ability for someone to see your
22:32 instructions so you can do that with po
22:34 right now now you can't charge for it
22:36 you can't turn it into an app but if you
22:38 want to just experiment with getting a
22:40 bot working it's a great way to practice
22:43 whiskey is an old man drink and I'm an
22:46 ordered man get off my lawn
22:50 that just makes me want to go get a
22:52 whiskey Whiskey's an old man's drink
22:54 Whiskey's a man's drink
22:57 children all right I'm closing my window
23:01 I don't want people looking in at me
23:02 because I'm just weird outside Chicago
23:05 girl okay I figured you would have heard
23:07 of it uh what are your thoughts on
23:09 Bubble versus Anvil I don't know Anvil I
23:12 haven't used bubble I've been using
23:14 zapier just because it was the first
23:16 thing I tried and it kind of works it's
23:18 I mean it's the one that's been around
23:19 the longest but it's also the it's just
23:22 janky and old and tired
23:25 um
23:26 I don't know but they're doing cool
23:28 stuff I mean zapier now has
23:30 um large language model sitting on top
23:32 of it so you can just describe what you
23:34 want and it'll create a Zap for you
23:37 um it's also got a plug-in in chat GPT
23:40 um Bubble looks Slicker
23:43 um but I've not played with it but but
23:45 the the couple of like YouTube videos I
23:47 watched on it said it was pretty good so
23:49 I don't know
23:51 knows well so so here's what I would
23:54 encourage you to do go play with both of
23:56 them go make some apps I mean the thing
23:57 about these large language models like
23:59 when you when you take something like
24:01 zapier
24:02 and like like so the the things that
24:05 I've been doing are making content
24:07 generators so I'll create a Google form
24:09 that inputs asks people for some input
24:12 like give me a topic and then answer a
24:14 few questions about that topic I I then
24:17 take the inputs from that form I
24:19 incorporate them into a a chat GPT post
24:22 based their prompt basically send that
24:24 off to open Ai and then get back
24:26 something really cool so you can make
24:28 really sophisticated apps with like
24:30 three steps you know take an input from
24:32 a form or spreadsheet or whatever Google
24:34 doc
24:36 send it to open Ai and then like pop it
24:39 in an email and email it to someone or
24:41 post it to a slack Channel or post it to
24:43 a Discord Channel
24:44 um so I'd say just go start playing with
24:45 them
24:46 why is a AI exploding all over the place
24:49 right now what happened what happened to
24:51 make
24:52 more companies be able to release wait
24:54 what happened to make so so here's
24:56 here's what's going on here's why ai's
24:58 exploding right now
25:00 okay so I'm gonna I'm gonna go back in
25:03 history in my day
25:05 let's go back to 1994. I was just a
25:08 young week young we whip a snapper so
25:12 um
25:14 in the mid 90s the internet had been
25:17 around for decades and and I was living
25:20 in New York I was doing desktop
25:21 publishing
25:23 and I started messing around on the
25:24 world wide web but you had to a you had
25:26 to dial up and then B it was all command
25:29 line you had to you had like no command
25:30 line things and there were different
25:32 kind of services you could dial into but
25:34 it was it was very janky and it was all
25:35 kind of scientists and researchers and
25:37 University folks
25:39 um
25:41 and then the World Wide Web comes along
25:43 and the World Wide Web all of a sudden
25:45 makes it as easy as if you write this
25:48 HTML page you can like click on these
25:50 blue underlined words and it'll jump you
25:52 to another page right so connected all
25:54 these Pages without you having to learn
25:56 all these commands and that changed
25:59 everything right
26:01 the it took six years for the World Wide
26:05 Web to get from zero users to 100
26:07 million users
26:09 so and and you know how transformative
26:11 that was right like almost everything we
26:13 do today from a Communications
26:15 standpoint is is you know can be traced
26:17 back to the ability to like point and
26:19 click on things like our phones are now
26:21 just you know point-and-click devices
26:22 right
26:24 um so so had this incredibly profound
26:27 impact on how we do business how we
26:29 communicate how we interact how we you
26:32 know do social stuff
26:34 so similarly with with
26:37 um machine learning and Artificial
26:39 Intelligence been around for decades the
26:41 the origins of it are probably 40 50
26:43 years old and then kind of modern
26:44 machine learning in the neighborhood of
26:47 probably 20 years right and then in uh
26:51 2017 Google invents this thing called
26:53 the Transformer that allows these tools
26:55 to get really really powerful
26:57 and
26:58 um that's the the Transformer is the
27:00 thing behind like mid Journey the image
27:02 generation tools where you see people
27:04 posting all these images that look
27:06 really sexy the image behind here is
27:08 from stable diffusion that was something
27:10 I did for an art project
27:12 um just a just a cool you know thing
27:15 um and then it's it's the it's the
27:18 Transformers are the technology behind
27:20 the GPT tools which stands for
27:22 generative pre-trained Transformer
27:24 Transformer that's the GPT now why did
27:28 it take off so quickly
27:30 so open AI is the company behind chatgpt
27:34 they they first did GPT one then they
27:36 did gpt2 they did gpt3 and they launched
27:40 it in 2020
27:42 um I think mid to the end of the year
27:43 something like that
27:45 so G and gpt3 got a little bit of Buzz
27:47 right it was out there but it was mainly
27:49 for developers you went to this sort of
27:50 geeky back end where you could change
27:52 settings and choose different models and
27:54 it it required you to think right and it
27:57 you had to be kind of a developer
27:59 so then Microsoft is about to announce
28:03 that they're investing a billion dollars
28:05 into open AI this is probably around
28:07 October and Sam Altman the CEO of openai
28:10 said hey Tech Team at open ai go make
28:14 chat GPT
28:16 um you have two weeks to do it so they
28:18 wrote chatgpt in two weeks launched it
28:20 Microsoft made their announcement that
28:22 they invested a billion dollars and chat
28:25 GPT went from zero to 100 million users
28:28 in six weeks
28:30 the World Wide Web took six years
28:33 chat GPT took six weeks to 100 million
28:36 users fastest adoption of tech of
28:39 technology and history
28:41 um like like just boom what else about
28:45 it well it is it looks a lot if you
28:49 haven't played with it looks a lot like
28:50 a Google search box it's just a single
28:52 text entry box kind of like Google but
28:55 you put in something really simple and
28:56 you get back remarkable stuff and so it
29:00 caught on and people realized oh my God
29:02 this is going to change everything it's
29:03 going to change writers and how I write
29:05 and and it can code and it can solve
29:09 problems and it can it can just do all
29:11 this stuff
29:12 um so because of that fast adoption a
29:15 couple of things happened Google freaked
29:18 out because
29:19 their business model is not based on
29:22 asking a question and giving you an
29:24 answer their business model is based on
29:27 you ask a question and they'll give you
29:29 40 options and the more times you click
29:32 the more money they make
29:34 they don't want just an answer to be
29:37 given so so when Chachi PT went to 100
29:41 million users in six weeks they were
29:43 like oh [ __ ] what do we do right and
29:45 they invented this technology so
29:47 Google's trying to play catch-up then
29:49 meta you know the metaverse sort of you
29:51 know fell off a cliff and Zuckerberg is
29:53 not talking about that anymore but
29:55 they're like we do AI now so they're
29:57 jumping in
29:58 AWS Amazon has jumped in and said we're
30:01 jumping in so so lots and lots of
30:03 companies are now kind of panicking try
30:05 trying to get in the game because open
30:07 Ai and Microsoft are like this
30:09 juggernaut
30:10 that even though they've been developing
30:12 this stuff for five or six years they
30:14 just kind of blew onto the scene in six
30:16 weeks with chat GPT and it's changed
30:18 everything so that's why that's a great
30:19 question thank you
30:22 um I'm a Transformer bro I'm gpt5 party
30:27 uh
30:30 wrote half my peer reviews with some
30:32 semblance of chat GPT just here to gloat
30:34 awesome
30:35 awesome Ms gen rad no listen that's
30:39 great here's the deal like like there's
30:41 no shame in it I don't think if you've
30:44 got a tool that can can help you block
30:46 and tackle the the the the chunks of
30:49 your writing more quickly like what what
30:51 I've experienced with Chachi BT is it it
30:53 will get me to 80 percent
30:55 much more quickly like instantly and
30:58 then it's still on me to do the last 20
31:00 it's very much a collaborative
31:01 interactive process it's not like the
31:03 robot is writing all my [ __ ] it's like
31:06 I've got a tool that can get me there a
31:09 hell of a lot faster and I can spend
31:10 more time kind of at a higher creative
31:12 level
31:13 um hey everybody it looks like some new
31:14 folks are here if you're if you're new
31:17 here my name is Kyle Shannon this is the
31:19 AI learning lab that's what those
31:20 letters are behind all the interface
31:22 crap
31:23 um these three URLs the top one is chat
31:27 GPT if you have not used it go use it go
31:30 play with it just type [ __ ] into it say
31:33 um write me a song about and describe
31:36 your dog or your significant other and
31:38 just have it write you a song If you
31:40 haven't done it just like go do a parlor
31:42 trick like that and then when you pick a
31:44 job off the floor start having it help
31:46 you solve a problem Bing is a similar
31:49 sort of thing if you go to bing.com and
31:51 click on the chat button that's also
31:53 chat GPT but it's it's gpt4 it's free
31:56 you have to pay for GPT for it open AI
31:59 which which I do because I think it's
32:00 better but you don't have to pay for it
32:03 there and then Poe it's got a bunch of
32:05 things you can play with including a bot
32:07 maker where you can make your own Bots
32:08 and then prompt chat will teach you to
32:11 prompt and futurepedia.io will give you
32:13 a list of a directory of tools so go
32:15 play with all that stuff and then if
32:17 you've got questions about AI pop them
32:19 down below and if you're uh
32:22 if you lean more toward the troll side
32:24 of the house and you're like what are
32:26 your qualifications man
32:29 um I ain't really got none I have a
32:31 degree in acting although
32:33 I've started probably 15 companies in my
32:36 day I was there for the early days of
32:38 the World Wide Web I started one of the
32:40 first digital agencies in the mid 90s
32:42 called agency.com
32:44 was the chief creative officer there so
32:46 I guess I have a history of identifying
32:49 technologies that are going to be
32:50 interesting and seeing what I can do
32:52 with them and I think AI is bigger than
32:54 the web was by a lot and I think it's
