aiLL 7/16/23 - The Future of Generative AI

Description

Chapters

Transcript

0:09 foreign
0:28 welcome to the AI learning lab
0:32 got a situation going on in the hair
0:34 hair department but you know
0:37 it's part of the situation
0:38 [Laughter]
0:43 welcome to the AI learning lab if you're
0:46 new to generative AI that's what this is
0:49 all about well even if you're not new to
0:51 generative AI it's still about this
0:54 we are talking about Chachi p and the
0:58 like and anything else we want to talk
1:00 about because that's how we roll
1:04 welcome welcome welcome happy Sunday
1:07 night
1:08 I have not been thinking about AI much
1:10 today so I don't have a lot of a ton of
1:13 new updates
1:15 but good evening very sorry I missed
1:17 your answer a couple of days ago about
1:19 best way to clone my voice oh yeah you
1:22 missed a good one I did it good
1:25 I did 11 labs
1:27 and did
1:30 I can show that again tonight
1:32 Mr Stone Mr Stone
1:38 see is it hot is it I mean it is hot
1:42 but is it cooling enough that I should
1:44 have the window open
1:46 or should I put on the air conditioner
1:48 hmm choices choices choices welcome
1:51 welcome welcome do me a favor if you
1:54 would hit the hit the like button share
1:56 the live let's see if we can get some
1:58 folks
1:59 out there
2:01 join them in the fun
2:04 my glasses are incredibly dirty
2:09 I'm just doing just doing life
2:10 maintenance over here
2:12 [Laughter]
2:15 welcome everybody to the AI learning lab
2:18 um Kyle Smith another Kyle in the house
2:23 looks like the numbers are not updating
2:25 again how many how many folks we have in
2:27 here now I'm showing 10.
2:29 and it looks like there's more than that
2:31 that if I'm seeing names pop up
2:39 nice likes oh yeah there's definitely a
2:42 day 132. holy crap oh yeah now it says
2:45 137 okay uh welcome everybody my name is
2:48 Kyle Shannon this is the AI learning lab
2:52 um Let me let me flip over so so here's
2:54 the deal I'm an entrepreneur uh my
2:57 background is I've started 12 or 15
2:59 companies
3:01 um one of the companies I started in the
3:03 mid 90s was called agency.com it was one
3:05 of the first digital agencies so I got
3:07 to experience the early days of the
3:09 World Wide Web
3:11 and got to ride that
3:14 um rocket ship up and down and through
3:17 through the.com Bubble Burst and uh
3:21 ended up selling that company to Omnicom
3:23 in 2002 but this generative AI stuff
3:26 this this Chachi PT thing
3:28 what is this chat GPT thing
3:31 uh it is very very similar uh to the
3:35 um to the early days of the web and uh I
3:39 think it's going to be significantly
3:41 more profound and I think it is coming
3:43 much much quicker than the web came the
3:46 web took about six years to get to 100
3:49 million users chat GPT took about six
3:51 weeks so this channel is all about
3:53 talking about kind of post chat GPT
3:56 inflection point sort of you know for
3:58 the last nine months we've been dealing
4:00 with this new generative AI thing what
4:02 is it you know what's it mean uh that's
4:04 what this channel is all about so if you
4:06 have questions about AI pop them in the
4:07 comments below I'll do my best to answer
4:09 them uh I swear occasionally so if you
4:12 don't like that get the [ __ ] out
4:16 hahaha
4:18 uh what else I tend to get a little
4:20 smarmy as the night goes on I do try to
4:22 answer all the questions so fair warning
4:25 I will get behind on the questions you
4:27 can make money with Chachi mate thank
4:29 you Chad
4:30 [Laughter]
4:32 um
4:33 yeah I don't like those people that say
4:35 that but uh it's been become kind of a
4:38 funny meme in the channel uh uh
4:41 I love you Kyle does anybody else love
4:44 Kyle thank you Matthew I appreciate that
4:49 you chase them away we're up to 170 now
4:53 we're down to 140. so I don't think they
4:56 do I think they're like who's this old
4:58 guy what the [ __ ] going on with his
4:59 hair why did he swear at us all right
5:01 let's get started here oh let me let me
5:03 uh tell you what the okay
5:06 so this used to be three sites now it's
5:09 five and it's probably gonna be seven at
5:11 some point and then I'm at some point
5:13 I'm gonna have to narrow it down but
5:14 basically
5:16 these are what I'm considering the major
5:18 players in the chat GPT game
5:20 um there's another one out there called
5:21 Pi Pi dot AI that is at least in the
5:25 neighborhood of a major player they just
5:26 raised 1.3 billion dollars so let's call
5:29 them a major player
5:32 um but their their model is is not not
5:34 quite as capable as is what a lot of
5:36 these are doing
5:38 um it's just different it's a bit more
5:40 conversational so chat.openai that's the
5:43 official chat EBT website if you haven't
5:44 played with chat GPT just go there go
5:46 there right now while you listen to this
5:48 old man ramble on about this [ __ ] and
5:51 start playing ask it to write you a poem
5:53 ask it to write you a song ask it to
5:54 write you a python application just ask
5:56 have it uh create a meal plan plan for
6:00 your family for the next week and then
6:01 turn that meal plan into a shopping list
6:03 of ingredients and then have it sort
6:05 those ingredients by section of the
6:07 grocery store do anything just go do
6:10 something
6:11 um the the thing down below here
6:13 prompts.chat that's uh a document
6:15 that'll talk to you about how to talk to
6:18 these tools these these are all sort of
6:20 different flavors of the same thing
6:23 um bing.com is also chat GPT it's gpt4
6:26 it's free so if you go to bing.com click
6:29 on the chat button that's gpt4
6:33 you can generate images there you can uh
6:36 read images there like you can upload an
6:38 image to it and tell it to describe it
6:40 to you or write advertising copy about
6:43 that image and it will do it uh it's
6:46 free it's connected to the internet it's
6:48 quite remarkable what you get to do for
6:50 free there uh gpt4 at open AI sites 20
6:54 bucks then another free one this one's
6:56 brand new this is Claude 2
7:00 um and it's anthropics competitor to
7:03 gpt4 it's got a hundred thousand token
7:06 context window meaning you can upload
7:08 about I don't know somewhere between 50
7:09 and 75 000 words into your prompt into
7:12 your chat so you could take your great
7:15 American novel pop it in there and start
7:17 editing it with that thing uh or
7:20 whatever it's amazing and then uh
7:23 poe.com has a bunch of different models
7:25 and bard.google.com is getting an
7:28 honorable mention right now it wasn't on
7:30 the list for a long time because it
7:32 sucks so bad but it's gotten better
7:34 um and and it's they're on their way
7:37 it's it's still not great but it's
7:40 they're on their way
7:41 so that's that and then if you really
7:44 want specific tools like something for
7:46 music or video or things like that
7:48 futurepedia.io is a big list of AI tools
7:51 so with that let's get to the questions
7:54 let's go to the videotape
8:02 I'm a physician I did AI research 30
8:05 years ago who's advancing the medical
8:07 front
8:08 um I think lots of people are advancing
8:10 the medical front I mean the the thing
8:13 about
8:14 about the AI World AI ai's been around
8:17 for like you say decades the the thing
8:20 that's different now is
8:23 um since the Advent of the Transformer
8:24 model that Google introduced in 2017
8:27 that's the thing that's been driving all
8:30 these
8:31 um
8:32 innovations that we're playing with
8:34 today so what the Transformer allowed uh
8:38 researchers to do is train these large
8:40 language models on massive data sets so
8:42 they trained them on you know kind of
8:44 all of the internet
8:46 um and so these tools are just in
8:48 incredibly powerful you don't have to
8:50 bring your own data you can you know
8:52 augment your data with these large
8:53 language models
8:55 um so I think there's lots of people
8:56 working on medical you know the people
8:58 that have been doing it before but now
8:59 you're going to have a whole rash of
9:01 people doing things on top of it Google
9:04 just announced three days ago a thing
9:06 called Med Palm let me get make sure I
9:09 got the name right
9:10 which is a large language model that's
9:13 been trained up on medical data
9:15 specifically
9:17 I think it's Google Google Med Palm
9:23 yeah Med Dash Palm so if you go to
9:27 sites.research.google slash medpalm
9:35 a large language model from Google
9:37 research designed for the medical domain
9:41 and
9:47 mm-hmm
9:48 [Music]
9:50 she's trying to look at when it came out
9:53 that was May 11th they had a they had a
9:55 new announcement
9:57 this week oh medpom too here we go
10:02 no that's April 13th
10:07 anyway go look up medpom
10:10 there's here's July 8th okay
10:13 this says this is from The Verge
10:15 Google's medical AI chat bot is already
10:18 being tested in hospitals the Mayo
10:20 Clinic has reportedly been testing the
10:22 system since April okay so it came out
10:25 in April and it looks like you know it's
10:28 it's in the world Mayo Clinic there you
10:31 go you know Clinic's doing it
10:39 hey man really important
10:42 I gotta stick
10:45 oh Joe Black you're cute
10:49 I'll work itself out it'll be tasty all
10:52 right please do it the daily summary at
10:55 the beginning middle and end of The Tick
10:58 Tock uh okay
11:00 the I would I didn't I was busy doing a
11:04 bunch of other stuff today so I didn't
11:06 do a ton of research on I didn't see any
11:09 major new things it was a big end of
11:11 last week so I don't have I don't have a
11:14 big update today on on news so
11:19 I mean I can talk about what I talked
11:20 about last week but I don't even have
11:23 the URL still up um
11:25 [Music]
11:34 [Music]
11:38 no I don't I don't have the uh I don't
11:40 have the article still up
11:42 so I'll get back to him
11:44 um Leonardo came out with a new model
11:46 check out leonardo.ai if you're not on
11:49 it
11:50 um get on it get on the wait list and
11:52 then apparently with leonardo.ai if you
11:57 get on the wait list and go to their
11:59 Facebook Facebook group and sign up for
12:02 their Facebook group they'll immediately
12:04 Grant you access uh to leonardo.ai
12:08 um but it's it's a very very good site
12:10 uh that is built on top of stable
12:12 diffusion
12:14 um and it's easier than mid-journey you
12:16 don't have to go to Discord it's just a
12:18 website uh and it's quite good and they
12:21 give you
12:22 150 credits a day and it and it's about
12:28 18 or 20 credits to produce four Images
12:32 so you know you can do it for free you
12:34 can do a handful of images and then if
12:36 you pay for it you can do lots
12:38 Lots
12:44 [Music]
12:48 I don't like Bard at all I hated Bard
12:51 for a while it's at least tolerable now
12:53 and it's you there's some interesting
12:55 things you can do there
12:57 um you can it can read images just like
12:59 bin Ken
13:01 um one of the guys in the AI Salon that
13:03 I that I run
13:05 um uploaded a sketch of a website like
13:07 they did in the in the open AI demo when
13:11 they when it when they announced gpt4
13:14 he uploaded a sketch of a website and
13:17 said build me a website you know that
13:19 looks like this and it did and one of
13:22 the guys in the salon did that with
13:23 bard.com so he uploaded a sketch of a
13:26 website and it actually wrote the code
13:29 um to make functional buttons on a
13:30 website
13:31 the T-shirt today is just some sort of
13:33 album
13:40 [Music]
13:47 a lot of butter jokes going on in the
13:49 comments here
13:50 I approve
13:51 [Laughter]
13:56 is somebody eating popcorn
14:00 is joining anybody figure out a way
14:02 around the internet being disabled
14:09 I think that might be a local issue I
14:11 think we'd heard about that on the news
14:14 let me see uh well let's wait well glass
14:19 Bing Let's go ask Bing I'm gonna try a
14:22 new thing today too
14:24 um
14:25 being
14:27 go to Bing chat
14:31 and
14:35 [Music]
14:38 has the internet been disabled
14:43 [Music]
14:44 all right Bing's gonna tell us
14:47 and I'm in more precise mode so it
14:49 definitely won't hallucinate
14:53 I'm sorry but I don't have any
14:55 information about the internet being
14:56 disabled globally it's possible that
14:59 there may be local issues
15:01 but all right we're okay it's okay Bing
15:05 told us the internet has not been shut
15:07 down we're good we're good for at least
15:09 an hour what are you talking about Moody
15:12 I have I have no reference for what
15:15 you're talking about x dot AI uh what
15:17 about it it's they made some
15:18 announcements it's that's Elon musk's
15:20 new competitor yeah they'll go up there
15:22 when they have something to play with
15:23 the reason I have these there these are
15:25 all things that I think anyone who
15:26 hasn't played with you know these tools
15:29 you know might want to experiment or
15:31 start with and there's a bunch of free
15:32 options there
15:34 claude.ai is is a definite true
15:37 competitor to chat GPT it's a bunch of X
15:40 openai people
15:41 um and that that site's actually quite
15:43 powerful the the language stuff isn't
15:46 quite up to the
15:48 the
15:50 um
15:50 the quality of gpt4 but it's it's quite
15:53 good it's quite powerful
15:56 and the fact that you can you know have
15:58 you have a a hundred thousand token
16:00 context Windows huge
16:03 uh what is your go-to prompt for getting
16:05 the AI to act like an expert in the
16:07 specific field so the reason I have go
16:10 to prompts.chat
16:12 that little second URL from the bottom
16:15 prompts.chat go there
16:17 that is a whole list of exactly what
16:20 you're asking for
16:22 act like a an accountant act like a
16:25 screenwriter act like a comedian act
16:26 like an idiot
16:29 um
16:32 I don't you don't have to get super
16:34 fancy with those generally what you want
16:36 to do is you want to give the thing a
16:38 roll so I want you to act like this kind
16:41 of role which kind of Narrows the
16:43 context a little bit
16:45 and then I'm going to be giving you
16:47 you know
16:49 some things to accomplish or you know
16:51 here's the things I want you to do and
16:54 then you give it the specifics and that
16:57 set of three contacts who it is you know
16:59 the role it plays
17:02 um
17:03 what is what its General tasks are the
17:06 milestones and then what what the
17:08 specifics are and then you get into
17:10 specific back and forth prompting
17:13 um the other thing you can do is very
17:14 often if you
17:16 if you prompt it and just say I want you
17:18 to act like this that and the other
17:20 thing and just hit return
17:22 it'll just start talking to you so one
17:24 of the things you can do is you can give
17:25 it these little contexts and say I don't
17:28 want you to answer yet I just want you