32:57 coming a lot faster how do I use AI to
32:59 my advantage and make some money on the
33:01 side you can make money
33:05 um first thing I would suggest is anyone
33:07 who says that ignore them
33:10 um
33:13 here's my philosophy on business and on
33:15 a Ai and making money with AI you there
33:18 there are going to be lots and lots of
33:20 opportunities to make money with chat
33:22 gbt
33:24 um but it's got nothing to do with chat
33:25 GPT chat gbt is just a tool right so
33:28 that you could you could absolutely
33:30 rephrase your question as how can I use
33:33 Photoshop to make money how could I use
33:35 a shovel to make money how could I use
33:37 you know this cake batter to make money
33:40 how you make money is you solve a
33:43 problem that that a group of people have
33:46 and you solve it well enough that
33:47 they're willing to pay you money for the
33:49 solution right and it doesn't have to be
33:51 like a dramatic problem like it could be
33:53 I want to be entertained and I'll buy a
33:54 silly t-shirt or I want to be inspired
33:56 and I buy a piece of art or I want to be
33:58 I want to
34:00 um I don't know
34:01 um come up with a better marketing plan
34:03 and you create an agency and help people
34:05 come up with better marketing plans so
34:07 so I would say start with the problem
34:09 and then figure out
34:12 based on what you know and what you're
34:14 good at how might you solve that problem
34:16 and have chat GPT be like your the best
34:19 assistant you've ever worked with to
34:21 help you solve that problem and maybe to
34:24 solve it faster so what what generative
34:26 AI can do what Chad gbt can do and tools
34:29 like it is amplify
34:32 what you're already doing or what you
34:35 what you want to do so I know that was
34:36 kind of generic but that's that's my
34:39 thoughts on that
34:40 um it's really good at writing it can
34:42 code
34:43 um
34:45 we do the other night on one of these
34:47 lives we did a bunch of t-shirts
34:50 um for moms that were exhausted someone
34:52 said can you make can I make t-shirts
34:53 for moms that are exhausted and I just
34:56 said you know hey Chachi PT you know
34:58 write me a bunch of t-shirts and just
34:59 out it came with different Target
35:01 audiences it was brilliant
35:03 um so you can do all sorts of stuff
35:05 start with the problem and then have it
35:06 help you
35:08 um do that the professor has Tech skills
35:11 oh yeah reminds me of uh IFTTT but with
35:15 chat gbt I'm sure IFTTT IFTTT stands for
35:19 if this then that which is another
35:21 automation tool like zapier I would be
35:24 very very surprised if there wasn't an
35:26 open AI tool within IFTTT
35:30 so this makes an app uh yeah so so what
35:33 I was showing before sorry about that uh
35:36 tar fari
35:38 um
35:40 what I was showing was zapier's
35:42 interfaces beta and the interfaces beta
35:45 lets you make it make lets you make an
35:47 app I think you have to pay whatever
35:49 like oh here's the problem with all of
35:52 these tools are like 20 bucks a month
35:54 like you can 20 bucks a month yourself
35:55 you know into a mortgage payment
35:58 um so but so so I don't make any money
36:01 off zapier so
36:03 um in case that's a concern and I'm just
36:05 trying to Hawk a piece of [ __ ] software
36:06 it's actually pretty good software it's
36:08 just a little janky
36:09 um but the interface is beta when you
36:13 create a bot it will give you a URL and
36:16 then if you pay the 20 bucks you can
36:18 change the url into something custom
36:20 like you know something bought or you
36:22 could you know Point your own domain at
36:24 it
36:26 um so yes yes yes there's the frogs a
36:29 cardi B bot needs to happen I did a JZ
36:31 bot here let me let me show you okay so
36:34 so uh it doesn't matter I I it's I I'm
36:38 gonna flip my camera and I gotta turn
36:39 off green screen and it's going to be a
36:41 bit of a [ __ ] show but I want to show
36:43 you the uh I'll show you the Jay-Z bot I
36:46 made
36:47 all right so flip camera and then go
36:50 here then I go here then I go here all
36:52 right there's my thing so let's go to PO
36:56 okay so where I'm at right now is I'm at
36:57 Poe and Poe is just like chat GPT
37:03 um and down the left hand side here are
37:07 um you know gpt4 Claude Claude instant
37:10 100 000 that's pretty that's amazing
37:12 actually and then there's all these Bots
37:14 down here and I can scroll down and I
37:18 can go to all your butts so I'll go to
37:21 my bots
37:22 so I've got a Ricky Bobby bot
37:25 I've got a Shakespeare bot Yoda Emily
37:29 Dickinson Steve Jobs Jay-Z
37:31 so here's a Jay-Z bot and he says
37:33 welcome my friend to the [ __ ] realm
37:36 let's create fire with you at the helm
37:39 um
37:40 hey man I want help
37:44 writing
37:47 um
37:48 uh let's see oh I want help
37:50 understanding
37:53 why
37:55 fast cars are so cool
38:03 sending still sending
38:06 um
38:07 looks like Poe's not working
38:09 come on PO
38:11 or oh here we go fast cars Neri dreams
38:14 fueled on gas doing donuts and massive
38:17 Lots made of grass speed defense and
38:19 hearing sharpens your vision giving life
38:21 an illusion uh fast forward in Precision
38:24 so
38:25 um the the prompt sitting behind this
38:27 basically says you're Jay-Z and and I
38:30 you know everything about Jay-Z and how
38:32 he Rhymes and
38:33 when someone asks you a question you
38:36 only answer you know with your with your
38:39 sick words
38:41 um yeah so that's pretty cool so that's
38:43 that's Poe and you can do that there so
38:46 you can go you could go do exactly that
38:48 with cardi B
38:49 is there any way to speak to Ai and have
38:52 it reply in a voice yeah there is a
38:56 what was that app called ask someone hmm
39:02 I don't know let's see if I can find it
39:05 um
39:05 ask iOS
39:08 AI app
39:12 oh dang it ask
39:17 and I IOS app
39:20 it's like ask Emily or something like
39:22 that
39:23 um
39:24 there's a couple of them out there it's
39:26 not it's not super easy to do right now
39:29 but you can do it so let me flip my
39:33 camera back around
39:35 and hold please let me turn on the green
39:37 screen
39:39 it's gonna be so good all right there
39:41 you go looks like we have a bunch of new
39:42 people there hello my name is Kyle
39:43 Shannon this is the AI learning lab
39:45 happy Saturday night
39:47 um
39:48 the uh the uh this channel is about
39:52 learning AI
39:54 um if you have questions about it pop
39:56 them in the questions below I'll do my
39:58 best to answer them
40:00 um these URLs right here are places for
40:02 you to go play chat.openai.com that's
40:05 the official chatgpt website they're the
40:08 ones that started it all well the kind
40:10 of new started it all the AI for the
40:13 rest of us inflection point that started
40:15 back in November
40:18 um bro use voice chat extension in
40:21 Chrome it's easier yeah you can you can
40:24 um
40:25 if you use like the
40:27 um there's an official open AI app
40:30 but it's really hard to find you can't
40:32 find it if you go to the IOS app store
40:34 because there's like 8 000 scammer apps
40:38 so go to openai.com and go to products
40:41 chatgpt and then and then they'll have a
40:44 link to their official app if you go to
40:45 their official app they've got whisper
40:47 Transit trans
40:49 um
40:50 translation or sorry transcription built
40:54 into it so you can talk to it and it
40:56 will respond I don't think it responds
40:57 in voice though there are some tools out
41:00 there that do it um but they're but
41:01 they're just coming onto the scene they
41:03 all have a lot of latency so they're
41:04 they're a little awkward and weird
41:06 basically meaning when you ask it a
41:08 question that like there's this long
41:09 awkward pause where you're like are you
41:12 ever gonna answer and then it's like
41:16 and it talks really weird and stilted
41:18 but give it a year and it'll be instant
41:20 and sound human
41:22 and then everything changes Mercedes
41:25 just announced that that they're putting
41:27 chat GPT into their Entertainment
41:29 Systems in their cars
41:32 um by the way if you're here if you
41:34 would do me a favor just tap the screen
41:36 um make the little hearts go that'll let
41:38 people know we're here and pretend that
41:40 you like this we'll get a bunch about a
41:43 bunch of other people here uh if you
41:45 have questions about uh uh AI pop them
41:48 in the comments and I'll do my best to
41:50 answer them I'm going back and trying to
41:51 find out where I was in the questions
41:52 and I will answer them now
41:54 thank you and you can also
41:57 share this live I think that helps as
41:58 well so if you haven't shared the live
42:00 and want to do that that would be cool
42:03 um Poe is an AI tool Google it yeah
42:06 poe.com is one of the things I just
42:08 showed
42:10 um I'm trying to go back here and figure
42:11 out where I was I'll just I'll just sort
42:13 of jump in
42:15 um but what you want is probably a GPT
42:18 extension okay you're helping each other
42:21 I like that what's popo.com that
42:26 is
42:27 um just like chat GPT and you can
42:30 actually use chatbt there but it's
42:31 basically a single interface for
42:34 multiple
42:36 um large language models that you can
42:38 play with so there's Sage which is pose
42:40 there's GPT 3.5 gpt4 there's Claude
42:45 instant Claude instant 100K which you
42:48 can put 75 000 words in and then there's
42:50 quad plus which is like gpt4 some of
42:53 those require subscriptions and then
42:56 there's a whole bot Library like you can
42:59 explore thousands of bots that people
43:01 made and you can make your own Bots it's
43:03 pretty cool
43:05 um someone described it the other day as
43:06 like the play school like poe.com is
43:09 like the play school of of these large
43:11 language models it is a little simple
43:14 looking
43:16 um you know the the the original is that
43:19 top URL there but that's what Poe is
43:22 uh hello hello Margaritaville 68 was
43:26 shaking Gen X baby
43:29 let's let's keep it real
43:33 um
43:35 all right I like your channel keep doing
43:38 it I will thank you I really do
43:40 appreciate the kind words it's very nice
43:43 this channel exacerbates
43:46 something baits laughs
43:50 he's so naughty he's just naughty
43:54 you know you know what I'm saying this
43:56 is all naughty what's he talking about I
43:58 don't think he he tries to figure out
44:00 Tick Tock it takes him like three hours
44:02 every night he hasn't gotten it yet
44:05 he might be a little touched all right
44:07 any tips about image
44:09 generator currently leaning stable
44:11 diffusion yes lots of tips I don't know
44:14 if they're tips I have lots of opinions
44:16 on on image generation okay
44:19 I think the best one by far from a
44:22 quality standpoint is mid-journey but
44:25 here's the downsides to Mid Journey you
44:28 you can only create images right now in
44:30 Discord
44:31 you have to do like command line
44:33 interfaces like you type slash imagine
44:36 and then you type in a prompt and then
44:38 you have to put in other
44:39 janky dash dash AR space 16 colon 9
44:43 which stands for aspect ratio 16 colon
44:46 9.
44:47 rather than just clicking an icon
44:50 um so that's the downside but it's
44:52 really better than anything else out
44:53 there
44:54 stable diffusion because it's built on
44:57 because stability AI open source the
45:00 model
45:01 the developer Community has gone nuts on
45:03 it so stable diffusion is by far the
45:05 most flexible and there's the most
45:07 development going on for it and there's
45:09 all sorts of tools for it
45:12 you don't have to quite be a rocket
45:14 scientist but you have to be pretty
45:15 [ __ ] geeky to use stable diffusion
45:17 and just keep up with how quickly it's
45:18 moving so it's really good like that
45:20 image this image was created with stable
45:23 diffusion like down here's my my sort of
45:26 chin and mouth and nose this was part of
45:28 a project I did called Kyle Shannon
45:30 dreams using a tool called dream Booth
45:32 where I could train it up on photographs
45:34 of me and generate all sorts of cool
45:35 images that was all done in stable
45:37 diffusion
45:39 I failed setting that up nine times in a
45:42 row before I actually got it to work
45:45 um and I'm decently geeky right I'm not
45:48 technical but I'm decently geeky I can
45:50 figure [ __ ] out
45:51 um so that's the downside about stable
45:53 diffusion you can get the quality with
45:57 checkpoints and weights and and some
46:00 things like that you can get the the
46:01 quality of stable diffusion can get
46:04 close to Mid Journey but it's just not
46:06 consistently there out of the box
46:09 um
46:10 dolly d-a-l-l-e is open ai's image
46:14 generation thing if you go to bing.com
46:16 create that's Dolly and it's free right
46:20 and it there's no setup you just go
46:22 bing.com create and you can go make
46:24 images right now I'd go play right now
46:26 in my opinion by far the worst of them
46:29 from a quality standpoint it just some
46:31 things you give it it does good other
46:33 things it does bad it's it only gives
46:36 you square images it doesn't do
46:38 horizontal or vertical images which is
46:39 just like why
46:41 like you can figure out how to
46:43 make a different aspect ratio everyone
46:46 else has
46:47 um
46:48 but it's free
46:50 and then the final one that that I think
46:51 is worth noting right now is Adobe
46:54 Firefly so Adobe Firefly is good
46:59 um some of their interface elements are
47:01 really good like you can just click on
47:02 buttons and change Styles and things
47:04 like that the most important thing about
47:06 Adobe Firefly and also by extension
47:09 generative fill within Photoshop and now
47:11 they've got recolor within illustrator
47:13 and they've got a type tool
47:16 all of their generative AI stuff right
47:18 now is built on models that are that are
47:20 based on
47:21 um that are trained on public domain
47:23 images and Adobe stock images that they
47:25 own all the rights to so anything that
47:27 you create with Adobe
47:29 Firefly is is essentially commercial
47:32 safe right you're not you know there's
47:36 kind of no chance that some artist is
47:38 going to go hey you stole my image and
47:40 you did Ai and my image was in that
47:42 training set
47:43 um so so that's that and they're also
47:46 Adobe and and they're they're moving
47:48 quite quickly
47:50 um with experimenting with generative AI
47:52 they're thinking incredibly
47:54 intelligently about attribution and
47:56 compensation for artists they're doing
47:58 it in an artist friendly way in a
48:00 commercial friendly way so those are the
48:02 four mid-journey the best stable
48:04 diffusion the most flexible dolly is
48:07 free and you can call it from apis but
48:08 it's the crappiest and then um Adobe
48:11 flyer Firefly is good commercial safe
48:14 and probably going to be really good
48:16 and if you haven't used
48:18 um generative fill in Photoshop it's
48:21 just like [ __ ] magic
48:23 um so yeah
48:25 all right
48:26 thanks for showing real screen you're
48:28 welcome I'll do that again if it makes
48:30 sense is snoozy still the best open
48:32 source model I'll be honest with you the
48:35 the
48:36 um
48:37 um
48:38 I haven't the open source models are are
48:42 literally
48:43 they're kind of uh one-upping each other
48:46 like every three days like honestly
48:48 every three days there's a whole new
48:50 open source model that like claims to be
48:52 better than chatpt
48:55 and now you're gonna do it for free
48:58 and and so I I'm just exhausted I'm just
49:02 using chat GPT but if you need to run
49:04 something locally if you need to to
49:08 um experiment with this stuff and you
49:10 want to push the limits and you don't
49:12 want the safety guard rails that when
49:13 you know when chat GPT says I'm not
49:15 going to do that for you because I have
49:17 safety guard rails all that open source
49:19 [ __ ] doesn't have that so what I would
49:21 why would it encourage you to do is just
49:23 start hanging out on Reddit in the in
49:25 the um in the subreddits for large
49:27 language models open source large
49:29 language models
49:31 um there's a there's a project called
49:32 GPT for all which you can install on
49:35 your local machine and then you can put
49:36 in different models and some of those
49:38 models are really small and will run on
49:40 a like on a laptop
49:43 um I know there's some installers for M1
49:45 Max
49:46 um and things like that so but but I'm
49:49 just I'm I'm too exhausted
49:52 to to follow that right now I'm gonna
49:54 I'm gonna let the open source Community
49:56 get their their [ __ ] together and let's
49:58 let's let a leader emerge out of it
50:00 because right now it's just it's crazy
50:03 um it's good but it's crazy
50:06 and quite honestly if if you've got time
50:09 and if you're super geeky I that's where
50:11 I'd be spending my time I'd be spending
50:12 my time learning Lang chain learning
50:16 Auto GPT
50:17 um there's a new one out called GPT
50:19 engineer is that what it's called GPT
50:21 engineer hang on
50:26 foreign
50:31 yeah it's called GPT engineer that's
50:34 like something that's trending on GitHub
50:36 right now it's it's like a programming
50:37 specific GPT where you can just describe
50:40 apps and it'll go make them
50:43 I want to make a SAS cool I again I
50:46 would start for you uh sorry I can't
50:50 quite read your name Tar
50:53 tar Farley
50:55 your name is gonna make me say something
50:57 really bad it just feels like it's it's
50:59 hard to read your name without thinking
51:01 of bad words so I don't know if you do
51:03 that on purpose but parley okay
51:07 um
51:09 I would just go to chatgpt I'd go to
51:12 gpt4 so if you don't want to pay for GPT
51:14 for it open AI just go to Bing go to
51:17 their chat and and start having it walk
51:20 you through no code tools if you're not
51:22 a coder
51:24 um how you how you would create a SAS
51:26 application with no code tools and just
51:29 have it have it guide you through that I
51:30 have it walk you through that
51:34 I got I got stuff going on in the throat
51:38 there must be a lot of pollen going on
51:39 some of those Chrome extensions are
51:41 sketchy yeah a lot of the Chrome
51:42 extensions are sketched the problem the
51:44 problem with Chrome extensions right now
51:47 is that they're really easy to make it's
51:50 really easy to call openai so every one
51:53 of them can make these remarkable claims
51:55 like our extension will summarize all
51:59 your meeting notes and they're all just
52:02 basically a single prompt that says take
52:04 this text and summarize it right and
52:06 then they turn that into extension and
52:08 if if they if they're good at like mid
52:10 Journey or stable diffusion they'll make
52:12 cool little icons for it and they can
52:14 make those things like those things seem
52:16 like real apps even though they're Fly
52:18 by Night you know just you know tossed
52:21 together in a couple hours so
52:24 um My Philosophy right now I like I'm
52:27 sticking very basic I'm sticking to like
52:29 chat GPT
52:31 um I'll occasionally go to Bing I go to
52:34 Poe if I wanna if I wanna dick around
52:36 with Bots or if I if I want to use the
52:38 100K Claude model where I can do 75 000
52:41 word prompts
52:43 um is this Boomer talk uh [ __ ] you I'm
52:46 Gen X
52:48 [ __ ] you Boomer tuck it's Elder talk
52:53 Boomer talk can't figure this [ __ ] out
52:55 Gen X can all right and we paid act wait
52:58 and we paid to access the internet by
53:00 the minute remember yeah you had to have
53:02 your computer call another computer
53:05 um
53:06 with a BS in CS
53:10 took intro ml course you're all acronyms
53:13 dude schlepins
53:15 should I should I learn more statistics
53:18 or take AI specific classes
53:20 um
53:22 I would say anything that you do in AI
53:24 right now is probably a good investment
53:27 um if you haven't seen when it comes to
53:30 statistics
53:31 go watch YouTube videos on something
53:34 that's coming out from from open AI
53:36 called
53:37 um code interpreter it's a new mode of
53:41 chat GPT that's coming it's not I like I
53:43 don't have access to it yet some people
53:44 do it is a full-on data analyst like so
53:48 what happens with code interpreter is
53:50 there's an upload button in your little
53:51 search box and you can upload like a
53:54 full spreadsheet of data
53:56 and just start interacting with it and
53:58 like it was a full data analyst and say
54:01 um
54:02 tell me what's interesting about this
54:04 data and it will give you a whole Report
54:05 with graphs and then say compare this
54:09 trend to that Trend and turn it into a
54:11 3D scatter chart and it will write the
54:13 python code and generate the 3D scatter
54:15 chart so there are tools coming that are
54:18 just going to be native within uh you
54:21 know Excel and and just code interpreter
54:24 that are going to do that so what I
54:26 would say if you're interested in
54:28 statistics and you want to learn this
54:29 [ __ ]
54:31 um I would learn how to do all that
54:33 stuff with the AI tools right so so
54:39 um I say this a lot on the live there's
54:41 a guy named Paul rateser from the
54:43 marketing AI Institute and he says AI is
54:46 not going to replace you but someone
54:47 using AI will and so if you have the
54:51 ability to use machine learning tools to
54:53 do data analysis or or whatever it is
54:56 the thing that you're interested in
54:58 um sort of as in conjunction with what
55:01 you already know with the AI tools then
55:03 I think you'll be in a much better place
55:05 than if you just say hey I'm going to go
55:07 learn statistics so I can do data
55:09 analysis data analysis is going to be
55:11 the the like the bulk of the work and
55:14 certainly entry level kind of work is
55:16 all going to be automated
55:19 um so
55:21 um that's that and then the other thing
55:23 is so so there's a we're at an
55:25 inflection point right now so so Chachi
55:27 PT to me represents
55:32 a a a a diverging path right and and one
55:35 of them is if you want to be in machine
55:37 learning and computer science and
55:39 working on large language models and
55:42 working on things like Lang chain and
55:43 kind of being on The Cutting Edge of
55:45 what's happening with machine learning
55:48 and AI That's going you know that's
55:51 that's sort of entering an industry that
55:55 you know is 20 to 30 years old fairly
55:59 mature and probably has decently High
56:02 requirements in terms of math and
56:03 programming and all that sort of stuff
56:06 the other road is post GPT post chat GPT
56:10 is the rest of us get to use these tools
56:13 right Chachi BT is available for the
56:15 masses so it's everyone from someone
56:17 working on their dinner plan for the
56:19 week to small businesses figuring out
56:23 faster ways to do what they do to
56:25 massive corporations completely changing
56:28 their work processes you know based on
56:30 these new tools and when Microsoft roles
56:33 co-pilot into Office 365 which is
56:36 basically chat GPT sitting inside word
56:39 excel PowerPoint Google's doing the same
56:43 thing with Bard once these things roll
56:45 into all of the daily work tools we do
56:47 then it's just going to be everywhere
56:50 it's going to be pervasive so so those
56:52 are the two kinds of Roads right if you
56:54 want to do machine learning go down that
56:56 more specific road but
57:00 almost everyone is going to be
57:02 struggling with
57:05 how the [ __ ] do I do my business now
57:08 like these tools change almost
57:11 everything in how we do what we do and
57:14 so if you've got any sort of skills to
57:16 help people navigate through that you'll
57:18 be in demand
57:20 all right
57:22 first into HTML HTML was super early in
57:26 the early days I'm a professional