17:29 to let me know if you understand and
17:31 it'll say yes I understand and then you
17:33 say okay here's another piece of
17:34 information you know
17:37 you know I don't want you to do anything
17:38 yet I just want to make sure you
17:40 understand so you can give it a set of
17:41 instructions and make sure that it
17:43 understands all of those and then start
17:45 to get into asking it questions doing
17:48 things like that so that's that's kind
17:51 of how I approach it the other the big
17:53 shift for me with large language model
17:56 prompting versus
17:58 Google searches
18:00 because they look similar but they're
18:02 not
18:04 Google wants you to put as concise a
18:07 prompt in there as possible right I just
18:09 want this thing and they're going to
18:11 give you back that answer
18:12 these large language models they need a
18:14 lot more context because because
18:17 they're the the way they the way they do
18:19 what they do they're pulling from a
18:21 massive Data Corpus
18:23 um not that Google's isn't massive but
18:25 what it what it's returning is a list of
18:27 links uh which is you know from an index
18:29 so it's a very different thing than
18:31 here's all of the words in all of the
18:34 languages in all of their different
18:36 semantic meanings
18:39 and I'm going to start doing
18:41 probabilities on the words so the more
18:43 you the more context you give to these
18:46 tools the better they get
18:48 um and so that's something that I had to
18:50 relearn is that it's much more like a
18:52 conversation with a very simple person
18:55 that's a genius like this is these are
18:58 like idiot savants they need a lot of
19:00 Direction they need a lot of focus but
19:03 then they know everything so I'm gonna
19:04 put on the air
19:07 it's getting a little warm in here
19:11 it's getting a little warm in the
19:14 learning lab
19:21 um
19:21 [Music]
19:26 what up homie how are you I'm doing good
19:29 Doug Lee
19:30 what kind of interesting stuff are you
19:33 doing in here real quick math in 2022
19:37 IBM divested and spun off their Watson
19:40 Health division yep
19:42 why did IBM Watson fail I don't know the
19:45 I don't know probably I mean I I would
19:48 think that there's a couple of things
19:49 behind it Watson Watson was machine
19:52 learning technology that was
19:53 pre-transformer
19:55 um everyone that I know that that was
19:58 working in AI you know for the past
20:01 decade
20:02 working with
20:04 working with
20:06 old modalities basically
20:09 um are a little Blown Away by what these
20:11 tools are doing
20:12 so it it you know there might have just
20:15 been some of the tech might have not
20:17 quite been there it might have been data
20:19 set I don't know I don't know the
20:20 details of it
20:22 um
20:25 but
20:27 the the the healthcare space I do a lot
20:30 of work so my company's called storybine
20:32 we do a lot of work in healthcare and
20:33 Pharma and it's such a regulated
20:35 environment and hospitals in particular
20:38 are so resistant to change
20:42 um it takes it takes such a different
20:44 mentality to innovate in that space and
20:47 it's really hard and it's really slow so
20:49 it might have just been a thing where
20:51 they weren't getting the traction they
20:52 needed and they said hey we just don't
20:54 you know this isn't you know this isn't
20:57 delivering the level of results we want
20:59 so let's just spin it off I find it
21:01 ironic that what was it four months ago
21:03 they they announced Watson X which is
21:06 which is a large language model
21:09 it's a it's a rebranding of Watson it's
21:11 a reintroduction of of Watson going we
21:14 can do AI too right so Chachi PT the
21:18 reason we're seeing all these
21:19 competitors and the reason Elon Musk is
21:21 starting a new company and the reason
21:23 inflection just raised 1.3 billion
21:26 dollars all of this stuff is happening
21:28 because chechi PT went from zero to 100
21:30 million users in six weeks
21:33 um Sam Altman didn't expect that
21:34 Microsoft didn't expect that Google
21:36 certainly didn't [ __ ] expect that
21:39 um so so Chachi BT being such an
21:42 immediate hit changed everything so
21:46 um one of the things that I heard about
21:48 IBM which which is uh fascinating is
21:51 that they created an AI Center of
21:54 Excellence across IBM you know they have
21:56 150 or 200 000 employees they have a lot
21:59 of employees
22:00 um they have a thousand people at IBM
22:03 writing prompts for this stuff for this
22:06 new stuff
22:07 um so
22:09 everyone's everyone's head in this
22:11 direction why because these tools are so
22:13 powerful
22:16 um you're live awesome thank you very
22:18 much I do appreciate that
22:23 uh yeah if you're new here uh pop any
22:26 questions you have down below I'll do my
22:28 best to answer them
22:33 where do you find all these AIS I don't
22:36 really have a life
22:39 I'm running my company and I kind of
22:41 spend every waking moment these days
22:43 trying to figure this stuff out and it's
22:45 you know I'm not so so Matt wolf who
22:49 does futurepedia.io there's like 3 700
22:51 last time I looked it was 3 700 AI tools
22:54 there
22:55 um I've given up on following any of the
22:58 open source things I've given up on
23:01 following any of the Tactical specific
23:04 tools like you know here's the music
23:06 generator here's the other than Runway
23:08 ml I'm not really [ __ ] around too
23:10 much with video
23:12 um
23:14 um like all the specific tools you know
23:16 specific tools for writing copy or
23:18 writing doing real estate marketing or
23:20 building websites I'm just I just go to
23:23 futurepedia.io I'm not paying attention
23:25 to any of those
23:26 where I'm putting my attention is the
23:29 core large language models so the the
23:31 large companies building these large
23:32 language models that all these other
23:34 applications are going to be built on
23:36 because these are going to be the
23:38 engines that drive
23:40 everything else
23:42 and I just like it for me it feels very
23:46 important to understand the nuances
23:48 between these what the different things
23:50 are good at
23:52 um and and try to read the tea leaves
23:54 about what's coming and what are the
23:56 implications like
23:58 for example if you haven't played with
24:00 code interpreter at chat GPT oh sorry
24:03 just ate dinner
24:05 um if you haven't played with code
24:06 interpreter chatgpt
24:09 um
24:10 it it
24:12 is
24:13 mind-numbingly powerful it's it's
24:17 insanely powerful and I know nothing
24:18 about data analytics
24:20 but I can see that what that is is you
24:24 know a superpower right for anyone with
24:26 any kind of data which is essentially
24:28 any business these days right
24:30 um
24:31 so I'm spending my time trying to look
24:34 at what these tools are what they do
24:36 what makes them unique like Claude
24:39 having a hundred thousand word context
24:41 window to put that in context Chachi BT
24:45 currently is a 4 000 word context window
24:48 which means it's got like this 3 000
24:50 word sliding window
24:52 that is its memory for your chat session
24:54 so if you've ever worked with chat gbt
24:57 and all of a sudden it seems like it
24:59 forgot the early [ __ ] that you were
25:00 talking about
25:01 that's because it forgot the early [ __ ]
25:03 you were talking about because that
25:05 early [ __ ] popped out of that sliding
25:08 context window right
25:11 um Claude has a hundred thousand token
25:13 context windows so it's 75 000 words
25:17 right so it's a lot bigger than chat gbt
25:20 so even if the writing capability of
25:25 Claude
25:26 isn't as powerful as gpt4 the fact that
25:29 you've got this massive context window
25:31 is a big big deal and it's a reason to
25:33 use it
25:35 um and and just all that sort of stuff
25:38 you know barred from what I understand
25:39 is better at code
25:43 um chat GPT now has code interpreter
25:45 which is just insanity if you if you
25:49 watch it do what it does
25:53 I'll show you it's it's a it's an
25:56 amazing thing so anyway uh and I find
25:58 them on uh I just pay attention to
26:00 Twitter I follow people on Twitter
26:03 um
26:04 I find people like Andre carpathy
26:07 um who's the one of the founders of
26:09 openai and he he posts a lot I find
26:13 people like Ethan Malik who's A Wharton
26:15 Professor who's been doing amazing posts
26:18 about all this stuff and and so I go
26:20 look at who they follow on Twitter and I
26:22 follow them so I just start following
26:24 people who are who who the people that I
26:27 respect are following and within about
26:31 15 minutes you can have a pretty solid
26:34 list to follow on Twitter and then just
26:36 pay attention to it look at it two or
26:38 three times a day and go play with stuff
26:42 and start to get a critical eye like you
26:44 know understand the strengths and
26:45 weaknesses of these things one of the
26:47 things we did the other night on the
26:48 live someone asked can these things
26:50 write a joke and I was like yeah
26:52 probably not like it ain't gonna be
26:54 funny but let's go see so I came up with
26:56 a joke premise and we told it to act
26:58 like a stand-up comedian or comedy
27:00 writer and write a joke about this
27:02 premise about how avocados you know are
27:05 not ripe and suddenly they're overripe
27:07 right
27:09 um and we and we went to all these
27:11 different tools and had it try to write
27:12 the joke and they were all bad except
27:15 for gpt4 was the only one that actually
27:17 wrote a joke
27:18 I mean Bing wrote a joke but it was as
27:21 if Microsoft wrote it it was really bad
27:25 um I was referring to chat gbt Moody
27:29 where was that before what were you
27:31 saying before
27:35 just joining anybody figure out oh oh
27:37 about chat GPT being dumb was it down
27:40 today code interpreter was down
27:42 yesterday or the day before
27:45 but that's just because they're fixing
27:47 stuff here I am a chat gbt oh yeah I'm
27:50 gonna instead of typing instead instead
27:52 of uh forcing you to watch me type which
27:54 I know is painful I'm starting to use my
27:56 microphone on my Mac
28:00 write me a poem that explains the
28:03 concepts of quantum mechanics like an
28:05 Eminem rap
28:06 [Music]
28:11 it's going
28:12 yo let's dive into the deep end my
28:15 friend quantum mechanics ain't no child
28:17 pretend
28:19 all right it's working
28:21 it's going it's doing its thing it was
28:24 probably just a hiccup
28:25 there I know they're they're
28:27 working on a lot of stuff over at open
28:30 AI right now I I have this strong strong
28:33 suspicion that there's a lot of uh
28:36 legal conversations going on over there
28:39 they cut a deal with the Associated
28:41 Press on Friday and they cut a deal with
28:44 Shutterstock on Thursday and they pulled
28:47 down browse with Bing the connection to
28:50 the internet last week uh because openai
28:54 was stealing content from behind
28:55 paywalls and I think they got a call
28:58 from time Inc that said hey
29:00 you're uh
29:02 you're sharing our content that people
29:04 should be paying for
29:06 and that's bad
29:08 so I I just I think there's a lot of
29:11 there's a lot of uh
29:13 holes being plugged right now is my
29:15 sense
29:18 but the good news I mean the the good
29:20 news about things like them cutting a
29:23 deal with the Associated Press means
29:25 that we're going to start to see
29:28 specific flavors of things like chat GPT
29:31 like imagine
29:33 a mode in chat GPT where you can
29:36 restrict
29:37 the the sort of world of information
29:39 it's looking at to the Associated Press
29:43 so I want to I want to learn about
29:45 Middle East peace but I only want it to
29:47 pull from Associated Press you know
29:50 documents
29:52 um we'll start to see things like that
29:55 um and I I think that's gonna that's
29:57 gonna make things really really
29:58 interesting
30:08 is Fran Drescher your president
30:11 um no I have a degree in acting and I've
30:12 written seven screenplays but
30:15 um I understand
30:18 why they're striking
30:21 I empathize with it
30:24 um
30:25 I think some of the AI issues are they
30:29 can fight it but they can only fight it
30:31 for so long I think it's inevitable that
30:34 what these tools are going to do to the
30:36 creative arts to writing to acting to
30:39 voice over
30:41 um
30:42 to post-production to pre-production
30:46 uh to marketing to thought leadership to
30:51 Consulting to everything
30:55 um are are going to be incredibly
30:57 disruptive and and ultimately Innovative
31:00 but
31:02 almost every business model is is gonna
31:05 is gonna be forced through through
31:09 um
31:10 a new Gauntlet of innovation so just
31:14 like when when
31:15 um when the streaming company started
31:17 streaming content and weren't paying for
31:20 any of it and then they cut deals with
31:22 the studios and did all that sort of
31:24 stuff and they were shitty deals but
31:26 like they're not going back because
31:27 you're not going to uninvent streaming
31:29 same same with this stuff so while I
31:32 empathize with what they're fighting for
31:34 and they should absolutely fight to get
31:35 as much as they can
31:37 um
31:39 you know like the the one thing that the
31:41 studio is saying you know we're gonna do
31:43 something groundbreaking with AI we'll
31:44 pay people for one day scan their
31:47 likeness and then use that likeness for
31:49 all of eternity and never have to pay
31:51 them again isn't that awesome and the
31:53 actors are like [ __ ] no
31:56 but what the studios are gonna do is go
31:59 okay if you don't want us to pay you for
32:01 a day then we'll just synthesize them
32:02 entirely like we don't need you like
32:04 this we're good we'll just go make it
32:05 mid Journey so so I don't know I don't
32:09 know what's gonna happen it's crazy
32:10 traffic
32:13 a lot of people need to learn this isn't
32:16 wait wait what why do people need to
32:19 learn this isn't it pretty
32:22 self-sufficient
32:24 why do people need to learn this this
32:27 stuff
32:28 isn't it pretty self-sufficient well it
32:30 is and it's not if you've got a so if
32:33 you're if you're an engineer if you're a
32:35 program or a coder and you understand
32:37 how to think
32:39 you know in in logical expressions
32:42 that's what you're doing here you're
32:44 just doing it with the English language
32:45 or whatever your native language is
32:48 people that are either that are either
32:50 not writers or not used to thinking in
32:53 in logical expressions
32:55 don't know what to do with this stuff
32:57 right it's just it's just a search bar
32:59 right so you just type in you know
33:03 um what was the score of the you know
33:06 who who won Wimbledon and Chachi BT goes
33:09 my latest information is September 2021.