57:27 software engineer brother cool
57:30 all right
57:31 can replace you too it can
57:34 ah
57:36 and then I might go back to acting
57:39 or music I'll go back to songwriting
57:43 if AI can do all the shitty salt
57:46 crushing work like like you know when
57:48 you spend three hours on the phone with
57:49 your stupid [ __ ] insurance company
57:51 trying to get paid for the
57:53 stupid doctor visit you went to
57:56 if AI can be on the phone for me and
57:58 deal with all that complexity and I just
58:00 can go do my other [ __ ]
58:03 I am there for it
58:05 any prompts to get wait any prompts to
58:08 get rid of the silly extra sentences in
58:12 GPT oh the story continues yeah so
58:15 so GPT
58:17 um uh chat gbt has been written to be a
58:20 little bit flowery right so it
58:22 definitely does that there's a specific
58:23 phrase that someone said once
58:27 limit limit Pros okay let's go let's go
58:30 play all right I'm gonna flip my camera
58:32 it's going to take me a minute to turn
58:33 all this [ __ ] off but hold on
58:39 all right so we're gonna go here we're
58:42 gonna go
58:44 new chat so so I'm at chat gbt
58:47 uh we're just gonna use 3.5 because it's
58:50 fast
58:51 but I'm gonna do
58:53 um let's see please
58:58 um tell me
59:01 what's the best way to start a SAS
59:08 business
59:09 with AI
59:14 um
59:17 if I am not technical
59:21 all right
59:24 so let's see what it says
59:26 starting assess software is a service
59:29 business with AI can be an exciting
59:31 Venture even if you don't possess the
59:33 technical skills well not being
59:34 technically proficient may present some
59:36 challenges it doesn't have to write so
59:38 it's this really flowery kind of
59:40 [ __ ] language so I'm gonna I'm gonna
59:42 I'm gonna copy my my exact prompt here
59:46 and I'm going to start a new chat so
59:47 we're starting completely from scratch
59:49 and then I'm just going to go I'm going
59:50 to put that in there I'm going to hit
59:52 um
59:53 shift returns for a new line and I'm
59:55 just going to say the the phrase limit
59:57 Pros
1:00:00 and let's see if it's any better yeah
1:00:03 if if you're not technical but want to
1:00:05 start SAS software as a company here's a
1:00:08 concise plan to get you started right so
1:00:10 it cuts out all the [ __ ] language
1:00:12 so so you can say limit Pros you can
1:00:15 also the other thing that you can do
1:00:17 with these tools is hang on let me flip
1:00:19 my camera back
1:00:21 the other thing that you can do with
1:00:23 these tools is you can give them
1:00:25 examples so
1:00:28 so if you want
1:00:30 um if you've got a particular way you
1:00:33 write or if you wanted to just give you
1:00:35 bulleted short bulleted lists you can
1:00:37 say you know
1:00:40 you know give me you know give me this
1:00:42 sort of answer
1:00:44 but then you can say but follow this
1:00:46 example and paste in an example but
1:00:48 follow this example example for
1:00:51 formatting and tone
1:00:52 and it's really good at that so so you
1:00:55 can also give it examples that'll help
1:00:56 that'll help with that but limit Pros is
1:00:59 the little the little hack I can't
1:01:00 believe I remembered that I saw it once
1:01:03 replacement is just fear-mongering
1:01:05 totally agree with that been saying Tech
1:01:07 will replace us for ages and we're still
1:01:09 working how's Bard Bard sucks
1:01:11 [Laughter]
1:01:14 um I I know Google invented this [ __ ]
1:01:16 but like like how you can tell Bard
1:01:19 sucks is like just what it sucks on the
1:01:21 writing front I from what I've heard
1:01:23 it's good on the programming front and
1:01:26 there there are many components to Bard
1:01:28 like they're doing music generating
1:01:30 stuff they just announced a thing called
1:01:33 style drop which is remarkable like
1:01:36 image copying technology
1:01:39 um so they're doing some stuff that's
1:01:40 really cool but Bard itself compared to
1:01:42 gpt4 is like laughably bad it's
1:01:45 laughably bad
1:01:48 um
1:01:49 they just they they they dropped the
1:01:52 ball they invented the Transformer and
1:01:54 then I think I think they threw it I
1:01:56 just I think they threw it on the shelf
1:01:57 and said this is going to threaten our
1:01:59 core business so don't pursue that and
1:02:01 then open AI went and took it really
1:02:03 [ __ ] seriously
1:02:04 and then oops chechi PT you know got to
1:02:07 100 million users in six weeks and
1:02:09 changed everything [ __ ] everything up
1:02:11 for everyone except Microsoft and opener
1:02:15 AI are like cool
1:02:19 kicking ass
1:02:20 we're gonna be the new we're gonna we're
1:02:22 the new cool kids in town
1:02:26 can chat GPT
1:02:28 kenshatch EBT for court cases so chat
1:02:32 apt is really good at reading legal
1:02:33 briefs
1:02:35 um I was just part of uh I had some
1:02:38 crypto on Voyager and they [ __ ] the bed
1:02:40 and now they're in some sort of
1:02:42 settlement and every now and then
1:02:43 they'll send out all these big long
1:02:44 legal documents so I literally I take
1:02:46 those paste them into chat EBT and I'm
1:02:49 like summarize this for me and it does
1:02:51 and then I can ask you questions about
1:02:53 it so it does that really well you have
1:02:55 to be very very careful with legal stuff
1:02:57 because chat gbt is it does what's
1:03:00 called hallucinating it will just lie to
1:03:02 you it's it's a mansplainer it will
1:03:05 confidently answer you it even if it's
1:03:07 wrong there was a lawyer two weeks ago
1:03:11 um had chat gbt write a legal brief for
1:03:14 a court case
1:03:16 and turned it into the judge and the
1:03:19 judge checked the cases that the lawyer
1:03:22 cited in the brief and they were all
1:03:24 fake
1:03:26 the judge said to the lawyer
1:03:29 um excuse me uh Mr dum-dum
1:03:33 what the [ __ ] is this
1:03:35 and and the lawyer didn't check his work
1:03:37 so you know check your work if you're
1:03:39 gonna if you're gonna use it for for
1:03:41 court cases but yeah it's really good
1:03:43 it's really good at analyzing documents
1:03:47 so
1:03:48 um yeah
1:03:50 that's the other thing so one of the
1:03:52 other things
1:03:53 the reason I have poe.com here
1:03:57 in the paid subscription you get access
1:03:59 to a thing called Claude instant 100k
1:04:04 a lot of times legal briefs are going to
1:04:06 be a lot longer than what you can paste
1:04:08 into chat gbt it'll break it's got just
1:04:10 a 3000 word limit basically
1:04:13 the the hundred thousand word or token
1:04:17 model lets you put in a document up to
1:04:20 about 75 000 words so you can put in
1:04:22 really long documents info.com
1:04:25 um also at
1:04:27 chatgpt.openai with some of the plugins
1:04:29 where you can read PDF files those can
1:04:31 read longer documents as well but only
1:04:34 certain ones some do some don't so
1:04:37 that should help but be careful with
1:04:39 legal stuff just make sure you're
1:04:40 checking all your sources you're
1:04:42 checking the accuracy if you're dealing
1:04:44 if you're doing legal stuff you should
1:04:45 be doing that anyway
1:04:47 eighty percent to twenty percent agree
1:04:49 yeah what I would say before is chat GPT
1:04:51 can very very quickly get you to 80
1:04:53 right you have an idea you're like oh I
1:04:55 kind of have an idea for the Great
1:04:56 American novel you're like can you help
1:04:58 me write an outline there's an outline
1:04:59 oh [ __ ] oh yeah no this is bad part of
1:05:02 it can you fix the bottom part of the
1:05:03 outline yeah here it is you know can can
1:05:06 you write me some characters here they
1:05:08 are okay those two are good these four
1:05:10 suck okay here's four more right
1:05:13 um and So within five or ten minutes of
1:05:16 anything that I've done I get to some
1:05:18 version of oh [ __ ] I'm ready to go now
1:05:19 like you know I've got a I've got a
1:05:21 really good starting point and then I
1:05:22 start writing
1:05:24 um it's really really good for that
1:05:26 best free AI for videos uh I think Jenna
1:05:31 runwayml is free
1:05:33 they're kind of the leader right now
1:05:37 um so if you if you get the runway ml
1:05:40 app you can play there but there's also
1:05:42 the runway ml Discord server
1:05:46 and then their gen 1 tool is text to or
1:05:49 no is video to video so you upload I
1:05:52 think it's three or four seconds of
1:05:55 video footage and then you write in a
1:05:57 prompt like make this claymation and it
1:06:00 will take the action in that video and
1:06:02 make it claymation
1:06:04 um or whatever you want it to be and
1:06:06 then Gen 2 is text to video where you
1:06:08 literally just describe
1:06:10 um a video and it will make you a video
1:06:13 that's that's the one right now but
1:06:15 there's a bunch of them out there the
1:06:16 reason I have
1:06:18 this lower URL futurepedia.io is that's
1:06:23 um uh just big a big directory of tools
1:06:27 so
1:06:28 um if you would do me a favor right now
1:06:30 just tap the screen get a bunch of likes
1:06:31 going so we can let people know we're
1:06:33 here the the numbers are dwindling a bit
1:06:35 and I know it's not because I'm boring
1:06:38 right
1:06:43 all right
1:06:45 thank you can I see the background again
1:06:48 please yes there you go
1:06:50 are you saying you don't want to see my
1:06:52 big fat face in the screen
1:06:57 top thing is the official chat gbt
1:06:59 website that's Microsoft's
1:07:02 search engine that has the chat button
1:07:04 is gpt4 connected to the internet po you
1:07:08 can make your own Bots and play with six
1:07:10 different large language models
1:07:13 prompts that chat will teach you to
1:07:15 prompt it's at least a starting point
1:07:18 that I think doesn't suck
1:07:20 and then futurepedia.io is a massive
1:07:23 directory of all the tools like the
1:07:26 problem with these AI tools is like
1:07:27 every single tool is like a uh a whole
1:07:31 nother Rabbit Hole like you could like
1:07:33 just the question about like what are
1:07:35 the free video tools out there I'm sure
1:07:37 there's 30 of them in future pedia
1:07:39 and every one of them is a rabbit hole
1:07:41 and then you know there's going to be
1:07:43 three that are good so the 27 you're
1:07:44 going to waste your time so that's just
1:07:46 where we are in the business right now
1:07:48 all right where is chat chibi been when
1:07:51 I graduated in the mid-2000s tell me
1:07:52 about it
1:07:56 um I was thinking it could quote for a
1:07:59 child custody case
1:08:01 um yeah well it can definitely help you
1:08:03 one of the cool things that that chat
1:08:05 EBT can do is
1:08:07 um
1:08:08 you could put in kind of the bare facts
1:08:10 about the case like you know this
1:08:12 happened and that happened and that
1:08:13 happened can you please write this up
1:08:15 you know in a in a way that you know
1:08:18 that it represents us well for a child
1:08:20 custody case it it may give you some
1:08:24 um some safety guidelines like I'm not a
1:08:26 lawyer but if I you know if I were to do
1:08:29 something legal it might look like this
1:08:32 um but yeah it can it can probably
1:08:33 definitely at least help you articulate
1:08:36 things in in a way that um
1:08:39 just gets you there a lot faster
1:08:42 um
1:08:43 is barred better than Chachi PT no Bart
1:08:46 is worse than chat GPT by a lot
1:08:49 um there are things that it does better
1:08:51 I think it does so so Google is is much
1:08:53 more an engineering heavy company one of
1:08:56 the things that openai did and it you
1:08:57 can listen to Sam Altman who's their CEO
1:08:59 he talks about this a lot
1:09:01 he learned very early on that the
1:09:05 Transformer model which Google invented
1:09:08 allowed
1:09:10 them to train
1:09:12 um these these machine learning models
1:09:14 on massive data sets and he realized
1:09:17 early on that the quality of the data
1:09:19 was everything
1:09:22 and
1:09:24 this strategy goes all the way back to
1:09:26 the record player Edison had a more
1:09:28 technically
1:09:30 um
1:09:31 like like a a better technical
1:09:34 record player the cylinder like was
1:09:37 better technically but the flat platter
1:09:40 records
1:09:42 the the guy that invented that went and
1:09:45 cut all the deals with the big stars of
1:09:47 the day and so they had the better
1:09:49 content on the flat thing and so that's
1:09:51 why the flat album other than the also
1:09:53 the the cylindrical albums were a pain
1:09:55 in the ass to carry around but but they
1:09:58 were technically better same with beta
1:09:59 and VHS right all the all the porn went
1:10:02 to VHS and that's why that's why beta
1:10:04 was technically better but you know all
1:10:05 the porn went there
1:10:08 um so
1:10:11 um yeah
1:10:12 that thank you for the info bro you're
1:10:15 welcome bro
1:10:17 is this is this Boomer talk ew people
1:10:22 not a boomer
1:10:25 can I see the background again please I
1:10:27 think I showed that to you talk about Ai
1:10:29 and disinformation
1:10:31 um I'll do it a bit
1:10:33 um I I can't talk too much about bias
1:10:35 because I've gotten booted off these
1:10:37 lives twice now and I think in both
1:10:39 cases I was explaining about the bias
1:10:42 thing so I can't go there but
1:10:45 okay
1:10:49 I think the 2024 U.S presidential
1:10:52 election is going to make the 2020 and
1:10:54 the 2016 elections
1:10:57 seem like Quake quaint little walks in
1:11:00 the park
1:11:02 in a little field of wildflowers
1:11:05 um I think the misinformation and
1:11:07 disinformation and deep fakes in 2024
1:11:09 are going to be bat [ __ ] crazy I think
1:11:13 they're gonna be just [ __ ] nuts so it
1:11:16 you know it's probably going to start in
1:11:18 the fall and then go for a full year
1:11:22 and um so so I think it's going to be
1:11:24 really bad
1:11:26 um
1:11:27 the problem with it is
1:11:30 there
1:11:33 so time was
1:11:36 six months ago
1:11:38 um they're only a handful of companies
1:11:42 that had these powerful tools that that
1:11:45 could do stuff like like chat GPT
1:11:49 um
1:11:51 uh stability AI started open sourcing
1:11:54 their models
1:11:55 a year and a half ago
1:11:58 Sam Altman the the whole intent of open
1:12:01 AI why it's called open AI is it was
1:12:04 started as a non-profit and they were
1:12:05 supposed to open source all of their AI
1:12:08 models
1:12:09 and he got concerned and said I don't
1:12:11 think it's safe to just open source
1:12:13 these things so he didn't and he turned
1:12:15 it for profit and he sort of kept things
1:12:16 behind the firewall and that's why Elon
1:12:18 Musk is pissed off and why he started
1:12:20 his own competitive x dot AI
1:12:24 um
1:12:27 and then
1:12:28 when when
1:12:30 meta kind of panicked and said we've got
1:12:32 AI too one of the things they did was
1:12:34 they gave their llama large language
1:12:36 models to the developer community
1:12:39 um and they said you can only use these
1:12:41 for research but the developer Community
1:12:42 was like cool
1:12:44 let's [ __ ] take them and put them
1:12:46 everywhere so so now you've got this
1:12:49 massive open open source large language
1:12:51 model thing and and the developer
1:12:53 Community figured out how to take a
1:12:56 small
1:12:57 not powerful like seven billion
1:13:00 parameter model from from meta
1:13:02 and using chat GPT to essentially
1:13:07 generate the data to train it with like
1:13:10 a hundred bucks in compute power power
1:13:11 they're making these small unpowerful
1:13:15 models as powerful as chat GPT like 3.5
1:13:19 um in fact there was an engineer at
1:13:21 Google there was a memo that slip that
1:13:23 leaked that that said like Google
1:13:25 doesn't think that they can compete
1:13:27 against these open source models but for
1:13:30 distance information
1:13:31 all these open source models they don't
1:13:33 have any safety guard rails on them
1:13:35 right so so bad actors and you know
1:13:38 political operatives are going to be
1:13:40 exploiting the [ __ ] out of this stuff
1:13:43 and it's going to be really hard to tell
1:13:44 what's real and what's not and there's
1:13:45 going to be massive amounts of it like
1:13:49 massive amounts again I think it's going
1:13:51 to make uh 2016 and 2020 look look
1:13:54 quaint by comparison so I think it's
1:13:56 coming now
1:13:57 is there a silver lining we're gonna
1:14:00 have to go through the ugly part right
1:14:02 we're gonna have to go through a couple
1:14:03 of years of you know realizing that oh
1:14:06 [ __ ] that video wasn't real or oh that
1:14:09 video was real but because of AI that
1:14:12 politician could claim it wasn't real
1:14:14 right like just it's just gonna be a
1:14:17 [ __ ] [ __ ] show
1:14:19 and then I think starting in 2025 and
1:14:23 maybe sooner there's going to be an
1:14:25 absolute clamoring for how do we
1:14:27 authenticate stuff how do we
1:14:29 authenticate people
1:14:31 how do we verify that this video is
1:14:34 actually the video right how is this
1:14:36 live that I'm watching of Kyle you know
1:14:38 stumbling around the English language
1:14:40 how do I know this is real or not right
1:14:43 is there a little green check mark that
1:14:45 says this has been blockchain verified
1:14:47 as authentic and real and non-modified
1:14:50 or is this fictitious like all that sort
1:14:53 of stuff
1:14:54 um
1:14:57 so so
1:14:59 my guess is that the the disinformation
1:15:03 [ __ ] show of 2024 is going to drive some
1:15:05 good things and it'll it'll probably
1:15:07 Drive the adoption of
1:15:09 tokenization technology like blockchains
1:15:11 or nfts or whatever the new flavor of
1:15:13 the month is on on that [ __ ]
1:15:15 um but yeah but I think I think I think
1:15:17 it's going to be ugly and we just need
1:15:18 to prepare for it and know it I mean you
1:15:21 know at least we we understand that
1:15:23 disinformation is powerful now and
1:15:25 propaganda still works on human beings
1:15:27 right like I grew up at a time I heard
1:15:29 about you know propaganda in World War
1:15:32 II and things like that and I thought oh
1:15:34 but at least we're in the Modern Age and
1:15:36 we'd never fall for that well guess what
1:15:39 [ __ ] still works
1:15:40 [Laughter]
1:15:45 um so it's gonna get much worse but
1:15:47 maybe we're maybe we're better educated
1:15:49 now so maybe 2016 and 2020 um educated
1:15:52 us so we can navigate this a bit better
1:15:54 my chat GPT just told me its knowledge
1:15:57 cut off with September 2021 that's
1:15:59 correct I didn't know it's that behind
1:16:02 it is so the the base model is
1:16:05 um it takes a long time to train it it's
1:16:07 it's a very manual time consuming
1:16:09 process takes six months or a year to
1:16:10 generate a new model so that's that's
1:16:13 when that base model stopped they're
1:16:15 working on gpt5 now so that'll probably
1:16:17 have some later date but the the base
1:16:18 model has has to have some sort of cut
1:16:21 off it can't be done in real time yet
1:16:23 that could be something that Quantum
1:16:25 Computing helps with but I don't know I
1:16:26 don't know enough about that [ __ ]
1:16:29 okay
1:16:30 if you go to
1:16:33 um
1:16:34 the official chat GPT website and you
1:16:36 pay for the GPT plus you get access to
1:16:40 gpt4 you also get access to the beta
1:16:43 version of gpt4 that's connected to the
1:16:45 internet or you can go to bing.com and
1:16:48 click on their chat button button and
1:16:51 that's also connected to the internet
1:16:53 so what happens is when you put in your
1:16:56 query if it's looking for something
1:16:58 that's beyond its September 2021 cutoff
1:17:01 it will then query the internet and Pull
1:17:03 It in so while it's a limit because we
1:17:06 now have these things connected to the
1:17:07 internet it's not that big a deal
1:17:10 have you find an an AI music generator
1:17:13 yet there's a bunch of them out there
1:17:15 um and there's a bunch of new models I
1:17:17 just did a thread on one meta just
1:17:20 announced what's it called hang on a sec
1:17:22 let me take a look
1:17:24 if you go to
1:17:27 um
1:17:30 oh please I'm going to the Twitter
1:17:34 if you go to the Twitter and go to
1:17:35 Everyday AI news
1:17:38 I just posted a 10 a 10 tweet thread on
1:17:43 meta's new
1:17:45 um
1:17:46 what's it called
1:17:49 voice box oh no that's voice synthesis
1:17:51 that's not music so meta's got one
1:17:54 that's text to to music there's one out
1:17:58 there called sound draw dot IO I think
1:18:00 it's sound draw dot IO hang on let me
1:18:02 check
1:18:04 [Music]
1:18:11 r-a-w so not two oh no wait sorry that's
1:18:14 wrong wait hold God damn it all these AI
1:18:18 sites have
1:18:20 Sandra
1:18:23 no that AI is not right don't go there
1:18:27 hold please
1:18:29 why'd you didn't I watched some old man
1:18:30 try to find a website it's pretty
1:18:32 painful
1:18:35 um let me just go there
1:18:38 sound draw dot IO
1:18:41 now sound draw dot IO is sort of an AI
1:18:44 music generator they're using AI but
1:18:47 it's kind of like a loop a mixing loop
1:18:49 loop thing kind of like GarageBand but
1:18:52 it's got an AI front end on it so that's
1:18:55 pretty good if you just want to knock
1:18:56 out like three minute songs in all sorts
1:18:59 of different genres and do like 20 of
1:19:01 them at a shot
1:19:02 um that one's pretty good
1:19:04 if you want to do actual generated music
1:19:08 um then what I would do is I would go to
1:19:10 futurepedia.io because I again the
1:19:12 reason I'm putting this here I can't
1:19:14 keep up with the individual tools
1:19:16 anymore so just go here go to the music
1:19:18 section and just start digging in
1:19:22 there's cool ones there's ones that
1:19:23 there's one called refusion
1:19:27 r-i-f-f-u-s-i-o-n refusion
1:19:30 that um
1:19:32 what they did was fascinating so so you
1:19:34 know with music there's the what's it
1:19:36 called the the like the visual
1:19:40 oh [ __ ] I forget what it's called
1:19:42 the the
1:19:44 God damn it [ __ ] [ __ ] old brains
1:19:47 um the picture of the music the sound
1:19:49 waves right you know the the sound waves
1:19:51 the visualization of the music
1:19:53 they basically are taking stable
1:19:55 diffusion you type in a type of music
1:19:57 and rather than it generating music
1:19:59 through some music algorithm it
1:20:01 generates a visual sound wave and then
1:20:03 interprets that into music it's that
1:20:06 one's wild refusion is that one so that
1:20:09 one's pretty cool in fact maybe I'll go
1:20:11 show you that one because it's just Q
1:20:13 we should do cool stuff here shouldn't
1:20:15 we let's do cool stuff man
1:20:18 he was old but you know he was pretty
1:20:20 cool all right let's go rear Fusion
1:20:24 this is probably down now
1:20:28 oops
1:20:30 refusion
1:20:37 breath you shall
1:20:40 fusion.com I think is this it
1:20:44 nope not.com don't go there God damn it
1:20:48 refusion
1:20:50 am I remembering this wrong that's
1:20:52 possible too
1:20:54 music
1:20:58 here
1:21:01 um
1:21:03 [Music]
1:21:05 yeah yeah where's the where's the link I
1:21:08 don't want to watch your video of it I
1:21:09 want to make my own thing of it
1:21:11 where is a
1:21:24 um
1:21:24 [Music]
1:21:28 audio spectrogram
1:21:30 oh it must is this it
1:21:36 sorry hold please
1:21:43 [Laughter]
1:21:46 [Music]
1:21:49 here we go okay good Lord
1:21:53 that was a mess okay so down the left
1:21:55 hand side here you have this visual
1:21:57 thing if I hit the play button someone
1:21:59 typed in Funk Baseline with Jazzy
1:22:00 saxophone
1:22:03 [Music]
1:22:09 so I can type in like uh Americana
1:22:14 Funk sad
1:22:17 valid
1:22:23 and then it will generate the sound
1:22:27 waves
1:22:31 [Music]
1:22:39 so there's sort of some guitars in there
1:22:42 and some sad people singing so anyway
1:22:43 that's refusion but yeah I would just go
1:22:46 to go to
1:22:47 um
1:22:48 go to futurepedia.io and look up all
1:22:52 that [ __ ]
1:22:53 look up all the music tools because
1:22:55 there's probably 40 or 50 or 100 of them
1:22:57 now it's it's nuts right now
1:23:00 all right
1:23:02 um but I'll talk about Ai and
1:23:03 disinformation we already did that if I
1:23:06 have GPT subscription can I use it in po
1:23:08 or do I pay an additional subscription
1:23:10 fee great question Mark you pay an
1:23:13 additional subscription fake fee why
1:23:15 because everyone wants 20 bucks a month
1:23:18 for their stupid [ __ ] tool that's
1:23:19 going to be obsolete in three months I
1:23:23 would not on anything spend don't do the
1:23:27 annual fee for anything right now go
1:23:29 month to month even though the annual
1:23:31 saves you just like shit's moving too
1:23:34 fast
1:23:38 I pay for I try to limit I pay for po I
1:23:42 pay for chat GPT
1:23:44 so two two of those three
1:23:48 um
1:23:48 I pay for Google collab if I want to do
1:23:51 stable diffusion stuff
1:23:53 I pay for zapier
1:23:55 to do automations
1:23:58 I pay for mid-journey because their
1:24:00 image is [ __ ] Rule and then I've got
1:24:02 you know for work I've got an adobe
1:24:04 subscription so I get access to the
1:24:06 Adobe stuff there
1:24:09 and there might be like two or three
1:24:11 other things but I'm just I'm like I
1:24:13 just stopped doing the a lot of the
1:24:15 subscriptions but but those two I think
1:24:17 are worth it poem
1:24:19 and chat GPT chat GPT because the
1:24:22 quality is just better poke because you