33:11 why do you ask
33:13 and like people are like well what is
33:15 this thing it's it's useless or they put
33:18 in you know
33:20 um do this math calculation and then it
33:22 does it wrong and they're like well this
33:24 is stupid it lied to me
33:26 you know write me an opinion piece on
33:29 this thing and give me the citations of
33:31 the Articles you used to come up with
33:33 that opinion and then the articles are
33:36 fake articles that don't actually exist
33:37 you're like well this is [ __ ]
33:41 so
33:42 so
33:45 people that have not been playing so so
33:47 here's the thing here's what this
33:48 channel is about I'm not a machine
33:50 learning guy I'm I'm an entrepreneur
33:52 I've started 12 or 15 companies I
33:56 um
33:57 my jam the thing that really lights me
34:00 up is once technology gets to the point
34:02 that the rest of us can use it that's
34:04 when I like to innovate
34:06 I don't like to create the technology so
34:09 if you've been in machine learning for
34:12 the past five years or 10 years or 20
34:15 years this [ __ ] seems trivial to you
34:18 but it's trivial in in a in an
34:21 interesting kind of way so in the early
34:22 days of the World Wide Web the internet
34:26 had been around for decades
34:29 and who was using it well it was mostly
34:31 researchers and scientists and computer
34:33 science people and programmers and
34:35 engineers and universities
34:38 and then Tim berners-lee invents this
34:40 thing called the World Wide Web where
34:42 you can just make a couple of tags and
34:44 you make this thing called a page and
34:45 it's got blue underlined words on it and
34:48 if you click on those they jump you to
34:49 some other page anywhere in the world
34:51 that was revolutionary because you just
34:54 had to do this
34:55 so all of a sudden a thing went from
34:57 command line you have to be a geek to
34:59 use it to I can use the internet click
35:02 click click click
35:05 and that transformed everything but it
35:07 wasn't until
35:09 it it made Escape philosophy escape
35:12 velocity beyond the Geeks and said hey
35:15 rest of the world there's this thing
35:17 called the internet you can click on
35:18 want to click on it
35:21 oh cool oh now I'm looking at something
35:24 that was written in Africa oh now I'm in
35:26 in China now I'm in Japan now I'm in
35:29 Texas oh that's cool this is exactly
35:32 like that it's been around for decades
35:35 machine learning neural networks AI all
35:38 this [ __ ] tensorflow Transformers
35:43 Gan networks it's been around
35:46 chat GPT
35:48 is this it's the equivalent of this was
35:51 you know World Wide Web to the internet
35:53 that's the difference and so why people
35:56 need to use it is they're like
35:58 what the [ __ ] is everyone talking about
36:00 why are they so excited about this
36:03 and and you know what I the the thing
36:06 the thing that really gets me excited is
36:10 what is when I'm in the presence of
36:12 people that have what I call the Kevin
36:15 McAllister moment right from Home Alone
36:20 it could do that
36:22 and so that's why I put these up here so
36:24 if you haven't played with chat gbt if
36:26 you're just sitting on the sidelines and
36:27 heard that the robots are going to kill
36:28 us
36:29 and you've seen too many tick tocks of
36:31 people going you can make money
36:34 and they annoy the [ __ ] out of you they
36:36 should
36:37 it's those people are jerks
36:41 but it doesn't mean you shouldn't use
36:43 these tools because these are the most
36:44 powerful tools ever put in the hands of
36:48 mere mortals
36:50 and so go there and start playing
36:54 chat.openai.com is the official chat GPT
36:57 website prompts.chat this little bottom
36:59 one right here
37:01 is we'll teach you how to talk to these
37:03 things
37:04 give you a little primer and just well
37:07 if you know if
37:08 this sounds interesting to you and maybe
37:11 you do want to make money with chat GPT
37:13 you got to start somewhere here's the
37:15 deal you're not as behind as you think
37:17 you are
37:19 and it doesn't require nearly as much
37:21 technical expertise as you think it does
37:24 it really is just talking with words
37:28 and asking for remarkable [ __ ] and it
37:31 does it
37:33 um and there's a whole bunch of issues
37:35 around it that we can talk about but if
37:37 you have questions pop them down below
37:38 and I'll do what I can to answer them
37:40 why don't you stream
37:42 acting as an AI why don't you do a
37:45 stream acting as AI I might do that I
37:48 know there's a couple of tools that that
37:49 can do it live I know Tick Tock can do
37:51 it live
37:52 um
37:53 I've I've thought about that
37:57 for me right now
38:00 I don't know a thing a thing that
38:02 actually is interesting to me is
38:07 is doing some sort of fully automated
38:11 persona
38:13 that
38:14 grabs news does education does something
38:18 does entertainment
38:20 um
38:22 but I haven't quite been inspired what
38:23 to do yet because it's the one of the
38:26 problems with this Tech is it's very
38:28 easy to get seduced by the tech and
38:30 think that the tech is the strategy the
38:32 tech isn't the strategy these are just
38:34 tools right so so when someone really
38:37 does ask the question well how do you
38:39 make money with AI
38:40 it's the same answer as how do you make
38:43 money with Photoshop I don't know you
38:45 [ __ ] solve a problem
38:47 good enough that someone's willing to
38:49 pay you money for it that like that's
38:51 all of business is someone has a problem
38:54 you use some tools to solve it or you're
38:57 this tool right to solve it and then
39:00 they pay you money
39:01 that's how this works
39:04 you use this and you came up with a
39:06 clever t-shirt and you did print on
39:08 demand print on demand is the tool you
39:10 know this is the copywriter
39:13 and then you stuck it on Tick Tock and
39:15 said look at my cool t-shirt well now
39:17 you can do that with AI and you can make
39:18 100 T-shirts instead of one okay that's
39:21 how you make money with chat CBT you
39:23 know but but that's not making money
39:25 with AI
39:26 that's making money making people laugh
39:28 and having a cool t-shirt that they want
39:30 to wear right
39:32 so so when I make the money when I make
39:34 the t-shirt for you can make money with
39:36 Chachi vitae
39:41 it doesn't matter like you know it's
39:44 about AI but it ain't right
39:46 all right
39:48 um where are you reading these questions
39:50 well class X here's the deal
39:53 I read all the questions
39:55 and
39:56 um it annoys the [ __ ] out of the younger
39:58 people because they're like because I'm
40:00 not reading the questions live at the
40:02 bottom I'm just sort of reading down
40:04 through them and so I just I just get
40:06 behind so I am live
40:07 and class X if you're still here I'm
40:10 reading your question now so I'm reading
40:12 them up where they turned gray
40:15 Google just added a great feature for
40:18 searches too it now has generative
40:20 searches so it will give you the
40:23 specific answer really is that it barred
40:31 um
40:32 let's see
40:37 um
40:56 hello
40:58 [Music]
40:59 uh
41:04 hello
41:07 hello
41:09 [Music]
41:10 um
41:12 what are the top
41:16 business
41:18 um
41:20 canvas
41:22 books in the past
41:26 five years
41:31 take a look at that
41:35 oh yeah that's interesting
41:41 so it looks like what it did
41:46 I asked that what are the best business
41:48 canvas books of the past five years
41:50 business model generation book it put an
41:53 image in there of it
41:55 value proposition design oh this is
41:57 really good
41:58 yeah so this is starting to get
42:00 multimodal so they're incorporating
42:02 search results into and then oh that's
42:04 nice you've got Google it down here
42:06 right so I can now Google it let's see
42:09 what that does
42:13 so that pops me over to a normal Google
42:15 search that's smart that's a that's a
42:18 really smart approach right so I can
42:20 have my kind of Standalone chat GPT or
42:23 this is more like Bing Bing like
42:25 experience
42:27 where where Google is now better than
42:29 Bing is Bing does lots of citations but
42:32 they just do it as like little you know
42:33 numbered citations
42:36 um here they're actually putting book
42:38 covers up here
42:40 um let me see let me see
42:41 um who let's see who
42:45 are
42:47 the top five uh
42:51 business cartoonists
42:55 in the past decade let's see if it
42:58 brings up any cartoons from the
43:01 cartoonists
43:03 let's see oh it does look at that Tim
43:06 Fishburn
43:08 Stephanie Pyro
43:11 Dave ramund
43:14 Mike Luke
43:15 Vic and Jim Boardman with examples
43:20 huh
43:22 give me
43:24 10 more
43:26 yeah this is impressive
43:28 I told you guys I said so so for months
43:31 Google has not been on my list of you
43:33 know things you should go play with
43:35 because for months Bard sucked really
43:38 bad but one of the things I said was
43:40 don't count Google out because they got
43:43 buckets of money they invented this [ __ ]
43:45 and they got a bunch of really smart
43:47 people there
43:51 yeah this is really really clever and
43:53 then what happened are any of these
43:55 linked
43:58 oh that's interesting this is linked to
44:00 just a New York New Yorker article ah
44:04 that's fascinating that's really cool
44:07 all right
44:08 Bard's getting there Bard hallucinates a
44:11 lot though that that I do know right now
44:13 so if you don't know what hallucinating
44:15 is that is when it lies to you when it
44:18 lies to you a lot like a dog it lies
44:21 like a dog
44:26 very cool thank you for sharing that
44:28 Tyler that's uh that's pretty neat yeah
44:31 that's that's new like that wasn't it
44:33 didn't do that yesterday it might not
44:36 have even done that this morning maybe
44:38 it was this morning and that's that's
44:40 the other thing about this technology
44:41 like it's being developed in real time
44:44 like we're getting access to these tools
44:46 and they're they're literally like the
44:48 engineers I can't imagine what it's like
44:50 to work in these companies they've got
44:51 to be you know coding 18 20 hour top
44:55 hours a day
44:57 um because this is this is insane it's
45:00 major major stuff going down
45:06 um how do you know if your AI answer is
45:09 violating copyrights um for the most
45:13 part they're not because
45:15 the nature of how the tools work is
45:17 they're writing your answer word by word
45:20 and it's just a probability engine now
45:23 if you tell it this is the the the
45:26 essence of Sarah Silverman's lawsuit is
45:30 if you ask for a specific quote I forget
45:33 what she did I think she might have
45:34 started the beginning of of a paragraph
45:36 in her book
45:38 and it completed it and completed the
45:40 rest of it
45:42 so in that case it's pulling content
45:44 from something it was trained on but
45:46 most of the time if you say I want you
45:48 to write me you know uh an opinion piece
45:52 on the value of generative AI for
45:55 writing marketing copy
45:58 it's it's going out you know across all
46:01 of what it's learned and it's just
46:02 probable probabilistically putting one
46:06 word after another
46:08 that has the highest probability of
46:10 answering your questions it's not
46:11 actually copying and pasting answers
46:14 from different things right so if you
46:17 purposefully try to get it to quote
46:20 something if it has that thing it will
46:23 do it very often it will tell you it's
46:25 giving a quote but it's lying it's not
46:27 even the real quote so that's something
46:29 you know where it might look like it's
46:32 something that's copy written but it's
46:33 actually not
46:35 um
46:37 so so most of it is is original content
46:41 that's been trained on all this all you
46:44 know all this other stuff
46:46 I wouldn't worry too too much about it
46:49 because these tools are going to be
46:51 rolled into like like chat gbt is
46:54 getting rolled into all of office 365.
46:56 Bard is getting rolled into all of
46:59 uh Google's office docs you know docs
47:01 and sheets and all that sort of stuff so
47:03 these tools are going to be ubiquitous
47:05 they're going to be spread throughout it
47:07 and it's you know it's not this similar
47:09 the the way these things create what
47:12 they create you know when you're when
47:14 you're typing up an email in in Gmail
47:17 and you write a sentence in like the
47:19 three gray words come at the end of it
47:21 it does that auto complete thing at the
47:23 end that's literally what these things
47:26 are just at a massive scale where it can
47:28 write you a whole paper instead of just
47:30 completing three words
47:33 um but it is generating original content
47:36 that's why it's called
47:37 uh the G in GPT is generative so it's
47:41 generating
47:43 original content based on pre-trained
47:45 information that's the p pre-trained
47:48 using Transformer technology which is
47:51 basically that's the model that they use
47:53 to be able to train it on the entire
47:55 internet
47:57 so
47:59 um but if you're doing this for a you
48:02 know for your company I would get
48:05 your lawyers to give you an opinion on
48:08 where the boundaries are
48:11 because I ain't a lawyer
48:15 um but I do know how they work they're
48:16 not copying and pasting copywritten
48:18 material they've been trained on
48:20 copyright material and lots of
48:23 non-copyright material and lots of blog
48:25 posts and Wikipedia articles and
48:28 websites and idiots talking on Reddit
48:31 you know it's all the [ __ ] right is
48:35 what's in the Corpus and then depending
48:37 on your prompt how much context you give
48:40 it you kind of narrow it down into
48:42 specific semantic Gardens
48:46 and then you know the more specific you
48:48 get it you probably run a higher risk
48:50 especially if you use someone's name
48:52 you know finish this paragraph from
48:55 Sarah Silverman's book called this
48:58 and it will do that because it's
49:00 somewhere in the Corpus you've narrowed
49:02 it down now to a point that it's it's
49:03 got a high likelihood of giving you
49:05 something that's copy written
49:11 I'm disappointed with tonight's shirt
49:13 Choice well at least you can read the
49:15 letters like like I I don't disagree
49:18 like all my good shirts are hard to read
49:20 text again so so
49:23 I mean what about the hair the hair it's
49:26 it's a problem never mind it's a we've
49:30 got issues we've got we've got fashion
49:32 issues we've got a coughing situation
49:38 Kyle did you pay to use chat gbt yeah I
49:41 pay for chat EBT and I pay for Poe and
49:44 I'm gonna I'm gonna unsubscribe to Poe
49:47 because it no longer has
49:51 the advantages it used to because Claude
49:53 came out
49:54 so the reason I paid for Poe is that you
49:56 could play with Claude 100k
49:59 um context window and now that's
50:01 available at claude.ai
50:05 thank you hair growth geek for the for
50:07 the rose I appreciate it I appreciate it
50:11 I appreciate it amazing but a little bit
50:13 scary yes totally agree wait till I show
50:15 you uh wait till I show you a code
50:19 interpreter if you haven't seen code
50:20 interpreter
50:22 who just mentioned Quantum Leap meta
50:24 Victory must be another gen xer
50:26 Quantum Leap references out of the out
50:30 of the archives
50:31 [Laughter]
50:42 hey my cuz O'Donnell that's a Rosie
50:45 O'Donnell reference meta victory
50:48 meta Victory saying I look like Rosie
50:50 O'Donnell
50:53 I'm getting Rosie O'Donnell the guy from
50:55 what is that Family Modern Family
50:59 uh skinny John Goodman fat Alec Baldwin
51:04 uh
51:07 uh what's his name
51:08 uh I don't know it doesn't matter
51:16 screenwriters fighting AI is a losing
51:18 battle I do fear so I've written a bunch
51:21 of screenplays so I feel for them but I
51:24 also I mean here's the thing that's kind
51:26 of [ __ ] up
51:28 this is the most powerful writing tool
51:32 in the history of humanity
51:34 as a writer as a marketer as a
51:38 screenwriter why would I not use the
51:40 most powerful writing tool in the
51:43 history of humanity
51:46 I remember I remember wanting to write a
51:49 screenplay
51:51 um about
51:53 um oh I forget her name it was Mary
51:55 something she was the first female
51:57 pirate they ended up making a movie
51:59 about it with
52:00 Gina Davis
52:03 um and like a decade before that I had
52:06 an idea to write a screenplay about this
52:08 this female swashbuckling pirate that
52:10 pretended she was a man
52:12 um
52:14 and it was from a like with ADD I'm
52:19 horrible at research right
52:20 so I went to the library and I got books
52:24 and I read them and I took notes and
52:26 that and it was just it was such a for
52:29 me it was such a burden to do the
52:32 research that I ended up not I didn't
52:34 even really finish the outline because I
52:36 realized it was just to do it well I
52:38 would have to basically be a dramaturg
52:40 and I just don't have the brain for that
52:42 well with these tools now
52:45 I could do all that research in an
52:47 instant right and and you know I might
52:49 need to go check is the is the research
52:51 that it did accurate right but now that
52:53 we've got things like Bard uh and Bing