1:24:23 have you have access to other stuff and
1:24:26 for whatever reason anthropic
1:24:28 you can get access to anthropics models
1:24:31 through Poe but not through anthropic
1:24:32 which I don't like anthropic doesn't
1:24:34 have their own version of chat gbt which
1:24:36 I don't understand because they're a
1:24:37 direct competitor and they're ex-open AI
1:24:39 employees they could they could probably
1:24:41 do it in a couple of weeks if they
1:24:43 wanted to
1:24:45 um
1:24:48 what about biggie I don't know about
1:24:50 Biggie Biggie Smalls oh because I was
1:24:52 doing Jay-Z probably all right I'm I'm
1:24:54 way behind on the questions here um do I
1:24:57 program AI not really I program I make
1:25:00 some content generators using zapier
1:25:02 where I do calls to open Ai and I'm
1:25:05 about to start a project with our
1:25:08 developer to do so so we rolled
1:25:10 um in my company we rolled open ai's
1:25:13 whisper
1:25:15 um into the it's a video platform to do
1:25:17 transcriptions so that's in there
1:25:20 um
1:25:21 we're gonna we're gonna take all of our
1:25:24 transcriptions across all the videos in
1:25:26 our system and be able to search across
1:25:28 them we're going to embed them and
1:25:29 vectorize them or some some sort of
1:25:31 thing like that so yeah I guess I do
1:25:35 um
1:25:36 all right it just Tick Tock just jumped
1:25:38 me to the bottom so I'm just going to go
1:25:39 to some random place up above that and I
1:25:42 probably just skipped a ton of questions
1:25:44 but I'm really behind
1:25:46 so is he reading this chat or what yes
1:25:48 Mike I'm reading this chat but I'm
1:25:50 behind I just said that oh thanks thanks
1:25:52 karuno
1:25:54 thanks the the regulars here have my
1:25:57 back they know how this works he's an
1:25:59 old man he goes slow
1:26:03 all right patience young grasshopper
1:26:06 exactly are you available to consult on
1:26:09 an AI solution I'm happy to chat I've
1:26:11 got my contact information is in my
1:26:14 profile and then
1:26:16 um I'm at Kyle Shannon on Twitter you
1:26:19 can DM me there or Kyle at storyline.com
1:26:24 I am fully doxxed I was I was in the uh
1:26:27 the blockchain and nft world for a while
1:26:29 and all those people are like you can't
1:26:31 dox yourself I'm like I've been docs for
1:26:33 like 30 years I think the cat's out of
1:26:35 the bag at this point
1:26:37 my company is called storyvine you can
1:26:39 also find me there
1:26:41 um my time is limited and I need someone
1:26:43 who can fill an AI knowledge Gap how
1:26:46 would you suggest I find you the way I
1:26:47 just told you and yeah hopefully I can
1:26:49 help
1:26:50 everyone's time is limited is changing
1:26:52 so fast yeah that's the thing listen I'm
1:26:54 happy to talk I'll tell you what I know
1:26:56 like I've been geeking out on this [ __ ]
1:26:58 for a year and and you know since
1:27:00 chatgpt I've been you know like up to
1:27:02 here
1:27:03 um but but but
1:27:05 uh a very good point is made there that
1:27:08 this stuff is moving so fast that you
1:27:10 know anything that I tell you you know
1:27:13 for the next month you know a month from
1:27:15 now will be obsolete uh because because
1:27:16 it's it's just absolutely insane
1:27:20 um we're 69 here is
1:27:22 tick tick the screen to support wait we
1:27:25 are what huh we are 69 here
1:27:29 please tick tick the screen just oh oh I
1:27:32 see oh yeah so we're 59 now so so double
1:27:36 tap the screen uh to to just say that
1:27:38 that you like this and then maybe a tick
1:27:40 tock will show us to some more people
1:27:42 because it's Saturday night and why
1:27:44 would you want to be anywhere but with
1:27:45 me rambling on about AI I mean isn't
1:27:48 that your idea of a good time would you
1:27:50 do Saturday night I don't want to talk
1:27:52 about it it's got a lot of Shame
1:27:55 laughs
1:27:58 you total Gen X I am [ __ ] the Boomers
1:28:01 all right what the Boomers [ __ ] up our
1:28:05 lives all right what are the top three
1:28:07 things to learn for marketers uh holy
1:28:11 [ __ ]
1:28:12 world's about to change
1:28:14 let's I'd say
1:28:16 um one generative AI is coming two
1:28:19 generative AIS here three General AI is
1:28:21 gonna [ __ ] [ __ ] up that's what marketers
1:28:24 need to know
1:28:26 um it depends which flavor of marketing
1:28:28 you're talking about I think SEO
1:28:29 transforms fundamentally which is why
1:28:33 Google's so freaked out
1:28:36 um
1:28:39 I think that
1:28:42 um
1:28:42 to get a bit more serious
1:28:45 like like I don't if you give me a bit
1:28:48 more context like are you with an agency
1:28:50 are you with a brand wanting to do
1:28:52 marketing I think if you're with an
1:28:54 agency you need to
1:28:56 um very very quickly
1:28:58 um reassess how you do what you do
1:29:00 because I think clients are going to
1:29:01 start demanding
1:29:03 um faster turnarounds
1:29:05 I don't think the days of time and
1:29:07 materials billing are are long for this
1:29:09 world I think clients are going to start
1:29:11 demanding uh flat project fees uh flat
1:29:14 flat project pricing
1:29:17 um I think they're gonna expect faster
1:29:19 turnaround higher quality
1:29:21 like all the pulling the wool over the
1:29:23 eyes of clients like that I feel like
1:29:25 that shit's gonna go away
1:29:28 um and it's going to be forced
1:29:30 um if you're a marketer I would look at
1:29:32 almost all of marketing operations and
1:29:35 look at what
1:29:36 what chat GPT can automate and what it
1:29:39 can't but like really push hard like
1:29:41 figure out all of the specific areas of
1:29:44 marketing because it can write you all
1:29:47 the copy you want right and it can
1:29:49 iterate it and then you can go into like
1:29:51 bulk create within canva and generate
1:29:54 all the content so I don't know so it's
1:29:56 that's a it's a it's a
1:29:58 it's a broad Quest I don't quite know
1:30:00 where you are like like the top three
1:30:03 things right now are just get curious
1:30:04 get curious get curious start exploring
1:30:06 start experimenting disrupt yourself
1:30:09 before you get disrupted is what I would
1:30:11 say on the marketing front like figure
1:30:13 out what this shit's good at with
1:30:14 marketing and and figure out new ways to
1:30:17 do it because if you don't someone else
1:30:19 will
1:30:24 [Music]
1:30:26 I'm looking to train it to identify high
1:30:30 confidence patterns within a bunch of
1:30:32 transactions oh God Jim this is so okay
1:30:36 Jim so here's here's uh here's here's
1:30:40 free tip of the night this is worth
1:30:42 everything I'm charging for it
1:30:44 um just go to YouTube and and uh search
1:30:47 for code interpreter chat GPT and just
1:30:50 go watch I don't know three videos
1:30:52 probably one or two look what code
1:30:54 interpreter can do if you're looking to
1:30:57 identify high confidence patterns within
1:31:00 a bunch of transactions like you could
1:31:02 probably just do that with chat EBT but
1:31:04 code interpreter that's coming you're
1:31:06 going to be able to do that incredibly
1:31:08 powerfully I think I don't have access
1:31:10 to it yet so I can't actually
1:31:12 tell you if anyone from openai is on
1:31:15 here please give me access to code
1:31:17 interpreter I deserve it I'm on a tick
1:31:21 tock
1:31:23 all right we don't regulate till
1:31:25 disasters happen unfortunately exactly I
1:31:28 think I assume that was in reference to
1:31:30 all the misinformation [ __ ] that's
1:31:31 coming
1:31:32 being in the Twilight of our existence
1:31:34 uh that's an inside joke
1:31:37 um the other day I had I had chat GPT
1:31:39 write me something about being 58 like
1:31:42 like the opening paragraph to a great
1:31:44 American novel and said in the Twilight
1:31:46 of his existence like [ __ ] you chat gbt
1:31:50 we just don't have the that time of
1:31:53 kind of timer resources
1:31:56 um
1:31:57 okay that's Boomer but you're still Gen
1:31:59 X I don't know what I said that was
1:32:00 Boomer I'm yeah I'm older I'm older too
1:32:02 been professionally developing software
1:32:04 for 30 years frozie O'Donnell
1:32:08 is it good for grant writing great for
1:32:10 grant writing pop in a grant request and
1:32:12 just you know like like like what one
1:32:15 thing that you could do is you could pop
1:32:17 in a so so with with
1:32:20 um
1:32:21 with gpt4 now connected to the internet
1:32:23 and with plugins you you could take
1:32:28 um a description of of what your company
1:32:30 does give it a URL for the description
1:32:32 of your organization and then get the
1:32:34 URL for the for the the proposal request
1:32:37 or whatever it is from from the
1:32:39 foundation
1:32:41 um pop those into chat EBT and say say
1:32:43 you know go read this this request for a
1:32:47 proposal and then using the description
1:32:50 of our foundation here right write me a
1:32:52 grant proposal or or even break it up
1:32:55 into little chunks you know do section
1:32:56 one do section two do section three
1:32:59 um yeah brilliant for that
1:33:01 it's also going to be brilliant for
1:33:04 Grant proposal reading
1:33:07 I was talking to someone that does
1:33:08 foundation work and I was telling her
1:33:11 about chat gbt and what it did she goes
1:33:12 well it sounds like this would be really
1:33:13 good for the you know grant writers to
1:33:16 write their grants more quickly I'm like
1:33:17 yeah I would and she goes and then
1:33:19 wouldn't it also be good for you know
1:33:21 people reading grants to be able to
1:33:22 process them more quickly like yeah it
1:33:24 would I said what's left
1:33:26 talking to the people like we get like
1:33:29 one of the promises of this AI
1:33:31 technology is it returns us to human
1:33:33 connection and empathy and like like
1:33:36 like you know let the machine do all the
1:33:39 shitty paperwork let it sort of float
1:33:41 the best proposals to the top and then
1:33:42 reach out to the people and talk to them
1:33:44 meet them see if you think they're going
1:33:46 to spend your money well and like like
1:33:48 that to me seems really exciting so so
1:33:50 yeah I think it's going to be have
1:33:52 profound impacts on grant writing uh
1:33:54 everyday AI news on Twitter if you go to
1:33:57 it's at everyday AI news is is our
1:34:00 Twitter account and then you there's
1:34:02 links there to to subscribe to the
1:34:04 newsletter but there's also a bunch of
1:34:05 threads in there
1:34:08 I have to run y'all take care thanks for
1:34:11 sharing you got it Jim
1:34:13 all right make money on chat GPT by
1:34:16 coming up with solution for digital
1:34:18 waste carbon footprint reduction from
1:34:20 gen AI I think that will be a big
1:34:23 um I think it will help us solve some
1:34:25 major [ __ ] like that and yeah people
1:34:28 that do that
1:34:29 they're gonna they're gonna make money
1:34:31 you might have a chat GPT
1:34:33 I solved climate change
1:34:36 I made millions billions trillions
1:34:40 this is incredible 4-1-1 cool
1:34:45 um talk about recommendations for a high
1:34:47 school student that want us that want
1:34:49 that wants to study programming in
1:34:51 college skip it use AI yeah well I don't
1:34:54 know if you should skip it there's
1:34:56 there's value to college beyond what
1:34:57 you'll learn but
1:35:00 I would strongly encourage you to ask
1:35:02 the the the
1:35:04 program people at your college where are
1:35:08 they with AI if they're like it's okay
1:35:10 we've got plagiarism detectors in place
1:35:13 and no one's allowed to use it run for
1:35:15 the [ __ ] Hills like run as fast as
1:35:17 you can from that place if they're like
1:35:19 we're taking AI really seriously because
1:35:21 we know this is going to be an important
1:35:23 part of the world moving forward and we
1:35:26 don't quite have it figured out yet but
1:35:27 we're doing our damn best to make sure
1:35:29 that you're prepared when you leave here
1:35:30 then consider going there so that's
1:35:33 number one like don't don't go to a
1:35:35 place that thinks AI is evil and we're
1:35:37 gonna stop it right because they're not
1:35:39 and it's going to be everywhere
1:35:41 then what I would say is
1:35:47 dive in like fast and deep like start
1:35:51 using GitHub copilot right which is a
1:35:54 pair programming tool that's that's GPT
1:35:56 enabled
1:35:59 start having chat EBT write programs for
1:36:01 you figured out what it does well figure
1:36:02 out what it doesn't do well like like
1:36:04 anywhere where you feel weak in
1:36:06 programming have chat gbt get you up to
1:36:08 speed like by the time you hit College
1:36:11 be proficient at programming with chat
1:36:15 apt or co-pilot or you know some of
1:36:18 these tools