52:56 that are connected to the internet that
52:59 I can check those citations and I can
53:02 just say hey tell me a tell me five good
53:04 stories about that particular character
53:06 in history oh that's good tell me more
53:09 about that one tell me about who else
53:10 was on the ship tell me about where they
53:12 were going it can just give me all that
53:14 stuff it can just feed me so if I'm if
53:16 I'm a screenwriter and I want to quickly
53:18 Knock Out Concepts why would I not use
53:21 this right and that's not what they're
53:23 fighting for what they're fighting for
53:24 is hey Studios don't use this to write a
53:28 shitty screenplay instead of using me to
53:30 do one
53:31 and if the studios are smart they won't
53:36 what the studio should do is go you know
53:39 give us more screenplays faster so we
53:41 have more options
53:43 feel free to use these tools
53:46 and give you know give me a bunch of
53:48 different options because that's what
53:50 exact film Executives love to do is they
53:52 love to get in there and act like
53:53 they're creatives and muck with the [ __ ]
53:57 well use the most powerful writing tool
54:00 in the history of humanity to help you
54:02 knock some [ __ ] out
54:06 I mean we're telling stories here
54:09 what are stories language what are these
54:12 things called large language models
54:15 large language models what are
54:18 screenplays language
54:20 what is business language what is
54:23 empathy language
54:27 this is like
54:28 these are Humanity amplifiers
54:32 they're amazing so yeah so I feel for
54:35 them and
54:39 it's yeah
54:42 here's a metaphor for you
54:46 the The Writer's Guild resisting catch
54:50 EBT
54:52 and the and the Actors Guild resisting
54:54 synthesized people
54:56 is the equivalent
54:58 of Florida declaring
55:02 legally with a proclamation that they're
55:04 no longer accepting hurricanes
55:08 I mean it's it's it's a noble law great
55:11 awesome congratulations by the way one's
55:15 coming
55:17 like you know this this is a force of
55:20 nature right this this can't be
55:22 uninvented
55:24 the fact that chat CPT got so popular so
55:26 quick and now literally trillions of
55:30 dollars of investment are heading that
55:32 way
55:34 it ain't stopping it's accelerating
55:37 right
55:38 so the storms the storms are Brewing
55:41 right the storms here
55:43 but it's like a tsunami where it's like
55:45 the you know the tide went out and it
55:47 went way out and it's like oh let's walk
55:49 out in the
55:52 let's let's walk out into Tampa Bay
55:54 because there's no water in it that
55:55 should be fun
55:57 um
55:59 so yeah yeah Florida is no longer
56:01 accepting hurricanes I I mean that's
56:03 that's the equivalent right now so
56:06 it's Noble and listen the The Writer's
56:08 Guild and the Screen Actors Guild should
56:10 get what they can get
56:13 um but
56:15 everything's changing everything
56:18 everything
56:21 the nature of intellectual property is
56:23 going to be fundamentally Rewritten
56:27 we're a few years from a world of
56:30 infinite personalized content generation
56:33 real time
56:35 personalized content generation
56:38 I like that Jay-Z song but I wish it
56:41 sounded more like Travis Tritt and
56:43 talked about me
56:45 here let me play that for you
56:51 that's coming eliminate hallucinations I
56:54 pulled email names from a scan but it
56:57 still had errors and no fillings yeah uh
56:59 so wait what's your name pickle
57:02 piccolochio all right eight out of ten
57:05 solid name I like it
57:07 um yeah the the hallucination thing is
57:10 just something we're gonna have to deal
57:11 with for a while so so any work that you
57:14 do with these if you're doing them for
57:15 work and where you know accuracy is
57:18 important check your work like your
57:20 homework now is not to not use these
57:22 it's use these and then make sure it
57:24 didn't lie
57:25 especially if you say a lawyer
57:28 can you help me run GPT for non-profit
57:31 experiment I've been trying to run yeah
57:33 tell me about it and I'll I'll maybe we
57:36 can do something live here Bing asked me
57:38 three times to be his girlfriend that's
57:40 really funny uh Rachel Studio what was
57:44 I'm curious what you were talking about
57:45 well that was the The New York Times uh
57:49 um reporter had a four hour conversation
57:52 with Bing it said its name was Sydney
57:54 and then it said it didn't just ask him
57:56 to be uh her boyfriend it asked him it
58:00 said he was in love with him and and he
58:02 should leave his wife
58:04 if for Sydney and then uh Microsoft
58:08 lobotomized old Sydney they shut that
58:11 [ __ ] down pretty quick once it made the
58:12 cover of the New York Times what do you
58:14 think is coming relative to AI how do
58:16 you see General businesses using them
58:20 um
58:23 [Music]
58:34 it's a great question
58:36 I don't have the answer but what I can
58:39 tell you is this what these things do
58:43 is
58:45 they can write
58:48 what a human would write about a hundred
58:51 times faster now
58:55 if you give it a really shitty prompt
58:57 it'll write about 40 percent as good a
59:00 hundred times faster
59:01 right it'll give you crappy stuff faster
59:05 if you get better at prompting
59:08 it'll give you like eighty percent of
59:11 what a human would written a hundred
59:13 times fat or it wouldn't be a hundred
59:16 times faster if you're giving it a good
59:18 well-constructed high context prompt
59:22 it can get to about eighty percent of
59:24 like what I would write as my my first
59:26 draft
59:27 in maybe 30 times faster not 100 times
59:30 faster but 30 times faster so what's
59:33 happening is businesses that are
59:35 aggressively leaning into using
59:37 generative AI are seeing 30 to 40
59:40 percent productivity increases
59:44 with an increase in quality right so 30
59:47 to 40 percent faster with an increase in
59:51 quality so it's as if they hired better
59:54 people better copywriters that might
59:56 charge more or and not just copywriters
59:59 like think any kind of knowledge work it
1:00:02 could be project management it could be
1:00:03 copywriting it could be ideation it
1:00:05 could be Consulting it could be talking
1:00:08 about HR issues it could be how do I
1:00:10 have a tough conversation one of the
1:00:12 things we've done on these lives is
1:00:14 um I want to quit my so someone wrote in
1:00:17 the comments
1:00:19 um
1:00:19 how do I press professionally tell my
1:00:22 you know current employers to go [ __ ]
1:00:23 themselves right and so we did a little
1:00:27 prompt for that and and came up with a
1:00:29 professional it was a really nice
1:00:30 professional letter that basically said
1:00:32 you don't listen to me and you know go
1:00:34 [ __ ] off
1:00:35 um so so all of that stuff
1:00:40 if if the business environment let's
1:00:43 just say just
1:00:45 um abstract it up completely
1:00:48 and let's take the the lower end of that
1:00:51 let's say there's let's say that on
1:00:53 average
1:00:54 companies that Embrace this stuff see a
1:00:57 20 increase in productivity with the
1:01:00 same staff
1:01:01 and a 10 increase in quality
1:01:08 either that company
1:01:11 can let go of 20 of the staff because
1:01:14 it's being 20 more productive some
1:01:17 companies will choose that right
1:01:18 especially the ones that are invested in
1:01:21 by private Equity firms that are just
1:01:23 like slash and burn it [ __ ] it let's
1:01:25 make more profit they'll just be like
1:01:27 get
1:01:28 the Smart Companies are going to say hey
1:01:32 we're seeing a 20 increase Here Without
1:01:34 Really having to do [ __ ] with our same
1:01:37 staff why don't we upskill our staff to
1:01:39 use even more of these tools and
1:01:41 reinvent some of our processes because
1:01:43 if we could take that from a 20 increase
1:01:45 to a 40 increase either we could take on
1:01:49 you know
1:01:51 two times or five times more clients or
1:01:54 we could lower our rates or we could
1:01:56 increase our value so that we retain our
1:01:59 clients longer term
1:02:01 so so at the sort of highest abstraction
1:02:04 level what what you what you move from
1:02:06 is you move from today where you've got
1:02:09 this many people producing this much
1:02:11 productivity
1:02:12 to the productivity starting to do this
1:02:16 and and and then and then what starts to
1:02:19 happen
1:02:20 so so here's where the disruptive [ __ ]
1:02:22 comes in
1:02:24 the the world that I know really well is
1:02:26 the agency world
1:02:28 so one of the ways that agencies make a
1:02:32 lot of their money right now and and the
1:02:34 holding companies that own you know lots
1:02:36 of the big agencies and lots of agencies
1:02:40 their model is based on time and
1:02:43 materials so a client hires them to do
1:02:46 write a bunch of social media content
1:02:48 come up with the strategy
1:02:50 do all the marketing for this particular
1:02:52 product
1:02:54 and they have a big Bullpen at the
1:02:57 agency full of copywriters and artists
1:03:02 and they come up with the strategy and
1:03:04 then they sort of dump the strategy into
1:03:07 the creative teams into the bullpen
1:03:10 and the more hours that those people
1:03:13 spend writing copy and making pictures
1:03:16 the more money the agency makes they're
1:03:18 charging 120 150 200 an hour for people
1:03:22 that they're paying
1:03:24 some amount less than that
1:03:28 so the agency
1:03:31 could take advantage of those
1:03:33 efficiencies but that's actually
1:03:35 contradictory to their business model
1:03:38 because in their business model the more
1:03:40 hours we spend the more money we make so
1:03:43 why would we want something that does
1:03:45 [ __ ] faster we wouldn't
1:03:48 but the clients
1:03:50 also are using this [ __ ]
1:03:54 so I've talked to two major
1:03:56 pharmaceutical clients in the past week
1:03:59 that have confided in me because I was
1:04:01 talking about this stuff have confided
1:04:04 in me
1:04:06 our agencies don't get it
1:04:10 we're asking them deliver faster and
1:04:14 charge us less money
1:04:17 and everyone we're talking to is like no
1:04:19 no we're good this is how we do business
1:04:25 so that's going to [ __ ] some [ __ ] up
1:04:28 because what's going to happen
1:04:30 the client's going to tolerate it for
1:04:32 what
1:04:33 maybe a year
1:04:35 till the next budget cycle goes around
1:04:37 or they're going to tolerate it until a
1:04:40 10 or 15 person agency that's all in on
1:04:43 AI
1:04:45 that can turn around in a week what an
1:04:48 agency can turn around in two months
1:04:52 they can turn around the same level of
1:04:53 stuff same level of quality
1:04:56 some aggressive AI agency is going to be
1:04:59 able to turn around in a week what a
1:05:01 traditional agency turns around in two
1:05:03 months at 10 times the cost because they
1:05:06 have no incentive to drop their cost
1:05:09 they want more hours
1:05:14 that's going to [ __ ] some [ __ ] up and
1:05:17 every single business category is going
1:05:20 to face when you have
1:05:24 a productivity
1:05:26 Spike like that there's it's just it's
1:05:28 gonna [ __ ] some [ __ ] up
1:05:31 so so in the end it's good it's we're
1:05:33 increasing Prosperity but
1:05:35 there's going to be jobs lost there's
1:05:37 going to be disruption there's going to
1:05:38 be business models they're going to be
1:05:40 completely blown up
1:05:43 sagin the wga the the Screen Actors
1:05:46 Guild and the writer's Guild right now
1:05:48 are basically saying
1:05:50 save the business model
1:05:54 the business model is over the business
1:05:57 models changed already
1:06:00 the contracts haven't caught up to it
1:06:01 yet
1:06:04 but it's this is this is a this is a
1:06:07 storm that's coming
1:06:10 so I don't know
1:06:14 it was a broad question to say you got a
1:06:16 broad answer sorry about that
1:06:18 uh seen you do it amazing uh Ethan Malik
1:06:22 is great love his post yeah he's awesome
1:06:24 um do all these companies own the usage
1:06:26 rights to AI responses yeah I'm pretty
1:06:29 sure they do at some point here's the
1:06:31 deal
1:06:32 okay you know who's happy right now do
1:06:35 you know who's pre-ordering their Yachts
1:06:36 the intellectual property lawyers are
1:06:40 [ __ ] ecstatic the lawyers are gonna
1:06:42 make a fortune off all this so
1:06:44 everyone's gonna Sue everyone
1:06:47 so yeah
1:06:49 Germany comedy that's funny
1:06:51 Yadi using the mic thank you Shelly
1:06:54 although I couldn't use it in bing for
1:06:56 some reason that was weird
1:07:00 was it a big oh no in Bard Bard I
1:07:02 couldn't use the mic
1:07:05 can I use it here
1:07:08 oh yeah there it goes
1:07:10 [Music]
1:07:12 yeah it's there okay
1:07:18 okay
1:07:24 how do you use your mic so you type in
1:07:27 your content been trying to figure out
1:07:29 how to do that I'm on a Mac so there's a
1:07:31 mic button F5 is a mic button
1:07:34 and I think you have to enable it and
1:07:37 you probably have to enable it for
1:07:39 the browser and then it looked like I
1:07:41 had to enable it there for that site and
1:07:43 so it
1:07:45 everything is more complicated than it
1:07:47 should be but
1:07:48 if you want to play with something cool
1:07:50 if you're on iOS for if you have an
1:07:52 iPhone
1:07:54 download the PI app pi
1:07:57 as in 3.1415
1:08:01 and it's got a mode where you can talk
1:08:03 to it so it's it's a it's like chat GPT
1:08:05 but you talk to it it talks back you
1:08:08 pick one of six voices
1:08:11 and then it's connected to the Internet
1:08:13 so I was asking it like what are the big
1:08:14 concerts in Denver this weekend it's
1:08:16 like well Taylor Swift is doing one and
1:08:19 she's on this big tour and people seem
1:08:21 to be loving your concert it's talking
1:08:23 back to me so that's a fun one but on
1:08:25 the Mac it's uh I forget what it's
1:08:28 called
1:08:29 the talking thing
1:08:38 have there been any reviews of the
1:08:40 Taylor Swift concert
1:08:42 in Denver
1:08:44 [Music]
1:08:48 I'm on bing.com so it's working for me
1:08:50 now so yeah that's better than me typing
1:08:52 that's less painful
1:08:55 let's there have been reviews
1:08:59 the concert took place on Friday
1:09:02 thousands of fans crowded the stadium I
1:09:05 drove by it it was insane
1:09:07 Swift's Talent discipline and cinematic
1:09:10 Poise was far more dazzling
1:09:13 then confetti Blaster fireworks could
1:09:16 ever be oh my God
1:09:19 this this AI writing is just it's a
1:09:21 little insufferable right now with
1:09:24 spectacle and emotion tightly braided
1:09:26 Swift Mega dose the faithful at the sold
1:09:28 out show with songs from across the
1:09:31 catalog while gracefully navigating
1:09:33 costume and set changes against a
1:09:35 massive LED screen would you like any
1:09:38 inform any more information I'd
1:09:41 like you
1:09:43 to be a better writer I could have said
1:09:47 that
1:09:50 responds to us a smarmy little troll
1:09:54 thanks for your feedback I'm always
1:09:56 working on improving my writing uh
1:10:01 all right
1:10:03 love the Bed Head yeah the glue the glue
1:10:06 broke it's yeah sorry about that
1:10:10 Skynet just everything screenwriters
1:10:12 fighting ai's a losing battle
1:10:14 Napster exactly this the shit's coming
1:10:17 out of touch to the max hi Kyle Ohio I
1:10:20 listen again I have deep empathy like
1:10:23 I've written screenplays I have a degree
1:10:25 in acting like you know I know people
1:10:28 that are you know deep in that world
1:10:33 so I feel bad I mean strikes strikes in
1:10:37 a business where I forget what the the
1:10:41 Actors Equity when I when I was acting
1:10:43 in New York the unemployment rate for
1:10:46 for Actors Equity which getting into
1:10:48 actors equities relatively hard right
1:10:50 you have to get a professional show to
1:10:51 be in it
1:10:53 um
1:10:54 had like a 96 unemployment rate like you
1:10:57 know the the the Arts is not you know
1:11:01 not everyone's Leonardo DiCaprio in
1:11:03 terms of their revenue
1:11:05 you know I think the average Screen
1:11:06 Actors Guild you know annual salary is
1:11:09 you know probably 20 grand or something
1:11:11 like that on the average Equity salaries
1:11:13 probably more like eight I don't know
1:11:15 what it is but actually let's ask let's
1:11:17 ask Bing that
1:11:20 okay I'll talk it I won't type it
1:11:28 can you tell me the average salary for
1:11:32 an Actors Equity actor comma a Screen
1:11:35 Actors Guild actor comma and a writer's
1:11:38 Guild screenwriter
1:11:41 [Music]
1:11:47 [Music]
1:11:49 and that's cool what it did was it took
1:11:51 my little prompt and it actually is
1:11:53 doing three different searches that's
1:11:54 kind of cool
1:11:55 this is Bing that I'm on right now
1:12:01 yeah annual average annual pay for
1:12:03 Actors Equity is 26 000 a year that's
1:12:06 for the ones that are working and it's
1:12:08 like you know
1:12:10 higher than 90 unemployment rate
1:12:14 um Screen Actors Guild 45 to 60 Grand a
1:12:16 year but most sag actors earn less than
1:12:19 five grand a year and most sag actors
1:12:22 are not working
1:12:25 and then the writer's Grill Guild spec
1:12:27 scripts range from 72 Grand to 136 Grand
1:12:30 the average around 100 000.
1:12:33 yeah I'm gonna say and what are
1:12:40 um the oh wait I forgot to talk it dang
1:12:43 it okay
1:12:45 and what are the employment rates
1:12:48 [Music]
1:12:55 uh
1:12:57 [Music]
1:13:09 employment of actors is expected to grow
1:13:12 eight percent
1:13:18 about seven thousand openings for actors
1:13:23 are projected each year so 7 000 jobs
1:13:27 for 70 million people that want to be a
1:13:30 star
1:13:33 it's it's a rough business so so listen
1:13:36 you know
1:13:37 I support them going for what they're
1:13:39 going to go for but
1:13:42 I don't know you just they're feeling
1:13:44 reactive and
1:13:46 doesn't feel good right now Hollywood
1:13:48 should everyone become an investor
1:13:51 on blockchain for movies
1:13:57 blockchain needs to get his [ __ ]
1:13:58 together
1:14:00 blockchain [ __ ] up
1:14:04 you know blockchain [ __ ] up black
1:14:06 blockchain needs to sit in the corner
1:14:08 for three years
1:14:09 but here's one blockchain's coming back
1:14:11 you ready
1:14:14 yeah I don't think it's a bad model to
1:14:17 invest in movies you know using that
1:14:19 kind of investment
1:14:21 um
1:14:24 nfts and blockchain and crypto I'll ask
1:14:27 to go sit in the corner because they
1:14:28 stole everyone's [ __ ] money because
1:14:30 they're a bunch of crypto Bros
1:14:32 and then
1:14:33 AI comes along it's going to be doing
1:14:35 its AI [ __ ]
1:14:37 and
1:14:39 the politicians and the consultants for
1:14:42 the politicians are going to start using
1:14:44 AI to do all sorts of misinformation and
1:14:47 deep fakes for the 2024 election and
1:14:50 then people are going to lose their
1:14:51 [ __ ] minds and go we've got to do
1:14:53 something about the misinformation this
1:14:56 is serious now and I mean it this time
1:15:00 and then the blockchain community is
1:15:02 going to go um we've got tokenization
1:15:04 that could help
1:15:05 what if we verified humans on the
1:15:08 blockchain and verified what they wrote
1:15:10 that it wasn't modified and it actually
1:15:13 was a New York Times article and those
1:15:15 sources actually were those sources
1:15:19 we could do that with nfts would you
1:15:21 like some we have all the technology
1:15:25 so
1:15:26 so I think AI is going to drive adoption
1:15:28 of all that [ __ ] I don't know what is
1:15:30 going on with my hair
1:15:35 O'Donnell
1:15:39 Rosie O'Donnell come
1:15:42 I want to control it not control us yep
1:15:44 I agree with that the fear that these
1:15:47 things are going to control us these are
1:15:48 our tools
1:15:50 and and one of the reasons that I love
1:15:52 pointing people in the direction of just
1:15:54 go start playing get curious start
1:15:57 exploring start thinking of this like
1:15:59 you're on a [ __ ] Adventure
1:16:01 is because if we don't it will be done
1:16:06 to us
1:16:07 right and not the robots will do it to
1:16:09 us other human beings using these tools
1:16:13 will do [ __ ] to us
1:16:16 so the more people that learn how to use
1:16:18 these tools and have that power we can
1:16:21 bend it to the light
1:16:24 bend it to the light bend it to the
1:16:26 light because if we just sit back and
1:16:28 wait and judge it and say it's not
1:16:30 perfect
1:16:31 and you know what I was saying is I went
1:16:33 on there and made a mistake I hear the
1:16:36 robots are going to kill us you know you
1:16:38 know what I told that AI why don't you
1:16:40 go [ __ ] yourself
1:16:41 if we're just doing that the whole time
1:16:43 then it's going to be done to us and
1:16:45 that's not a feeling I want to have
1:16:49 I use earbuds
1:16:51 I was on a live with a developer and he
1:16:53 used chat GPT to fix something on the
1:16:56 web 3 wallet or something yeah you can
1:16:58 do things like you can write smart
1:16:59 contracts it can analyze smart contracts
1:17:02 and tell you where their vulnerabilities
1:17:03 are
1:17:05 um yeah it's pretty slick and I would
1:17:07 imagine putting smart contracts in the
1:17:09 code interpreter where it's actually
1:17:10 writing python code and and doing things
1:17:13 like that there could be some really
1:17:14 interesting
1:17:16 things to play with there
1:17:18 do you think the best questionnaire
1:17:19 would make wait
1:17:21 would make the best of this
1:17:24 keep
1:17:27 do you think the best questionnaire
1:17:29 would
1:17:29 not a victory I don't know what that
1:17:32 sentence means
1:17:36 uh
1:17:37 is it missing punctuation do you think
1:17:39 the best questionnaire would make the
1:17:42 best
1:17:43 of this keep the best oh well I don't
1:17:47 understand the question so maybe
1:17:51 if I understood the question
1:17:53 smart to Smart to get ahead yep it's
1:17:56 important to get ahead if your talent is
1:17:58 limited to allowing the capturing
1:18:01 reflected light
1:18:03 they are trying to sell buggy whips yeah
1:18:05 K Sanders I I mean this thank you for
1:18:08 the game controller that's a new one I
1:18:10 haven't seen that before
1:18:11 um this is one of those things
1:18:15 that um
1:18:17 foreign
1:18:20 like I don't know what the future of
1:18:23 Cinema and entertainment is and it you
1:18:25 know part of it kind of looks like
1:18:27 whatever the [ __ ] the metaverse is going
1:18:29 to turn into like it isn't going to be
1:18:31 Mark Zuckerberg's metaverse because he
1:18:34 obviously [ __ ] that up
1:18:36 you know you spend all that money and he
1:18:38 made something that looked like second
1:18:39 life from 2001. I'm like
1:18:42 yeah maybe we could come up with
1:18:44 something that's [ __ ] not horrible
1:18:48 um
1:18:50 so
1:18:55 yeah the the only the the the place I
1:18:58 that's been spinning in my head for the
1:19:00 past month or so is just this phrase of
1:19:03 radical personalization that that
1:19:05 entertainment becomes radically
1:19:07 personalized to each of us individually
1:19:10 and so then what does that do for for
1:19:13 narrative where you've got a director
1:19:15 telling you know his or her story you
1:19:17 know in the following genre with the
1:19:20 following script and the following
1:19:22 actors
1:19:23 when you can take that basic thing and
1:19:25 you can swap all those elements out
1:19:28 and say I want to watch that movie but I
1:19:30 want different actors in it and I want
1:19:32 it in a different genre
1:19:34 and you can just watch that movie as
1:19:36 [ __ ] insane as that sounds and far
1:19:38 out as that sounds we're not that far
1:19:40 from that
1:19:42 so who become
1:19:44 who are the movie makers
1:19:47 that are crafting those personalization
1:19:49 engines
1:19:51 that understand narrative
1:19:53 and understand performance and nuance
1:19:56 and all that and find ways to capture
1:20:00 that stuff in a creatively
1:20:05 um
1:20:06 at a creatively high level right
1:20:09 like in order for that [ __ ] to be good
1:20:11 it can't be like a shitty College film
1:20:16 it's got to be like the Hollywood
1:20:18 Blockbusters we know and love right so
1:20:20 and maybe that's more like 20 years away
1:20:23 maybe maybe five years from now we have
1:20:26 the shitty version of this because it's
1:20:27 just technologists figuring it out
1:20:30 and then over the next 15 years after
1:20:32 that then the artists start
1:20:34 reintegrating into the technology and
1:20:36 you know who becomes the Steven
1:20:38 Spielberg of
1:20:41 the hyper personalized entertainment
1:20:44 thing
1:20:46 I don't know
1:20:47 I guarantee it's someone that's [ __ ]
1:20:49 around with this stuff right now
1:20:57 how are you innovating using AI right
1:21:00 now I'm using it
1:21:02 um I've generated some content
1:21:04 generators within my business
1:21:07 I've got a video technology business and
1:21:10 one of the things that I'm doing is I'm
1:21:11 I'm creating an omni Channel content
1:21:14 generation engine so for every video in
1:21:16 my system I'm generating content around
1:21:19 it like a whole world of content around
1:21:21 it I'm using it for
1:21:24 creative development with our clients
1:21:27 I'm using it for efficiencies in things
1:21:30 like writing emails and summarizing
1:21:33 meeting notes
1:21:36 um
1:21:39 I'm radically rethinking
1:21:42 so so my company's 11 years old it's
1:21:45 called storyvine and it's about human
1:21:47 beings telling stories right it's a
1:21:49 story-centric authentic storytelling
1:21:52 platform automated video editing and
1:21:56 it's about people telling stories
1:22:03 and I I feel like we're really good at
1:22:05 that like we understand it we understand
1:22:07 what it is
1:22:09 but one of the challenges we have is
1:22:12 getting people to tell their stories
1:22:13 well what if I didn't have to get people
1:22:15 to tell their stories
1:22:17 what if I could synthesize those but
1:22:19 what if I could do it in a way that that
1:22:20 it was more authentic than the other
1:22:22 [ __ ] that was out there
1:22:24 that's that's something that's maybe
1:22:26 interesting to me or what if I could
1:22:28 take
1:22:30 what if I said
1:22:33 hey would you be willing to share your
1:22:36 story sure because everyone says sure
1:22:38 but then when it comes time to really
1:22:39 record their story they're like well
1:22:41 does my hair look good oh my God my hair
1:22:43 is a disaster what do I do oh I don't I
1:22:46 don't think I talk good and I don't want
1:22:48 to do it so what if I just said
1:22:50 fill out this question there
1:22:52 and answer these give me some bullet
1:22:54 points of your story
1:22:56 and then
1:22:58 um just record a 30-second
1:23:01 a 15 second hi my name's Bob and I was
1:23:06 diagnosed with this disease in this year
1:23:09 and I take that 15 seconds of video I
1:23:11 synthesize Bob I synthesize Bob's voice
1:23:14 I take the bullet points of his story I
1:23:17 turn that into a compelling narrative
1:23:20 and then I make a video of Bob where Bob
1:23:22 didn't have to say [ __ ]
1:23:25 and it looks like a storybine video but
1:23:28 I don't have to you know break my back
1:23:31 or the client doesn't have to break
1:23:32 their back or the agency doesn't have to
1:23:34 break their back to get people to tell
1:23:35 their story
1:23:37 and then what if I could do that instead
1:23:39 of 10 stories with real people I could
1:23:43 generate a thousand stories that are
1:23:45 relevant to a thousand different
1:23:47 audiences and it gives the agency a
1:23:50 massive amount of organic content
1:23:56 so
1:23:58 yeah
1:24:01 and you know I've done some experiments
1:24:03 and and we're closer to that than
1:24:07 I thought we were
1:24:08 I'm pretty close to that
1:24:13 hyperlink was the leap yeah the
1:24:15 hyperlink was a big [ __ ] deal IRC
1:24:17 chat there's there's a there's a 1995er
1:24:21 watch the Quantum Leap
1:24:23 um I tried everything to get access to
1:24:25 dream boot is booth it wait dream Booth
1:24:30 is it really difficult to get up and
1:24:31 running yeah yes Robert Robert Smith 333
1:24:34 listen
1:24:36 so the the image right there that little
1:24:38 eyeball with the chin on it that's
1:24:40 that's one of my favorite dream Booth
1:24:42 images
1:24:43 I tried
1:24:46 nine times
1:24:48 to get stable diffusion and dream Booth
1:24:51 up and running and I I failed nine times
1:24:54 and this wasn't like nine times in one
1:24:56 night this was like over a series of
1:24:59 like a month and a half because I didn't
1:25:01 know what Google collab was I didn't
1:25:03 know how to run it
1:25:04 you know I would go on to Reddit and
1:25:06 they're like here's here's an automatic
1:25:08 11-11 you know Google collab notebook
1:25:11 and you just all you do is go here and
1:25:13 push the play buttons and I would go
1:25:14 there and I push the play buttons and
1:25:17 then one of them would give me an error
1:25:19 and I wouldn't I don't I I wouldn't know
1:25:21 what to and it took me nine tries until
1:25:25 I finally got it running and then I
1:25:26 finally got it running I was like yes
1:25:30 um so yes so so here's here's the deal
1:25:32 Robert
1:25:34 um
1:25:35 if you know Discord mid-journey is is
1:25:38 really high quality and and it's just
1:25:40 some relatively simple commands in
1:25:43 Discord to do it
1:25:44 um it's the best one from a quality
1:25:46 perspective stable diffusion is by far
1:25:49 the most um flexible because it's open
1:25:51 source and the develop the developer
1:25:53 Community is just going nuts on it so
1:25:56 there's all sorts of cool things it does
1:25:57 but the thing I would recommend if you
1:26:00 want to use stable diffusion is go get
1:26:02 on the wait list for
1:26:04 um
1:26:05 leonardo.ai
1:26:06 which is a web front end it's it's like
1:26:09 automatic 11 11 but you just go to
1:26:11 leonardo.ai you sign up for the wait
1:26:13 list and then you go to the Facebook
1:26:17 um group for leonardo.ai sign up for
1:26:21 their group and I think they fast
1:26:22 tracked you into the wait list it's
1:26:24 really good here I'll show it to you
1:26:30 so this is leonardo.ai so it's got
1:26:32 across the top here they just added a
1:26:34 bunch they just did an upgrade update on
1:26:36 this it's like version three
1:26:38 um so across the top these are different
1:26:40 models that you can use so like you know
1:26:43 this one's sort of like a cartoony kind
1:26:46 of thing this one is like ice and match
1:26:48 isometric sci-fi building so you pick
1:26:50 one of these and then you can
1:26:52 anything that you do within it is going
1:26:54 to kind of look look like whatever the
1:26:56 style of the model is
1:26:58 so there's some cool characters
1:27:00 there's a pixel art one
1:27:03 isometric fantasy illustration that
1:27:06 looks like someone did it in Adobe
1:27:08 Illustrator here's paper Art here's
1:27:10 vintage photography so it's got a bunch
1:27:11 of these selected models you can also
1:27:14 fine tune your own models so these are
1:27:20 um
1:27:21 like this is one I did on my photography
1:27:23 I I've done some you know landscape
1:27:26 photography with dramatic clouds out
1:27:28 here in Colorado and so I took a bunch
1:27:30 of my my best photos and I put it up
1:27:32 there and so now I can generate images
1:27:34 based on that so this is this is
1:27:36 basically dream Booth but without dream
1:27:38 Booth you just upload some images
1:27:41 um it does in painting and out painting
1:27:43 it's doing texture generation for 3D
1:27:47 um it's got a Community Feed of you know
1:27:50 here's all the images that people have
1:27:51 created and like any one of these you
1:27:54 can click on see what prompt they used
1:27:56 you can either copy The Prompt some of
1:27:58 them you can just go use the model that
1:27:59 they used
1:28:01 so it's really really good
1:28:06 so I would go play with leonardo.ai
1:28:10 and I'm not being paid by leonardo.ai so
1:28:13 I don't give two shits if you go use
1:28:15 it but it's a hell of a lot easier to
1:28:17 install than stable diffusion automatic
1:28:19 1111.
1:28:22 um
1:28:24 [Music]
1:28:30 but thank you for sharing that you were
1:28:32 having difficulty with it because it
1:28:34 like stable diffusion is powerful but
1:28:36 man is it a [ __ ] show if you're not a
1:28:37 technologist I mean this this is one of
1:28:40 the one of the frustrations with all new
1:28:42 technologies is that especially if
1:28:44 you're a creative and you have a
1:28:46 creative mind
1:28:49 like
1:28:51 to what you know to what level of
1:28:54 technical proficiency do you need to get
1:28:56 to be able to be creative right
1:28:59 and if you if you're not good with
1:29:01 virtual machines in Google collab
1:29:04 because you haven't been programming for
1:29:06 the past five years
1:29:09 then it's a [ __ ] pain in the ass
1:29:15 all right
1:29:18 amazing and a bit scary yes you make it
1:29:21 even more exciting I don't know what it
1:29:22 is but thank you yes I've been
1:29:24 researching that I'm just kidding you
1:29:26 have a great sense of humor Kyle thank
1:29:27 you Joy he does appreciate that
1:29:30 he was kind of funny but looks ain't
1:29:32 everything
1:29:33 [Laughter]
1:29:36 Hayden seemed to take a joke
1:29:38 he can see him take a joke yeah you know
1:29:40 we pick on him he calls it trolling I
1:29:43 don't know what trolling is like the
1:29:44 troll troll under the bridge it's like
1:29:46 troll under the bridge he's like that
1:29:48 but he doesn't he doesn't he doesn't
1:29:50 seem to get too upset about it but I
1:29:52 I'll tell you what I bet he does I bet
1:29:54 he hangs up from these lives he probably
1:29:55 just goes and cries because that's
1:29:56 that's what I would do I would just go
1:29:58 and cry
1:29:59 they talk you know what they said they
1:30:00 said you look like Rosie O'Donnell I I
1:30:02 just I think that's rude I think it's
1:30:04 rude he thought it was funny
1:30:07 all right
1:30:10 very true merely Immortals 80 20. don't
1:30:12 know what that means but okay yes 80 20.