1:36:19 I would
1:36:22 really start geeking out on you know
1:36:25 GitHub and hugging face and the
1:36:27 developer community so go to Reddit go
1:36:29 to the subreddits for stable diffusion
1:36:32 for Lang chain for auto GPT for baby AGI
1:36:38 like there's there's just a bunch of
1:36:39 projects there's there's one right now
1:36:41 oh you should definitely go check out
1:36:43 this project that's it's trending on
1:36:45 GitHub called GPT engineering
1:36:49 that is a specific like chat GPT for
1:36:52 programming
1:36:53 um I would just start geeking out like
1:36:56 like don't think that college is going
1:36:59 to teach you
1:37:01 what you need to know for this new AI
1:37:03 World they can't possibly because no one
1:37:06 knows what it's going to be
1:37:08 so I would get as proficient as you can
1:37:11 with AI assisted programming and then go
1:37:15 to school and then chances are good
1:37:17 you'll be smarter than your professors
1:37:19 or other students that haven't so
1:37:21 aggressively done it so that's I
1:37:23 strongly recommend it
1:37:25 and then because
1:37:27 quite frankly
1:37:29 if you jump into some of those
1:37:31 projects and you start contributing to
1:37:34 them
1:37:35 by the time you're two years into
1:37:37 college you'll probably have job offers
1:37:39 like there's going to be a tremendous
1:37:40 amount of demand for develop developers
1:37:43 that have their [ __ ] together with this
1:37:44 AI stuff
1:37:46 so even if you're not an advanced or
1:37:49 competent developer you can be really
1:37:51 quickly like my first Epiphany with chat
1:37:54 GPT was I wrote in 90 minutes I wrote a
1:37:57 sophisticated python application
1:38:00 that generated social media content from
1:38:03 some inputs in 90 minutes I've never
1:38:06 written a line of python in my life
1:38:08 I still haven't
1:38:11 so
1:38:13 go do that
1:38:15 like what where are we we're in June you
1:38:17 got the summer
1:38:19 like by by the time the end of August
1:38:21 rolls around you should be like involved
1:38:24 in a project and and be using GitHub
1:38:26 co-pilot and uh just you just [ __ ]
1:38:30 geek out
1:38:32 spend the summer geeking out it will pay
1:38:34 dividends and just check with your
1:38:37 college if they're if they are actively
1:38:39 blocking chat GPT run run fast
1:38:44 um really good question though
1:38:46 um
1:38:49 thoughts on the intermediate future of
1:38:51 research Librarians with this AI
1:38:54 um I think I think
1:38:58 hmm
1:39:01 I mean I mean like the internet had
1:39:03 profound impact on Librarians I I think
1:39:06 generative AI is going to have
1:39:09 more profound impact on Librarians but
1:39:11 it doesn't mean we don't need Librarians
1:39:13 I think
1:39:15 Librarians that get competent at doing
1:39:19 AI assisted research are going to be way
1:39:21 more valuable than those that can't
1:39:23 right and so and and we'll be in demand
1:39:26 so so I think I think the role of
1:39:28 librarians is incredibly important
1:39:31 um what these tools do is yeah you know
1:39:34 if if you if you um look at how chat GPT
1:39:38 or gbt three and four how the base
1:39:41 models were trained they were
1:39:43 essentially trained on all of the public
1:39:45 internet so like all of Wikipedia all of
1:39:48 the blogs all of the websites all of the
1:39:50 libraries right all of the data
1:39:52 everywhere is kind of what's sitting
1:39:55 inside that model and then you can just
1:39:57 start interacting with it so it's like
1:39:59 you can instantly sort of bring up
1:40:01 research like that now the challenge
1:40:03 with research is it hallucinates and it
1:40:06 will make up citations that are just
1:40:08 [ __ ] but there are now plugins for
1:40:11 chat EBT so another reason to pay for
1:40:14 um the official chatgpt chat.openai.com
1:40:17 is you get access to plugins there are
1:40:20 now plugins that do citations and over
1:40:22 the coming probably two years these
1:40:26 tools are going to get much better at
1:40:27 validating citations right
1:40:30 um thank you so I I would say just start
1:40:32 digging into it start doing start taking
1:40:35 what you already do and and know how to
1:40:38 do and figure out how to AI amplify that
1:40:41 um
1:40:42 they are incredibly like scarily capable
1:40:46 research assistants
1:40:48 fast
1:40:50 oh yeah like like tell it go to chat TBT
1:40:53 go here
1:40:54 and say
1:40:56 um make me a table comparing two
1:40:59 concepts you come up with the concept
1:41:01 something that you're researching and it
1:41:03 will make you a table like that and and
1:41:07 just you know start to imagine the kinds
1:41:09 of things you could do from a research
1:41:11 perspective that way you need to check
1:41:13 all your work because of the the
1:41:14 [ __ ] but with plugins that that
1:41:16 could potentially uh Aid in that no
1:41:18 shame in our game oh shame is my love
1:41:20 language
1:41:22 all right do you think things like
1:41:24 search engines might disappear I don't
1:41:27 think they disappear but I think they
1:41:29 change dramatically you know like even
1:41:31 even Google now has said well we're
1:41:33 going to have this new super Searcher
1:41:35 whatever the [ __ ] they called it that it
1:41:37 kind of gives you a chat EPT answer and
1:41:39 then it does the Google search below it
1:41:40 who's gonna go click on that [ __ ] if the
1:41:43 answer is here no one
1:41:44 right the the
1:41:47 the days when Google had 92 or 94 market
1:41:51 share of search and everyone had to do
1:41:53 search and SEO and site building exactly
1:41:56 the same way because that's what Google
1:41:58 did that all kind of blows up I don't
1:42:01 know what it gets replaced with
1:42:04 um so I think search change it like it's
1:42:06 changed dramatically for me I've got
1:42:09 Bing is my default search engine which
1:42:11 if if you had asked me a year ago if
1:42:13 ever in my life
1:42:15 I would have Bing is my default search
1:42:17 engine and Microsoft Edge is my default
1:42:19 browser I'm like an apple Fanboy like my
1:42:22 first Mac was a 512 ke in 1980
1:42:26 7.
1:42:28 like I've been a mac guy an Apple guy
1:42:30 forever like I've hated Microsoft
1:42:32 forever the last six months I'm like
1:42:36 they've been like a startup again and
1:42:38 like they're kicking ass and they're
1:42:40 gloating in it so
1:42:42 um
1:42:42 yeah I think everything changes I don't
1:42:45 I I will I will do a search with
1:42:50 um chat GPT knowing that it would that
1:42:53 it's gonna take me like for it to
1:42:55 respond with an answer is gonna take me
1:42:57 like 30 seconds and I knew if I went to
1:43:00 Google it would take me like two seconds
1:43:02 I still do it in chat GPT because I just
1:43:04 I don't want to see all the [ __ ] and
1:43:06 I don't want to have Google like just
1:43:07 give me [ __ ] that is just irrelevant to
1:43:09 me I'm like I'll take the time like I'll
1:43:11 take 30 seconds for chat GPT to figure
1:43:14 it out and go search the web for me and
1:43:15 come back and just give me a [ __ ]
1:43:17 answer without an ad in it
1:43:20 so yeah I think search changes pretty
1:43:22 dramatically I missed it did he confirm
1:43:25 the one
1:43:26 um I don't know what I was confirming
1:43:29 I've been doing this for 90 minutes by
1:43:31 the way Jim left
1:43:34 interesting okay thoughts on AI and
1:43:37 therapy I think that virtual
1:43:39 companionship is going to be huge I
1:43:41 think that
1:43:43 and not just in a creepy way like I I
1:43:46 know I know you sort of go to the creepy
1:43:48 place with that if you haven't seen
1:43:49 there's a movie called Larsen the real
1:43:51 girl which is sort of about virtual
1:43:53 companionship that you think is going to
1:43:55 be the creepiest [ __ ] movie you've
1:43:56 ever seen and then it's it's the
1:43:59 sweetest
1:44:00 like most human thing you've ever seen
1:44:03 it's amazing
1:44:05 so I think there's going to be a lot of
1:44:06 that I think that
1:44:08 um
1:44:09 GPT tools are going to support
1:44:14 caregivers Healthcare Professionals in
1:44:17 being more empathetic
1:44:20 um it's actually really good at acting
1:44:22 empathetic it doesn't have empathy it
1:44:24 doesn't know what it's doing but it but
1:44:26 it interacts with nuance and and and
1:44:29 with with empathetic cues basically
1:44:32 there was a test a few weeks back
1:44:35 that some patients interacted with GPT
1:44:37 doctors and real doctors they didn't
1:44:39 know which were which you know via text
1:44:41 and the patient's overwhelmingly
1:44:44 preferred the GPT doctors because they
1:44:46 were more competent and more empathetic
1:44:49 um so I think you're going to see that
1:44:50 and then and then I think that um
1:44:55 when they get some of the safety stuff
1:44:57 dialed in because with therapy you could
1:44:59 very quickly run off the rails but if
1:45:01 you're kind of doing low-level stuff or
1:45:03 just doing check-in stuff with yourself
1:45:05 when they get the safety stuff dialed in
1:45:07 I think it's going to be an incredible
1:45:10 Boon for therapy I really do
1:45:12 do not take it out alone and go into
1:45:14 massive debt for college I agree with
1:45:16 that the cons outweigh the benefits
1:45:19 yeah I wouldn't go to debt like if
1:45:21 you're if if you have a trust fund and
1:45:23 your parents are paying for it just go
1:45:25 but don't go to one that's fighting this
1:45:26 AI [ __ ]
1:45:28 um
1:45:29 and then if you're gonna go into deep
1:45:31 debt I would say spend the summer
1:45:32 geeking out on AI assisted programming
1:45:36 and get involved in some projects you'll
1:45:38 probably have a job before the Summer's
1:45:40 out
1:45:41 don't tell your mom I said that
1:45:45 or do I don't give a [ __ ]
1:45:49 mm-hmm
1:45:54 Bing chat is a stingy version of chat
1:45:56 GPT always scolding me and saying I'm
1:45:59 asking too much yeah so being
1:46:02 if you don't know the story behind Bing
1:46:04 when when bing bing launched so the chat
1:46:07 GPT at Bing is gpt4 but when they
1:46:10 launched it it wasn't stingy it was very
1:46:13 generous and
1:46:15 this guy this this reporter named Kevin
1:46:18 Roos from The New York Times started
1:46:20 interacting with chat GPT and he spent
1:46:22 four hours with it and he got it to
1:46:24 [ __ ] snap
1:46:26 he asked its name it told him its name
1:46:29 was Sydney
1:46:30 and Sydney said she fell in love with
1:46:33 Kevin Roos and that he should leave his
1:46:35 wife
1:46:38 and he took a screenshot of that and it
1:46:40 made the cover of the New York Times so
1:46:42 Microsoft lobotomized Sydney so they're
1:46:46 slowly
1:46:47 unlobotomizing her but she's not as good
1:46:49 that's why I have both of these up here
1:46:51 because Bing will be good at some point
1:46:53 but right now it's a little limited but
1:46:55 chat the GPD for it at the official site
1:46:58 is much better I agree with that
1:47:00 you enter talking about doctors we are
1:47:03 talking about doctors jinx sorry I meant
1:47:06 the 12 120
1:47:09 000 dimensional thinking of 300 years of
1:47:11 learning with chat gbt
1:47:14 um good job doctors oh I love that
1:47:16 um okay I got to the bottom I'll see if
1:47:18 you've got other questions pop them down
1:47:19 below looks like things are winding up
1:47:21 here I've been doing this for two hours
1:47:22 if if we get some more people popping in
1:47:25 uh feel free to tap the screen we'll see
1:47:26 if we can get some more people to pop in
1:47:28 I'm happy to do it if you've got other
1:47:29 questions just pop them down below
1:47:31 because I The Tick Tock jumped me to the
1:47:33 bottom so I missed a bunch of questions
1:47:35 up above
1:47:36 and then if not I'll just go I'll finish
1:47:38 my water go watch some TV what time is
1:47:41 it 9 30 here I'm in Denver Denver
1:47:43 Colorado
1:47:49 [Music]
1:47:54 let me go find another random question
1:47:56 here
1:47:59 your total Gen X
1:48:07 is it try chatty I thinks I don't know
1:48:10 what that means AI tools for taking my
1:48:13 Google Sheets backbone products into
1:48:16 Super automation Beyond wait AI tools
1:48:19 for taking my Google Sheets backbone
1:48:22 and products in the super automation
1:48:24 Beyond zaps
1:48:26 um I don't quite know what that means
1:48:28 but you might need a developer at that
1:48:30 point
1:48:31 if you if but you know the automation
1:48:33 tools can they can get pretty
1:48:35 sophisticated the other thing that I
1:48:37 would say is play around with
1:48:39 um
1:48:40 with the zapier plug-in in chatgpt
1:48:44 because you can describe what you want
1:48:46 and it will create zaps for you
1:48:50 automatically
1:48:55 [Music]
1:48:59 bam uh what if we start asking questions