1:30:15 mere mortals we are
1:30:17 I had one downloaded
1:30:20 and it I guess went away so you
1:30:23 recommend one what do they cost Tam the
1:30:25 obsession by the way uh seven and a half
1:30:28 out of ten on the name I like it
1:30:30 but I don't know what your question
1:30:32 means
1:30:34 I had one downloaded and I guess it went
1:30:36 away what was the one you had downloaded
1:30:38 and which so I'll recommend something if
1:30:41 I know what you're talking about
1:30:44 do you mean one of these you don't have
1:30:47 to download any of these all these are
1:30:48 just websites you go to
1:30:50 and Claude is free Bing is free
1:30:54 chat gbt has a a free level Poe has a
1:30:58 free level part is free
1:31:01 so those are all free and you don't have
1:31:03 to download anything
1:31:04 Kyle did you pay to use chat EBT yes I
1:31:07 pay for chat CBD could be largely
1:31:10 because of access to their new features
1:31:12 like code interpreter
1:31:15 how about building a virtual automation
1:31:17 bot on this channel okay we can do that
1:31:20 I'll do a zapier
1:31:22 it's the problem with doing a zapier on
1:31:24 this channel is it'll get really boring
1:31:25 really [ __ ] fast because zapier is
1:31:27 really janky but maybe what I can do is
1:31:29 I'll go show you one that I've already
1:31:30 built and I'll talk about kind of the
1:31:32 building blocks I'll talk about it
1:31:33 conceptually and and get you in there
1:31:36 um and if you if you want to see a bot
1:31:39 that I built go to content evolution.net
1:31:44 and there's a thing on the home page of
1:31:46 that site called The Challenge engine
1:31:48 and it's it's for business people it's
1:31:49 it's content evolution is a like a
1:31:51 Federation of marketing companies
1:31:54 and I'm the chief generative officer
1:31:57 there first Chief generative officer in
1:32:00 all of the world as far as I know huh
1:32:04 but
1:32:05 um
1:32:06 if you if you go play with the challenge
1:32:09 engine what it is is it's a Google form
1:32:12 where you put in your emails so we can
1:32:14 send you an email back you put in your
1:32:16 email your name
1:32:17 and a business problem you're having and
1:32:20 then you answer a few questions about it
1:32:21 like what does it cost you like what's
1:32:23 the impact of this problem and what
1:32:25 would your business be like if you
1:32:27 improved it all that sort of [ __ ]
1:32:30 and you answer those questions you hit
1:32:32 submit and then 30 seconds later you get
1:32:34 back essentially like a Consulting
1:32:36 report that says hey if you're trying to
1:32:37 solve this problem here's three
1:32:40 questions in six different categories
1:32:42 like strategic customer engagement
1:32:45 branding you know you know you should
1:32:48 you should know the answer to this
1:32:49 question that one now and so so we kind
1:32:51 of give back this Consulting report in
1:32:53 30 seconds based on your input so that's
1:32:55 a thing that I built so if you want to
1:32:57 go go play with that and see what it
1:32:59 does and then I'll I'll show you a
1:33:00 smaller version of that
1:33:03 um that I that I was I was working on
1:33:06 last week with a couple of guys
1:33:08 so we can do that
1:33:11 um how do you force AI to stop making up
1:33:13 info when automating a task like
1:33:15 transforming a table okay
1:33:18 so
1:33:23 large
1:33:25 largely you you you can't 100 but there
1:33:29 are things that you can do to make it
1:33:30 better
1:33:33 if you're using Bing Bing's got three
1:33:35 modes creative balanced and precise use
1:33:39 precise it'll hallucinate less it'll
1:33:41 make less [ __ ] up
1:33:44 um
1:33:46 the most important thing to do to reduce
1:33:48 hallucinations is to reduce your context
1:33:51 or or actually expand your context which
1:33:54 will reduce
1:33:56 the
1:34:00 areas of knowledge that it will draw
1:34:02 from so so here's here's
1:34:05 all right let me give you
1:34:07 a little bit of kind of large language
1:34:09 model 101 like like super high level
1:34:13 Theory
1:34:15 so the way these tools work
1:34:17 they use this technology called
1:34:19 Transformer technology and they can
1:34:21 basically dump massive amounts of data
1:34:24 into this machine
1:34:27 and so the massive amounts of data that
1:34:29 open AI decided to dump into this
1:34:32 machine is all of the public internet so
1:34:34 they spent years tagging finding data
1:34:38 sources making sure that they were in in
1:34:41 appropriate whatever structures or
1:34:43 formats to be able to be ingested by
1:34:45 these machines
1:34:47 so GPT 3.5 and gpt4 the underlying large
1:34:51 language model is essentially trained on
1:34:53 as much of the public internet as they
1:34:55 could get their hands on so it's a lot
1:34:57 it knows
1:34:59 everything
1:35:04 so here's kind of how it works the the
1:35:06 metaphor that I've come up with is this
1:35:08 imagine that you're in a helicopter tour
1:35:10 over the Grand Canyon
1:35:12 and the Grand Canyon is filled to the
1:35:14 rim with all of the words of the
1:35:17 internet and and every each little sub
1:35:20 Canyon is a little semantic grouping of
1:35:23 words meaning this
1:35:25 you might have the word dog well what
1:35:27 does the word dog mean well the dog is
1:35:29 the fluffy thing that you pet at night
1:35:31 and you know take out to pee and poo
1:35:34 well but a dog could also mean the guy
1:35:37 that did his girlfriend wrong he's a dog
1:35:39 oh that's a completely different meaning
1:35:41 so so when these machines go out and
1:35:44 look at all the words they're looking at
1:35:46 all the words in context of how they're
1:35:48 used so so there's these semantic clouds
1:35:50 basically so think of these little sub
1:35:53 Canyons
1:35:54 so if you type in a really generic
1:35:58 um
1:35:59 prompt like make me a table comparing
1:36:02 this concept and that concept your
1:36:05 helicopter is really high above the
1:36:06 Grand Canyon that can see all the words
1:36:08 right
1:36:10 and so
1:36:12 it can see all of the words that are
1:36:15 accurate about the concept that you want
1:36:17 to learn about and it can see all the
1:36:18 words that are inaccurate about it the
1:36:20 the example that Andre carpathy used
1:36:21 when he talked to the Microsoft
1:36:23 developers conference was he said
1:36:25 if you're asking chat GPT to solve a
1:36:28 mathematical equation
1:36:30 and and and you don't have the the
1:36:32 enough context to sort of limit those
1:36:36 semantic regions
1:36:37 he said you're kind of looking at the
1:36:39 entire Grand Canyon he didn't use Grand
1:36:41 Canyon he's like you're looking at the
1:36:42 entire Corpus and within that Corpus
1:36:45 you've got the professor who solved that
1:36:47 mathematical equation correctly you've
1:36:50 also got the 30 students who solved that
1:36:53 mathematical equation incorrectly So
1:36:56 when you say just solve this equation it
1:36:59 kind of you know pulls from all of the
1:37:02 places where someone's talking about
1:37:03 that problem it kind of melds them
1:37:05 together in some sort of probabilistic
1:37:07 average and just types one word after
1:37:09 another based on all of the Corpus that
1:37:12 it has and half of that [ __ ] or you know
1:37:14 ninety percent of that shit's wrong and
1:37:17 so that's going to hallucinate more
1:37:18 you're going to get more mistakes in
1:37:20 that but if you tell it to act like a
1:37:23 Stanford Professor that that specializes
1:37:25 in this branch of mathematics and you
1:37:28 know and is obsessed with with never
1:37:31 talking about equation unless it's right
1:37:33 and I'm and I'm going to tell you to
1:37:35 work on this equation and I want you to
1:37:38 show your process all of the stuff that
1:37:40 you do to give it more context what
1:37:43 you're doing is you're kind of lowering
1:37:45 that helicopter from seeing all of the
1:37:48 Grand Canyon down into one of those sub
1:37:50 canyons and potentially down into a
1:37:52 micro Canyon a little path where there's
1:37:55 just there's there's not that much for
1:37:58 it to pull from and it's got a higher
1:38:00 likelihood of pulling the right [ __ ]
1:38:03 and then still check your work
1:38:06 because here's the other thing that that
1:38:08 is not obvious about this stuff
1:38:11 because it's doing word by word
1:38:14 probabilities it's got the the
1:38:17 intellectual awareness of what it's
1:38:19 doing of a calculator
1:38:22 so if you say to a calculator hey
1:38:25 calculator
1:38:26 how do you know how to do math
1:38:29 that's ridiculous right
1:38:31 it doesn't it's just Pro like it's
1:38:33 circuits know how to do math right so
1:38:37 when you type in two plus two and hit
1:38:39 equals it makes it four and it just
1:38:41 knows how to do that
1:38:43 that's kind of what these large language
1:38:45 models are doing it's saying based on
1:38:46 the prompt you wrote the next most
1:38:48 probable word
1:38:51 is whatever word is at this mathematical
1:38:54 Vector space in this thousand
1:38:57 dimensional
1:38:59 word cloud
1:39:02 with all of the words in all of their
1:39:04 meanings
1:39:06 and it's whatever word is at Vector 342
1:39:09 comma six eight nine point three two
1:39:12 seven eight point four two
1:39:14 equals dog and then it types the word
1:39:17 dog it doesn't know it typed the word
1:39:19 dog it just knows that mathematically
1:39:21 the word dog is the one that's next and
1:39:23 then it does that for every single word
1:39:25 so what it presents to you looks like it
1:39:30 knows what it's talking about
1:39:32 but it it's not going back and looking
1:39:35 at what it wrote and going is this right
1:39:38 so so it literally doesn't know what it
1:39:41 wrote
1:39:42 which is counterintuitive because it's
1:39:44 these things are so sophisticated
1:39:47 that it's uncanny how much it seems like
1:39:50 well holy [ __ ] it knows what it wrote
1:39:53 right nope
1:39:56 oh that's it's doing copywritten stuff
1:39:58 no it's just it's written words
1:40:01 probabilistically
1:40:03 in a way that looks like what
1:40:05 Shakespeare would have wrote written
1:40:09 so I hope that helps
1:40:13 does your AI address you no you know who
1:40:15 addresses me Tick Tock because they're
1:40:17 stupid [ __ ] little letters I can't
1:40:18 put some sort of covering like like a
1:40:21 gray area underneath the letters so if
1:40:23 I'm wearing a shirt that has nice
1:40:26 patterns on it then I can't read the
1:40:28 words and everyone complains I can't
1:40:30 read it and we're all it all makes us
1:40:32 very fad
1:40:33 my wife does my dressing for me my my um
1:40:37 my fashion
1:40:39 um strategy is what's on top
1:40:43 like of the of the closet or of the door
1:40:46 of the drawer
1:40:48 just pull it out but now I have to pick
1:40:51 dark shirts so we can all read the the
1:40:53 little words on screen
1:40:56 um I haven't tested the Claude AI how
1:40:58 good is it compared to gpt4 it's it's
1:41:01 different the the writing of Claude is
1:41:03 not as good but it's but it's solid it's
1:41:06 it's good it's like GPD 3.5 good or
1:41:08 better
1:41:09 um it's not GPT for good but
1:41:12 you can upload up to five like PDFs and
1:41:16 interact with them it's got a hundred
1:41:18 thousand token context window so you
1:41:21 could literally copy and paste a 50 000
1:41:24 word novel into your prompt and start
1:41:26 interacting with it
1:41:28 I'm disappointed with tonight's shirt
1:41:30 Choice
1:41:31 ah I love my trolls you're the best
1:41:35 the first one is great don't know what
1:41:37 that means
1:41:40 Google search seems to not have
1:41:42 pagination anymore infinite scrolling oh
1:41:45 that's interesting
1:41:46 yeah Google listen man Google is uh
1:41:51 I can't even imagine the conversations
1:41:54 going on there
1:41:57 A disruption
1:42:01 you know they invented the thing
1:42:03 that allowed open AI to exist and then
1:42:06 they sat on it for seven years because
1:42:08 they're like you know
1:42:10 we probably shouldn't make a thing that
1:42:12 completely invalidates our entire
1:42:14 existence for being
1:42:21 so now they're trying to figure out okay
1:42:23 what is search
1:42:25 how do we make money
1:42:29 when our old model was predicated on
1:42:31 people clicking 40 times and now no one
1:42:33 has to click anymore because we give
1:42:34 them the answer
1:42:37 that's a that's a problem
1:42:40 and what's fascinating is Microsoft
1:42:42 doesn't have that problem because
1:42:44 they've got all their software
1:42:46 they've got their operating systems to
1:42:47 sell they've got Office 365 they've got
1:42:49 Dynamics they've got their cloud
1:42:51 services
1:42:53 so Microsoft has an entire Suite of
1:42:56 software applications
1:42:59 that they can just charge for chat gbt
1:43:02 Within
1:43:03 you know almost all of Google is
1:43:06 Advertising based on clicking
1:43:09 they want you clicking more
1:43:12 and they invented a thing that doesn't
1:43:13 make you click at all
1:43:21 got to be some fascinating conversations
1:43:24 going on
1:43:25 I'd like to be a fly in those beatings
1:43:29 actually I wouldn't
1:43:33 that old dog dog dog dog dog dog dog dog
1:43:35 dog
1:43:37 I would say most everyone could benefit
1:43:40 from code interpreter because of the
1:43:41 file use yep totally agree
1:43:44 code and interpreter super cool oh Tick
1:43:47 Tock just jumping to the bottom again
1:43:49 it's one of my pet peeves does anyone
1:43:51 here work at Tick Tock hey Tick Tock
1:43:53 you're listening to this right you're
1:43:54 transcribing this in real time
1:43:56 and doing a psychological evaluation of
1:43:59 me in real time yes and everyone in here
1:44:03 could you [ __ ] fix your scrolling bug
1:44:05 thanks that'd be awesome
1:44:10 so while I'm scrolling to figure out
1:44:12 where I was in my questions I like
1:44:14 answering all the questions and then if
1:44:17 I'm scrolling when someone makes a
1:44:19 comment it jumps me to the bottom and I
1:44:21 lose my place and then I have to go back
1:44:23 and it makes me cranky
1:44:25 so
1:44:27 and you're like why don't you just go to
1:44:29 the bottom I don't know
1:44:31 because I'm 58. and I'm cranky and I'm
1:44:36 bitter and I
1:44:38 you kids get off my lawn
1:44:42 you sons of [ __ ] don't you tell me
1:44:44 how to tick tock
1:44:46 Marge
1:44:48 I'm on a tickety-tock again they're
1:44:50 making fun of me
1:44:52 I don't look like Rosie O'Donnell do I
1:44:54 tell me I don't look like Rosie
1:44:56 O'Donnell
1:45:01 all right
1:45:03 this is like a Taylor Swift concert
1:45:05 except
1:45:10 the talent and the entertainment
1:45:17 slightly smaller crowd but
1:45:19 we have our fun
1:45:24 I feel like I've been here more time
1:45:29 [Laughter]
1:45:33 oh lordy lordy
1:45:36 I'm coming Martha
1:45:39 what's that from Sanford and Sons I'm
1:45:41 coming Martha you remember that God that
1:45:44 was a great show
1:45:45 all right does your AI address you I
1:45:47 found I found I got back to the troll I
1:45:48 was at before okay back to New questions
1:45:53 and I know I haven't started zapier yet
1:45:55 I'll show you zapier I gotta get in the
1:45:57 mood for it I gotta figure out how I
1:45:58 want to do it so it's not boring as [ __ ]
1:46:02 um they code with chat EPT
1:46:04 who comes with chat gbt
1:46:07 what are you talking about Big D
1:46:10 I would say most everyone could benefit
1:46:12 from code interpreter okay that's where
1:46:14 we were
1:46:15 um do you have to pay for a code
1:46:16 interpreter you do it's at
1:46:18 chat.openai.com you pay the 20 bucks a
1:46:20 month for GPT uh chat gbd Plus
1:46:23 and then you have to go into your
1:46:25 settings and turn on beta features and
1:46:27 you have plugins and you have code
1:46:28 interpreter turn those on and then you
1:46:30 have it and then just upload if you if
1:46:33 you if you're like me and and do not
1:46:36 have a history in data analytics which I
1:46:39 don't
1:46:40 um
1:46:41 go to kaggle
1:46:44 kagle.com and it's just a big pile of
1:46:47 data sources
1:46:49 and so you can go find like crime stats
1:46:52 for your city or
1:46:54 um IMDb TV show statistics for the past
1:46:57 decade
1:46:58 and then just download it and even if
1:47:01 it's in zip format this is a this is a
1:47:03 good thing to know about code
1:47:03 interpreter so you can upload one file
1:47:06 right so sometimes when you download
1:47:08 data sets sets they're in like a dozen
1:47:11 single CSV files so I was thinking like
1:47:14 oh I've got to find a way to get all
1:47:15 those csvs into a an Excel sheet with
1:47:18 multiple workbooks nope you can upload a
1:47:20 zip file to code interpreter so you
1:47:23 could have a zip file full of 100 files
1:47:25 so as long as it doesn't overload the I
1:47:27 think it's 100 megabyte uh maximum
1:47:31 so you can just upload an archive.zip up
1:47:34 to code interpreter it will write python
1:47:37 code to unzip it and then it will look
1:47:39 at what's inside it
1:47:41 and then it will go analyze all that
1:47:42 data it's it's just [ __ ] insane it's
1:47:45 outrageous
1:47:47 try identity code interpreter I have
1:47:49 I'll show it here in a second
1:47:51 I use chat GPT for schoolwork about to
1:47:53 finish my bachelor's awesome it's great
1:47:55 for rough drafts yeah so Brandon that's
1:47:57 what uh what I'm using it what what my
1:48:00 experience consistently is with these
1:48:02 things is I get to 80 like that I get to
1:48:05 80 in five minutes writer's block is a
1:48:08 thing of the past ideation is no longer
1:48:11 intimidating and like I like ideation
1:48:13 but you know sometimes if you're like I
1:48:16 gotta come up with six ideas for this oh
1:48:18 God I guess I gotta erase the [ __ ] off
1:48:21 the Whiteboard which means I got to take
1:48:23 a picture of their whiteboard which
1:48:24 means I gotta put it somewhere and I
1:48:26 can't ever find my whiteboard pictures
1:48:27 again so maybe I need to go get a
1:48:29 whiteboard picture organizer maybe I
1:48:31 need to find a thing that like you know
1:48:33 how that you ever had that conversation
1:48:35 anyone have ADD here
1:48:37 and now I don't have to do that now I'm
1:48:39 just like I gotta come up with an idea
1:48:40 oh yeah let me go to church EBT bang
1:48:42 eighty percent five minutes and then I'm
1:48:44 off to the races
1:48:46 um
1:48:47 so it is it's great for rough drafts
1:48:50 um
1:48:54 was shaking wait for your younger AI
1:48:57 coming soon yeah exactly I need it you
1:48:59 on fleek dude don't listen to the haters
1:49:02 thanks I assume you on fleek is a good
1:49:04 thing
1:49:05 but I'm an old dude so I think I've seen
1:49:08 that in one of those here's what the
1:49:10 young kids say
1:49:13 let's ask a boomer about about what the
1:49:17 Millennials are saying I think is on
1:49:19 fleek Millennials I think on fleek is uh
1:49:21 gen Z right huh
1:49:23 I am culturally relevant
1:49:27 so could I have an AI
1:49:30 I could have an AI continuous stories
1:49:33 from books for me
1:49:35 Ed would be upset I don't know who Ed is
1:49:39 and
1:49:41 was I mentioning Ed before
1:49:43 but yes you could have
1:49:47 you could have so there's a there's a
1:49:49 really cool project called
1:49:51 um
1:49:51 GPT author
1:49:54 and it's a it's a Google collab notebook
1:49:57 so you do have to use Google collab
1:50:00 um which which if you've tried to
1:50:02 install stable diffusion you'll be
1:50:04 familiar with
1:50:06 um but what GPT author is is you give it
1:50:11 your open AI key for the words and you
1:50:14 give it your stable diffusion key for
1:50:16 for cover art
1:50:18 and then it asks you a couple of
1:50:20 questions like what do you want the
1:50:22 topic of the book to be I think I think
1:50:23 it's set up to do fictions like fantasy
1:50:25 fiction
1:50:27 so you give it the topic and then you
1:50:29 give it the writing style
1:50:31 and you give it the number of chapters
1:50:33 and then you hit submit and it
1:50:35 starts writing a book it starts out
1:50:37 writing an outline
1:50:39 and then it writes chapter one and then
1:50:41 it writes chapter two and then it writes
1:50:42 chapter three and then it writes chapter
1:50:43 four and it does up to X number of
1:50:45 chapters and then based on the book it
1:50:48 wrote it generates cover art
1:50:51 and and turns it into to ebook format so
1:50:54 you sit there and watch it and it's also
1:50:57 cool it gives you the cost of each
1:50:59 chapter and it's like two cents for this
1:51:01 one four cents for that one it gets
1:51:03 increasingly expensive because I think
1:51:05 it's sort of building on it takes you
1:51:08 know the outline in chapter one and then
1:51:11 makes chapter two based on all that and
1:51:13 it takes all that makes chapter three so
1:51:15 it gets more expensive as it goes down
1:51:17 but it's like a dollar to write a book
1:51:18 so yeah you could write a book an hour
1:51:20 automatically if you wanted to
1:51:23 uh with the head of hair thank you for
1:51:25 all your input you're welcome code
1:51:27 interpreter is good for basing future
1:51:29 School responses based on zip previous
1:51:32 file work oh that's interesting yeah
1:51:35 yeah you take you take a bunch of
1:51:36 previous work that's one of the things
1:51:38 that these these tools are really good
1:51:40 at so so here's an early frustration
1:51:43 that I had and there's there's a way
1:51:44 around it
1:51:45 you tell it that you you want it to
1:51:48 write a poem and it writes you a shitty
1:51:50 poem you're like well no but I like
1:51:52 poems that are like this and then you're
1:51:54 like write a poem like T.S Elliott then
1:51:56 it sort of does that well but then you
1:51:57 want it to write poems like you write
1:51:59 poems but they're not like T.S Eliot and
1:52:01 it doesn't know who you are
1:52:03 well it's really good at like you can
1:52:05 take one of your poems and toss it in
1:52:06 there or ten of your palms and toss them
1:52:08 in there and say write me a new poem
1:52:11 about puppies in the rain
1:52:14 um in my style of the poems up above and
1:52:17 it's really good at that
1:52:19 so with code interpreter you can just
1:52:21 pile a bunch of documents like I was I
1:52:23 was actually thinking about this I've
1:52:24 got
1:52:25 seven feature screen plays a one-man
1:52:28 show
1:52:29 and probably
1:52:31 15 to 20
1:52:34 like like
1:52:35 partially completed screenplays or
1:52:37 outlines I could tell us all of that
1:52:40 into a zip file and put it in code
1:52:43 interpreter and basically start
1:52:45 interacting with all the [ __ ] I've
1:52:46 written you know
1:52:48 over the past couple of decades that's
1:52:51 kind of fun calling me out Gen X I know
1:52:53 I see you I see you
1:52:56 I know I know who's in here you look
1:52:59 like Sammy Boyle Jackson just talk like
1:53:02 him
1:53:03 [ __ ] snakes on this
1:53:05 [ __ ] plane I actually did
1:53:08 marketing for Snakes on a Plane
1:53:10 that was fun I was working for uh I was
1:53:12 working for a Hollywood marketer we did
1:53:15 snakes on the plane was one of our
1:53:16 projects it was good
1:53:18 I also got to do I did uh digital
1:53:21 photography for one of the Quentin
1:53:23 Tarantino films the uh hateful eight
1:53:26 and was down at Comic Con taking
1:53:27 pictures you know in The Press Room that
1:53:29 was fun
1:53:31 Kyle's approaching the end of our
1:53:32 questions ask more uh oh
1:53:34 I want to make a website to upload my
1:53:37 screen my screenplay wait I want to make
1:53:39 a website to upload my screenplay and
1:53:41 get coverage comps cast workflow
1:53:43 thoughts dude okay
1:53:45 toast post okay uh 8.5 out of 10 on the
1:53:49 name toast post
1:53:52 it's it's solid it's close to a nine
1:53:54 it's an 8.5
1:53:57 um go to claude.ai right now or or if
1:54:00 you if you're a GPT plus subscriber you
1:54:04 can do it there with code interpreter
1:54:05 but just go to claude.ai
1:54:08 you can upload files so just upload your
1:54:11 screenplay or copy and paste the text of
1:54:14 your screenplay paste it into the prompt
1:54:16 and then just say based on this
1:54:18 screenplay
1:54:20 write me a Hollywood log line you know
1:54:24 you know give give me you know uh act
1:54:28 act as a as a uh
1:54:31 um a Hollywood script reader right who
1:54:34 writes coverage for high-powered agents
1:54:37 so I want you to write write the uh I
1:54:39 want you to read the screenplay and
1:54:41 write write some high covered you know
1:54:42 some high scale coverage I want comps I
1:54:45 want cash breakdowns I want plot
1:54:48 breakdowns I want I mean you could
1:54:50 anything yeah go do it right now I did
1:54:52 it
1:54:53 um I did exactly that
1:54:55 three days ago it was it was [ __ ]
1:54:57 brilliant
1:54:59 there's a screenplay that I haven't
1:55:02 thought about in
1:55:05 25 or 30 years
1:55:07 and it gave me the synopsis of it it
1:55:10 gave me the character breakdowns and
1:55:13 what happened to them it was like [ __ ]
1:55:15 it was just it's it's amazing it's
1:55:18 it's just
1:55:21 you owe it to yourself to play with
1:55:23 these tools until you have this moment
1:55:28 and then once you have that moment you
1:55:30 can't go back you can unsee it
1:55:33 I'm gonna take the red pill are you
1:55:35 gonna take the blue pill
1:55:40 we ain't uninventing it
1:55:42 it's gonna [ __ ] things up for you
1:55:44 whether you're using it or not so you
1:55:47 might as well start using it
1:55:52 have you heard of AI cannibalism I read
1:55:54 an article about it today
1:55:58 um is that like the the model collapse
1:56:00 like if if you've got AI training on
1:56:02 synthetic data that they're going to
1:56:04 have model collapse that the AI
1:56:06 basically it all becomes
1:56:08 self-referential [ __ ]
1:56:10 um I that might be what you're talking
1:56:12 about but I I haven't heard that term AI
1:56:14 cannibalism but it sounds like that
1:56:16 model collapse kind of concept
1:56:18 use your personal expertise to generate
1:56:20 and know that you are creating quality
1:56:22 work with AI yeah yeah exactly the the
1:56:26 you're thinking about this right Brandon
1:56:29 all the stuff that you're saying tonight
1:56:30 feels very right to me like
1:56:33 understand these tools like the best
1:56:35 thing you can do right now in my opinion
1:56:37 in any profession no matter what you're
1:56:40 doing is learn what these things make
1:56:41 possible and start doing good work with
1:56:44 them
1:56:45 don't get precious about like I I was on
1:56:47 here one night I don't know a month or
1:56:49 so ago and someone was like I feel
1:56:52 really guilty I'm using mid-journey to
1:56:54 make patterns that I'm selling on Etsy
1:56:57 and I feel really guilty
1:56:59 I'm like why
1:57:02 if you're making patterns that you like
1:57:04 right you're taking the time to learn
1:57:06 how to use mid-journey well enough to
1:57:08 make patterns and you're probably making
1:57:10 a hundred patterns for anyone that
1:57:13 you're gonna sell you probably got a
1:57:15 hundred that you're not going to sell
1:57:16 right so you're taking the time to
1:57:18 generate a bunch of work pay for that
1:57:21 curate it make sure that it's truly
1:57:24 seamless
1:57:26 put a product page together sell it on
1:57:29 Etsy
1:57:30 why why feel guilty about doing
1:57:33 something that just
1:57:35 I mean the guilt is maybe I know that if
1:57:38 I were doing this with traditional tools
1:57:41 that would take me eight hours or four
1:57:44 hours per pattern and with this tool it
1:57:47 takes me 30 minutes
1:57:49 so
1:57:51 if it's good work
1:57:54 if it's stuff that you think adds value
1:57:56 to the world
1:57:57 who cares what tool you used
1:58:02 if you were to cut out patterns with
1:58:04 linoleum that would take you a week and
1:58:08 with Adobe Illustrator it takes you four
1:58:10 hours and with this it takes you 30
1:58:13 minutes who cares it's a tool
1:58:17 so yeah
1:58:20 figure out what you want to do figure
1:58:22 out what's the best tool to do it and
1:58:24 [ __ ] do it I just heard all the union
1:58:26 actors are mandated to third image and
1:58:29 then they're done
1:58:31 no that was what the what the what the
1:58:35 producers Guild what the what the
1:58:37 Hollywood producers basically
1:58:40 said to Sag is
1:58:43 we've got a groundbreaking proposal and
1:58:46 that's that's the words they used
1:58:49 and their groundbreaking proposal is for
1:58:51 every background actor which is
1:58:53 basically all the people that stand
1:58:54 behind the people that talk
1:58:56 you know like the people in the cafe
1:58:58 that are like pretending to talk and
1:59:00 occasionally they'll look at the camera
1:59:01 them
1:59:03 what the producers were saying is we
1:59:06 have a groundbreaking idea
1:59:08 we'll hire them for one day we'll pay
1:59:10 them for a full day and then what we'll
1:59:13 do is we'll just scan them in
1:59:15 and we'll never pay them again but we'll
1:59:17 use their likeness forever for any
1:59:21 Project without having to pay them
1:59:23 without having to tell them about it
1:59:24 awesome right
1:59:27 and that's when sag went go [ __ ]
1:59:29 yourself and went on strike
1:59:33 produce is not so smart
1:59:36 they could have positioned that a little
1:59:37 we Tad better because for them it's
1:59:40 groundbreaking
1:59:42 I mean the thing that's really [ __ ] up
1:59:44 about it is that
1:59:47 The Producers could say well listen
1:59:50 not for nothing we said we would pay
1:59:53 them for a day
1:59:55 but now we're just gonna go synthesize
1:59:58 them completely and we're not going to
1:59:59 pay anyone so we're just so background
2:00:01 actors are now just gonna be and and one
2:00:03 of the things that sag will probably
2:00:05 fight for and probably get some
2:00:07 some version of is
2:00:11 um there's got to be some percentage of
2:00:14 real background actors
2:00:17 in in a scene right so if you've got a
2:00:20 scene with you know 30 people behind the
2:00:22 actor maybe five of those have to be
2:00:24 real people and the rest can be
2:00:26 synthesized or something like that like
2:00:27 they'll probably get to some sort of
2:00:29 agreement like that but I don't [ __ ]
2:00:31 know I mean it's it's just nuts it
2:00:33 really is just nuts
2:00:36 um
2:00:38 you know any good stock to AI stock to
2:00:40 invest in no I'm I have a degree in
2:00:44 acting don't listen to me for for
2:00:45 investing advice I here's your Clues
2:00:49 every night I've been I've been doing
2:00:51 these lives for two or three hours every
2:00:52 night and I have I have a bunch of
2:00:54 companies listed here start there
2:00:56 and then look at all the companies
2:00:58 related to them
2:01:00 um yeah I I think anthropic claude.ai
2:01:04 they're they're very impressive to me
2:01:07 um I don't think they're investable
2:01:08 right now because they're not public but
2:01:11 um I think Microsoft is kicking ass I
2:01:14 think when when Microsoft rolls all this
2:01:15 [ __ ] into their into their entire
2:01:17 application Suite that's really
2:01:19 interesting Nvidia seems to have a near
2:01:22 Monopoly on gpus
2:01:26 um
2:01:27 open AI
2:01:29 inflection
2:01:31 um
2:01:33 Google's on their heels right now but
2:01:36 you know
2:01:37 they got a lot of money a lot of smart
2:01:39 people
2:01:40 interesting just saw Brian Tracy say
2:01:43 number one cause for failure is focus on
2:01:46 Tech versus customer Solutions so yeah
2:01:49 this is this is a lesson I learned over
2:01:51 and over and I and I this is a mistake I
2:01:54 see happening over and over and over
2:01:55 again every time a new technology comes
2:01:57 along it happens
2:02:00 um
2:02:01 when I started agency.com in the 90s in
2:02:04 the mid 90s
2:02:08 a huge part of the conversation
2:02:12 for like
2:02:14 three years
2:02:15 was
2:02:17 CEOs and you know big companies would
2:02:20 come to us and like we need a website
2:02:23 and what most people would say that
2:02:25 we're building websites back then you
2:02:28 can make money with the worldwide web
2:02:29 those people
2:02:32 is they'd say sure we'll build you a
2:02:33 website
2:02:35 website and then they build them a piece
2:02:36 of [ __ ] website we never said we'll
2:02:39 build you a website when they came to us
2:02:41 and said we want a website we would ask
2:02:43 why
2:02:44 why do you want a website well I was
2:02:47 golfing with my buddy and he's got one
2:02:49 okay that's a [ __ ] horrible reason
2:02:52 well but also we'd like to have better
2:02:54 customer service okay now we're getting
2:02:56 somewhere
2:02:57 you know what are you trying to
2:02:58 accomplish as a business
2:03:00 what do you need what do you need this
2:03:02 website to do
2:03:03 who are your customers what are your
2:03:05 business goals
2:03:06 does anything over here fit with
2:03:08 anything over there
2:03:12 you know
2:03:15 they weren't all like that some of them
2:03:17 were like we're gonna [ __ ] figure
2:03:19 this web thing out and we're gonna do it
2:03:22 world class like we we worked with
2:03:24 British Airways they were like that they
2:03:26 [ __ ] had their [ __ ] together
2:03:28 that was the toughest client I've ever
2:03:30 had they were they were insanely hard to
2:03:32 work for but fair
2:03:34 and they know what they [ __ ] wanted
2:03:36 so understand first what you're trying
2:03:39 to accomplish
2:03:41 and then you go which tool
2:03:43 right
2:03:44 you want images for social media
2:03:48 do you need AI or could you just go to
2:03:50 canva and pull up one of the one of
2:03:53 eight million [ __ ] templates that are
2:03:56 pre-designed and ready to go
2:03:58 and they've got their own stock
2:03:59 photography do you really need to go
2:04:01 generate something here for that no
2:04:05 if you just want
2:04:06 good high quality Instagram [ __ ] go to
2:04:10 canva
2:04:11 which by the way is all full of AI Nows
2:04:13 so you can automate the [ __ ] out of it
2:04:16 but but don't mistake
2:04:19 the tool
2:04:21 for strategy
2:04:23 this comes after strategy this comes
2:04:25 after why
2:04:29 what are you trying to do
2:04:32 and then
2:04:34 look at this and if this can help you do
2:04:37 it 30 times faster or 30 times cheaper
2:04:39 or 30 times better [ __ ] awesome
2:04:42 and that's why people are so excited
2:04:44 about it because when they actually
2:04:46 start to apply like the parlor tricks
2:04:48 are cool write me a country song write
2:04:50 me an Eminem rap make me a table write
2:04:52 me some code that's all fun and games
2:04:54 but when you go oh wait a minute I'm a
2:04:56 project manager let me see if this