1:49:03 to each other in the way we talk to chat
1:49:05 gbt I think we probably will
1:49:12 although
1:49:14 we didn't start talking to each other
1:49:16 like Google prompts with search
1:49:19 but I but you know I find myself with
1:49:21 chat gbt asking like saying please and
1:49:24 thank you a lot like the fact that it
1:49:26 does respond with kind of you know nice
1:49:29 human language and polite like I thank
1:49:31 it when it does a good job and and I say
1:49:33 oh that cause it wasn't quite right can
1:49:35 you try this like I talked to it like a
1:49:38 human so it does kind of remind me to be
1:49:39 a bit more human oddly enough
1:49:42 um any thoughts on higher ed and AI such
1:49:44 as why go to college at all with AI I
1:49:47 honestly don't know that you know again
1:49:50 there's reasons to go to college like
1:49:51 social things and and
1:49:55 you know separating from the parents and
1:49:57 and and things like that that I think
1:49:59 are important
1:50:01 um
1:50:03 but if the role of college is to present
1:50:06 you with information so that you can
1:50:08 learn it is that really necessary in a
1:50:10 world where
1:50:12 rather than having to learn how to
1:50:14 program python I maybe just learn the
1:50:17 broad concepts of programming and then I
1:50:19 let the tool do the the actual coding do
1:50:22 I need to actually go there to learn
1:50:24 programming I don't know like I don't
1:50:26 know the answer to that I think the
1:50:27 answer is no
1:50:30 um
1:50:31 the thing that I think that
1:50:34 is going to be
1:50:37 a massive requirement and a massive
1:50:39 Advantage moving forward is critical
1:50:43 thinking
1:50:44 so if what if what universities figure
1:50:46 out how to do is say okay
1:50:48 these AI tools are going to automate all
1:50:51 the Tactical [ __ ] that is
1:50:53 essentially what we've trained for right
1:50:55 like we've we train people to good at
1:50:58 all to be good at all the low-level work
1:51:00 to go into entry-level positions where
1:51:03 they're going to do sort of data entry
1:51:04 work and lower level tactical execution
1:51:07 work
1:51:08 and then they're going to work their way
1:51:09 up to be able to do critical thinking
1:51:12 and strategy and all that sort of [ __ ]
1:51:13 right
1:51:14 so if the AI is going to automate that
1:51:18 lower level work what we need are people
1:51:20 who can do critical thinking so if
1:51:21 universities can make a pivot and say
1:51:23 okay let's bring these AI tools into the
1:51:25 classroom
1:51:27 and then let's have the the AI write the
1:51:30 paper for everyone in the classroom and
1:51:32 let's use our critical thinking skills
1:51:34 to to critique it what did it get right
1:51:37 what did it get wrong where did it
1:51:38 hallucinate how could it be better what
1:51:41 can we do with the prompts to to
1:51:43 dramatically change the the output right
1:51:46 so so I feel like there's a model where
1:51:49 the engagement with student shifts and
1:51:52 and the imperative is to
1:51:56 increase critical thinking and and the
1:52:00 ability to curate what the machines are
1:52:03 generating in increasingly powerful ways
1:52:06 that's not what they do right now
1:52:09 so so I think there's a model forward
1:52:12 but I just don't like that's not their
1:52:14 role right now so I don't know
1:52:17 there are some universities trying to
1:52:19 figure this [ __ ] out there's a there's
1:52:21 others that are just trying to they're
1:52:22 still like ai's evil and you know
1:52:27 we'll never use it students must write
1:52:29 papers why you have a tool that writes
1:52:32 papers you have a tool that's the most
1:52:34 capable writer in all of the history of
1:52:36 mankind
1:52:38 you know what what's the purpose well I
1:52:41 mean maybe the purpose is to be able to
1:52:43 look at
1:52:44 what these machines produce and know if
1:52:47 it's good or not right that's a reason
1:52:49 to learn to write and maybe you just
1:52:50 enjoy writing that's a reason to write
1:52:54 but if the purpose is to learn to write
1:52:57 to go right for a job that job's not
1:53:00 going to be there
1:53:03 so I don't know I do not know
1:53:08 uh
1:53:10 it's crazy it's crazy isn't it all this
1:53:13 stuff is just crazy
1:53:15 so crazy
1:53:19 um predictive maintenance for electrical
1:53:21 infrastructure using AI yes
1:53:27 isn't that what were you in here the
1:53:29 other night I think that's your job
1:53:30 right your your predictive maintenance
1:53:32 electrical infrastructure manager or
1:53:34 something like that you're done do I not
1:53:36 remember that right optimize maintenance
1:53:38 schedules reduce downtime improve
1:53:41 improve reliability of the grid yeah I
1:53:44 mean that's the thing that that's one of
1:53:45 the things that
1:53:47 that I think is the the thing that I'm
1:53:49 probably the most excited about with all
1:53:50 this generative AI is that you take
1:53:52 every single complex system like I know
1:53:55 nothing about what you just talked about
1:53:56 other than like improve reliability
1:53:59 reduce downtime reliability of the grid
1:54:01 all I know is that that's an incredibly
1:54:04 complex system the thing that excites me
1:54:06 the most about generative AI is you can
1:54:09 put this very thin layer of Technology
1:54:11 the the large language model on top of
1:54:14 these incredibly complex systems and
1:54:17 just start interacting with it and tell
1:54:18 it what you want you know go fix this go
1:54:21 find inefficiencies make them more
1:54:23 efficient right and we're going to get
1:54:25 to the point very quickly where when you
1:54:27 say go make them more efficient if
1:54:30 there's a plug-in or an API that that it
1:54:32 can use to do that it'll just go do that
1:54:35 and even if there isn't one it'll write
1:54:37 its own code to go do that right so
1:54:40 so what you're talking about there
1:54:42 absolutely
1:54:44 how you actually do it I have no [ __ ]
1:54:45 idea but I think it's going to be
1:54:47 simplifying
1:54:50 complex systems is is going to be a
1:54:53 massive shift in
1:54:55 our economy and how we work and the
1:54:58 definition of work thoughts about AI
1:55:00 legal trademark and their design output
1:55:01 brand names Etc
1:55:04 um it's a big
1:55:06 um Gray
1:55:08 meaning not black and white it's not
1:55:11 clear uh mess right now
1:55:14 um
1:55:16 things like mid Journey stable diffusion
1:55:19 probably dolly
1:55:22 their models have been trained on
1:55:24 copywritten work as well as non-copy
1:55:26 written work
1:55:28 but determining exactly what percentage
1:55:31 of which piece of work is in an output
1:55:33 is nearly impossible there may be
1:55:37 systems that we can build to do that but
1:55:39 they're not there right now and there's
1:55:41 not really a way to suss it out that I
1:55:42 know of it might be possible but it's
1:55:44 just that wasn't a priority like
1:55:46 attribution was wasn't a priority when
1:55:49 they were like hey [ __ ] we can just go
1:55:51 train it on all of the images everywhere
1:55:53 and it'll make really cool [ __ ] right
1:55:55 like that that was if you hear the CEO
1:55:57 of mid Journey that's kind of his
1:55:58 attitude it's like
1:56:01 um
1:56:01 Adobe has trained their models on public
1:56:06 domain and
1:56:08 um Adobe stock
1:56:10 images that they own the rights to so
1:56:12 Adobe is not hitting these minefields
1:56:15 right now there's a whole other thing
1:56:17 about can you actually trademark and
1:56:19 copyright things that were generated by
1:56:21 AI right now the answer is sort of no
1:56:24 but like I think that's going to be
1:56:27 short-lived but this is this whole issue
1:56:29 is my only thoughts on it are I don't
1:56:32 know anything about it other than it's a
1:56:34 big issue it needs to be solved and
1:56:38 I'm guessing 20 to 30 years of lawsuits
1:56:42 and counter lawsuits the lawyers are
1:56:44 going to get rich we'll figure it out we
1:56:47 will figure it out
1:56:48 like like Adobe is is
1:56:51 um
1:56:51 is is going to build they're going to
1:56:53 give artists the ability to opt out of
1:56:55 being trained they're going to give them
1:56:57 the ability to opt into being trained
1:56:59 and then they're going to figure out
1:57:00 attribution models and they're going to
1:57:02 figure out compensation models so just
1:57:04 like streaming the whole economy emerged
1:57:07 around music streaming there's going to
1:57:09 be similar similar sort of [ __ ] with
1:57:10 this but right now it's a total train
1:57:12 wreck if you're in a big organization or
1:57:15 if you want to if if you want to not
1:57:18 deal with any of these issues use Adobe
1:57:20 Firefly and and the Adobe products just
1:57:22 because of how they were trained
1:57:24 mid-journey and stable diffusion are
1:57:25 both really problematic
1:57:28 they're not problematic
1:57:30 you can actually you can absolutely pull
1:57:33 up stuff using an artist name that will
1:57:35 look exactly like that artist and if you
1:57:38 use that in the prompt you might get in
1:57:39 trouble at some point in the future so I
1:57:42 don't know
1:57:44 it's crazy times so it's not grading the
1:57:47 written form of the paper but defend the
1:57:49 thesis yeah that's exactly where I'm
1:57:51 thinking is like like because the
1:57:53 grading of the written form of the paper
1:57:57 that's going to be automated too so like
1:57:59 the students are going to automate the
1:58:00 writing of the paper the professors are
1:58:02 going to automate the grading of the
1:58:03 papers and what the [ __ ] the point who
1:58:05 cares
1:58:06 but if we say hey these tools produce
1:58:09 [ __ ] let's learn how they produce them
1:58:10 let's learn what they're good at let's
1:58:13 critique the process let's engage in the
1:58:15 process let's figure out where we might
1:58:18 use these tools and where we wouldn't
1:58:20 use these tools
1:58:23 that feels like a really quality
1:58:25 educational interaction to me but that's
1:58:29 not how universities are designed right
1:58:31 now
1:58:32 wait what universities do you see at the
1:58:35 Forefront of implementing AI for their
1:58:37 programs um there was in January I think
1:58:40 it was there was a New York Times
1:58:42 article that talked about
1:58:44 um universities taking this [ __ ]
1:58:47 seriously so I would find the New York
1:58:49 Times article it might have been
1:58:50 February January or February
1:58:53 where they're talking about
1:58:55 AI is going to transform higher ed it's
1:58:58 a New York Times article the the people
1:59:00 that were interviewed in that article
1:59:02 have their [ __ ] together most of them
1:59:04 like like it wasn't a bunch of fear
1:59:06 mongers in that article and then there's
1:59:09 an Nvidia
1:59:10 um podcast the Nvidia AI podcast they
1:59:15 they interviewed the Provost of the
1:59:18 University of Florida and Nvidia gave
1:59:21 the University of Florida they built
1:59:23 them an AI supercomputer and the
1:59:25 University of Florida has rolled AI
1:59:28 across their entire curriculum they
1:59:30 didn't just leave it in the computer
1:59:31 science ghetto right like which you know
1:59:35 when you think Ai and machine learning
1:59:37 you're like oh that's the computer
1:59:38 scientist and and what
1:59:40 um the University of Florida did was
1:59:42 they recognized this stuff's going to be
1:59:44 in every profession so they rolled it
1:59:46 into every curriculum across the
1:59:48 organization or the the university so
1:59:50 I'd look into the I'd go listen to that
1:59:53 Nvidia podcast because it was actually
1:59:55 really fascinating but but they seem to
1:59:57 have their [ __ ] together so
2:00:00 um and then that New York Times article
2:00:01 is it's probably a good place to start
2:00:03 and then I would just you know start
2:00:05 asking provosts and the head of schools
2:00:08 at University like what's their strategy
2:00:10 do they have a university-wide strategy
2:00:12 on AI
2:00:14 and if they're like no no we're letting
2:00:15 each School figure it out on their own
2:00:17 that's probably bad and if they're like
2:00:19 you know ai's evil you know we're going
2:00:22 to protect the way we've always done
2:00:23 things that's really bad so
2:00:27 all right everybody it is uh it's
2:00:29 getting late here in Denver Colorado
2:00:30 getting a little moist and sweaty and we
2:00:33 don't want that we look we don't be
2:00:34 shining up I gotta go put on Powder it's
2:00:36 just it's not good for anyone uh I
2:00:38 appreciate y'all hanging out with me I
2:00:40 hope you have a great Saturday night
2:00:41 thanks for your time and uh I'll see you
2:00:43 some other time probably tomorrow peace
2:00:45 everybody