2:04:58 thing's any good at project management
2:04:59 holy [ __ ]
2:05:03 that's the Kevin McAllister moment
2:05:08 that's when this stuff gets starts to
2:05:10 get powerful
2:05:11 I mean sure it's powerful to show your
2:05:14 friends but it's really powerful when
2:05:16 you say ha I wonder if if I have to
2:05:19 fundamentally rethink the value
2:05:21 proposition of my 11 year old business
2:05:25 that's what I've been doing for the past
2:05:27 three months
2:05:29 and it was four months before that I was
2:05:32 just trying to figure out
2:05:33 is this AI stuff at all related to what
2:05:36 we do and the more I learned about it
2:05:38 the more it became clear to me oh yeah
2:05:41 it's like existentially related to what
2:05:44 you do you better [ __ ] figure it out
2:05:48 so that's what I'm in the middle of
2:05:49 right now
2:05:50 stressful
2:05:52 and exciting
2:05:53 excited scared
2:05:56 The Genius of and you can have both
2:06:00 did you answer my Hollywood blockchain
2:06:03 question
2:06:04 everyone investor I did
2:06:06 I said that I think blockchain's got to
2:06:09 go sit in the corner for another two or
2:06:10 three years and then it's going to get
2:06:13 important
2:06:14 um
2:06:15 I
2:06:16 I think that um
2:06:20 funding films using blockchain is fine I
2:06:22 mean it's it's crowdsourced funding so
2:06:24 there's that's been around for a bunch
2:06:26 of years
2:06:27 um
2:06:29 you know
2:06:31 I I
2:06:34 the the crypto Bros [ __ ] things up for a
2:06:37 bunch of people for a bunch of years so
2:06:39 they just need to go sit in the corner
2:06:41 and and let let that [ __ ] calm down I
2:06:44 but I think it's a I think it's a fine
2:06:45 mechanism
2:06:47 um
2:06:48 there's a credibility problem there that
2:06:50 like the the Hollywood film investing
2:06:53 business is already a [ __ ] show so I
2:06:56 feel like blockchain and nfts complicate
2:06:59 the [ __ ] out of it and it's just it's a
2:07:01 it's I don't know yes it's possible yes
2:07:04 it's fine technology but they have they
2:07:07 have a PR problem right now
2:07:10 which is we [ __ ] stole everyone's
2:07:14 money
2:07:16 and now we're starting another company
2:07:18 and we just raised you know 15 million
2:07:21 dollars for that or 50 or 100 whatever
2:07:23 it was
2:07:26 um I use GPT to rewrite emails to sound
2:07:29 professional
2:07:30 and have received comments even I'm
2:07:32 telling you these these these tools are
2:07:34 going to help us be better writers more
2:07:36 empathetic all that sort of stuff
2:07:40 I'd be in on the non-profit also I'm not
2:07:44 sure which one I talked about for work
2:07:46 that is my first thought is how
2:07:49 dangerous it could be
2:07:50 um can be dangerous there's a thing
2:07:52 called alignment um go look at the
2:07:54 there's a blog post from openai I'm
2:07:57 talking about super alignment so open AI
2:07:59 the makers of chat gbt
2:08:02 um are dedicating 20 of their compute
2:08:05 resources I think it's over the next
2:08:07 five years to a thing called super
2:08:09 alignment
2:08:10 which is basically making sure that as
2:08:13 these things get more powerful
2:08:16 that they are aligned with our
2:08:19 imperatives
2:08:22 because if these have a mind of their
2:08:24 own like we want to be the most powerful
2:08:26 things in the world and at some point
2:08:28 they go these things are getting in our
2:08:32 way
2:08:32 that's where it gets dangerous
2:08:36 um but there's all sorts of other ways
2:08:37 it can be dangerous like the
2:08:38 misinformation in the 2024 election here
2:08:41 in the states is going to be a [ __ ] show
2:08:44 in the spreadsheet world this is about
2:08:46 to take those that do not know anything
2:08:48 to be somebody this will take those that
2:08:51 do not know anything to be somebody
2:08:53 um yeah exactly is Kyle Frozen am I
2:08:55 Frozen
2:08:58 um
2:09:01 yeah this is this is
2:09:04 this is like everyone gets to be a data
2:09:06 analyst everyone gets to be you know
2:09:09 everyone gets to be in some sense
2:09:11 whatever the [ __ ] they want to be
2:09:15 because I don't have the talent to be an
2:09:17 illustrator
2:09:18 but I can make illustrations that
2:09:21 accomplish my creative goals now I could
2:09:23 never do that before
2:09:26 I can write but I don't have but I'm
2:09:30 I'm not good at rewriting and I'm not
2:09:32 good at research well I am now
2:09:35 I sort of know coding
2:09:39 well now I can write full-on
2:09:40 applications
2:09:44 I don't know anything about investing oh
2:09:46 now I can create trading Bots that merge
2:09:49 six different philosophical approaches
2:09:51 to investing into a single trading bot
2:09:54 just for fun
2:09:58 you couldn't do that nights ago
2:10:02 I mean you could if you knew how to get
2:10:04 into the opening eye playground but like
2:10:07 chat GPD that's what these things have
2:10:08 opened up for everyone
2:10:11 what about where creativity is concerned
2:10:15 sag girl or essay girl
2:10:18 um
2:10:18 where creativity is concerned
2:10:20 everything's going to transform
2:10:22 everything
2:10:25 everything's going to transform there's
2:10:27 going to be nothing there's writing will
2:10:30 never be the same
2:10:32 illustration will never be the same
2:10:37 um
2:10:38 video animation 3D World building
2:10:41 character development 3D modeling
2:10:45 architecture none of it none of it will
2:10:48 be the same
2:10:50 now in a lot of ways it's going to be
2:10:52 better
2:10:53 like like the
2:10:56 if you just if you just erase the
2:10:58 transition so let's say there's a
2:11:00 transition from now to five years from
2:11:02 now if we just jump to five years from
2:11:04 now
2:11:06 we're gonna be in a creative Renaissance
2:11:08 the likes of which we can't even imagine
2:11:10 right now
2:11:11 architecture is going to be through the
2:11:13 roof fashion is going to be through the
2:11:15 roof writing music
2:11:18 film TV
2:11:21 gaming
2:11:23 is is going to be on a creative level we
2:11:25 can't even imagine right now
2:11:28 and how you produce all of that will
2:11:32 almost in no ways resemble how you
2:11:35 produce it today
2:11:37 so what that means is
2:11:41 we can't just jump to the future we just
2:11:43 can't jump to a new state where we know
2:11:45 what the new jobs are and we knew we
2:11:47 know what the new creativity is and we
2:11:49 know what the new Creative Expressions
2:11:51 are
2:11:52 so we have to go from
2:11:55 okay
2:11:57 musicians just barely survived the
2:12:01 transition
2:12:02 from from Studios to streaming for well
2:12:06 from Studios to digital to streaming
2:12:12 and now they're entering a world where
2:12:14 fan generated songs might be more
2:12:16 popular than the artist's songs
2:12:20 why because
2:12:22 we're in a new place where the rules are
2:12:24 different
2:12:26 so every single creative
2:12:29 discipline is going to go through some
2:12:31 sort of radical transformation where the
2:12:33 business model changes
2:12:35 the value of work changes what the work
2:12:38 actually is changes
2:12:42 the the creative and particular
2:12:47 is a really interesting one because
2:12:55 there's a guy I mentioned on the live a
2:12:56 lot guy named Paul rateser and he talks
2:12:58 about the value of writing in his life
2:13:01 he's a writer
2:13:03 and he says
2:13:06 for me writing helps me think
2:13:10 the act of writing an article about
2:13:12 something allows him to process the
2:13:14 ideas
2:13:16 and so for him it's the act of writing
2:13:18 is that's where the value is
2:13:21 but it's a certain kind of writing it's
2:13:24 the kind of writing on on topics that he
2:13:26 wants to learn something about he's
2:13:28 still writing by hand he's not using
2:13:30 chat EBT but all of the other writing
2:13:33 that he does
2:13:34 he's using Chachi BT
2:13:36 so I think every creative is going to
2:13:38 have to confront
2:13:43 do you get your personal enjoyment out
2:13:46 of the creative product or the creative
2:13:48 process
2:13:50 and if you get your enjoyment out of the
2:13:51 creative process you're probably not
2:13:53 going to use these things but you need
2:13:55 to understand that you're going to be
2:13:57 competing in a world where there's other
2:13:59 creatives that are more interested in
2:14:01 the creative product
2:14:03 and they're going to be able to make a
2:14:05 hundred of their creative product in the
2:14:07 time that you can do one of yours
2:14:09 because you're committed to the creative
2:14:11 process
2:14:13 right so if the value is in the process
2:14:16 keep doing it the way you're doing it
2:14:18 keep doing illustration by hand keep
2:14:20 drawing if you get your
2:14:23 the benefit for you as a human is is in
2:14:26 the act of drawing itself keep drawing
2:14:31 but if you're selling those
2:14:33 illustrations in a commercial setting
2:14:36 and you could knock out a hundred of
2:14:38 those in the time it could take you to
2:14:39 draw one
2:14:42 that's a choice you got to make
2:14:44 and there's going to be a lot of people
2:14:45 where that
2:14:47 compromise
2:14:50 is soul crushing
2:14:53 right
2:14:55 that's [ __ ] up
2:14:57 but everyone's gonna have to face it
2:14:59 because you're increasingly competing in
2:15:01 a world
2:15:03 where people that haven't been an
2:15:06 illustrator for 40 years are all of a
2:15:08 sudden able to illustrate like you
2:15:09 illustrate and the clients are just
2:15:11 going to say well I just need something
2:15:13 good enough for this ad
2:15:14 I mean I like you Fred but it takes you
2:15:16 a month to make a drawing
2:15:19 and I need 10 next week can you do that
2:15:22 no I enjoy the process sorry
2:15:25 this guy over here does something that's
2:15:27 close
2:15:28 and like you're gonna look at that guy's
2:15:30 AI generated work and you're gonna
2:15:31 [ __ ] hate it you'll be like that's
2:15:32 not anywhere near what I do it's not and
2:15:36 that client needs the 10 drawings next
2:15:39 week
2:15:41 that's you know what does that mean I
2:15:43 don't know
2:15:44 it ain't pretty
2:15:49 but it's exciting like go to
2:15:53 I had to leave Thelonious I had to leave
2:15:55 to unfreeze Kyle am I still frozen I
2:15:58 I've seen screen Taps you're still
2:16:00 hearing me right I'm not frozen
2:16:04 yeah tap the screen like make the hearts
2:16:06 go and yeah okay
2:16:09 I think yeah okay
2:16:11 our old-fashioned expert system still
2:16:13 around built with something like Pike I
2:16:15 don't thanks thanks hair growth
2:16:18 um
2:16:18 I don't I don't know Pike I don't know
2:16:20 the old expert systems but these things
2:16:22 are kind of meta expert systems so so
2:16:26 all sorts of expert systems like the one
2:16:29 that's the the pet peeve of mine it's
2:16:32 it's not really an expert system but
2:16:33 it's a very dedicated specialized kind
2:16:35 of system our phone trees
2:16:38 phone logic
2:16:41 thank you for calling atnt
2:16:45 press one
2:16:48 for the thing you're not interested in
2:16:50 if you're interested in this other thing
2:16:52 you're not interested in press two
2:16:55 for the thing you couldn't give a [ __ ]
2:16:57 about press three
2:17:00 for another thing that you hate and
2:17:02 don't even want to think about press
2:17:03 four
2:17:05 if you're a commercial vendor of a t
2:17:08 press five get to my thing
2:17:12 that's an expert system
2:17:15 every single one of those should be
2:17:17 [ __ ] turned off tomorrow
2:17:21 and replaced with a chatty little
2:17:24 synthesized voice that knows everything
2:17:26 that just answers the phone hey Kyle
2:17:29 how's it going last time we talked you
2:17:32 were having trouble with connectivity
2:17:34 with your Wi-Fi that all working okay no
2:17:37 it's [ __ ] up again oh I'm sorry to
2:17:39 hear that
2:17:41 have you tried resetting your router
2:17:43 yeah I've done all the stuff okay so it
2:17:46 sounds like you might want to talk to a
2:17:47 real person yeah that would be great
2:17:49 okay I'll synthesize all of what we've
2:17:52 talked about the last four calls and I'm
2:17:54 going to put you through to Sally now hi
2:17:56 this is Sally
2:18:01 so [ __ ] expert systems
2:18:04 if they're like if they're like phone
2:18:06 trees
2:18:08 um
2:18:11 and that's you know I
2:18:13 I think that people that have been in
2:18:15 machine learning and Ai and a lot of
2:18:18 this stuff expert systems for a long
2:18:20 time I think I think this Transit
2:18:22 transition is as profound for them as it
2:18:25 is for all of us on you know on the on
2:18:28 the non-technical side of this
2:18:31 um
2:18:32 because rather than I mean if you think
2:18:35 about the complexity of a phone tree I
2:18:37 mean
2:18:39 I looked at how to program a phone tree
2:18:42 for my you know seven person company
2:18:46 and
2:18:47 I I could not believe how complex it was
2:18:51 like I can't even imagine what it would
2:18:53 be like to program one of those things
2:18:55 to think about all the logic of all of
2:18:57 the different scenarios and all the
2:18:59 different levels of authority that
2:19:02 people have within an organization
2:19:05 all of that has to be
2:19:07 thought of and and
2:19:09 programmed in manually every logical
2:19:13 decision
2:19:17 the thousands and that probably hundreds
2:19:20 of thousands of decisions in that phone
2:19:22 tree
2:19:23 in traditional systems have to be
2:19:25 programmed in and Along Comes this and
2:19:28 you can just dump all of your corporate
2:19:30 data into
2:19:32 the customer service
2:19:36 reposit you know knowledge repository
2:19:39 and this thing will just start talking
2:19:41 it'll just start listening to what Kyle
2:19:43 has to say
2:19:46 if you've programmed it cleverly it'll
2:19:48 sound like Kyle or it'll sound like
2:19:50 someone in Kyle's neighborhood
2:19:54 it'll remember the last time you talked
2:19:57 because that goes into the repository
2:19:59 too
2:20:03 and it'll it'll have all the answers
2:20:06 and it will be smart enough to know
2:20:10 if Kyle has a problem that is beyond its
2:20:14 answers
2:20:15 it connects it with a person
2:20:18 so
2:20:21 so what happens to those people that
2:20:23 program foundries like when that
2:20:25 transition happens when it becomes clear
2:20:27 that
2:20:28 that model has to go the way of the fax
2:20:30 machine what happens to those people
2:20:32 what happens to those systems what
2:20:34 happens to those companies
2:20:37 many of them are going to be like Kodak
2:20:40 and say no no phone trees are still very
2:20:43 vital people love our phone tree we've
2:20:45 got a very different kind of phone tree
2:20:47 we don't say press one we say interact
2:20:50 with one it's a very different thing no
2:20:53 it's not
2:20:56 so I don't know what happens to all
2:20:57 those companies some of them are going
2:20:59 to be replaced by a couple of little 23
2:21:02 year old snot-nosed kids that are like
2:21:04 wouldn't it be cool if we'd have bought
2:21:05 that just answered the phone and told
2:21:07 you everything about the company you
2:21:08 needed to know and had the authority to
2:21:11 make decisions and give refunds and [ __ ]
2:21:13 like that that'd be cool man
2:21:17 and they'll get funded for 200 million
2:21:19 dollars and they'll create that system
2:21:20 and that company that's been writing
2:21:22 phone trees for the last 40 years
2:21:25 will lose their
2:21:27 170 million dollar a year contract
2:21:32 with their customer of 25 years
2:21:36 and they'll be like what just happened
2:21:40 I think I'm going to go golfing and not
2:21:42 come back from it
2:21:48 I don't know I don't know
2:21:52 I'm just messenger
2:21:54 I'm just a lowly messenger with bad hair
2:21:58 and moderately humorous humorous acting
2:22:01 skills
2:22:03 I could have been a contender
2:22:05 agree they should take this chance to
2:22:08 redefine a new paradigm
2:22:11 new paradigms are good and all but
2:22:12 they're painful to get to Smart
2:22:14 contracts are replacing a lot of Ip
2:22:17 lawyers yeah that's it I I think the
2:22:19 smart contract thing is really
2:22:20 interesting and the the Dows
2:22:23 conceptually all the blockchain and nft
2:22:26 stuff the Dow stuff the the fintech
2:22:28 stuff is fascinating they just they just
2:22:31 kind of screwed themselves the biggest
2:22:32 issue I've with AI and marketing wait
2:22:34 the biggest issue I have with AI and
2:22:37 marketing because it's over the top
2:22:39 language geez yeah I know so Silver Fox
2:22:42 one of the things you can do there's a
2:22:44 phrase that you can use I saw on a tick
2:22:46 tock and I tried it and it's interesting
2:22:49 um the last sentence you could put in a
2:22:51 prompt is say limit Pros
2:22:54 and it will not be so wordy
2:22:59 so you could try something like that try
2:23:00 to tell it you know I don't care about
2:23:03 your damn words just give me the
2:23:06 just the facts ma'am
2:23:11 be a better reviewer love Bing
2:23:14 it's funny some people are like love
2:23:16 Bing
2:23:17 other people like hate Bing love Bard
2:23:19 hate Bard which is good that means that
2:23:21 means we're in a competitive space now
2:23:23 right when there's only one option
2:23:24 everyone's like well Chachi BT's the
2:23:26 bump and now all these other things are
2:23:28 are different enough and have enough
2:23:31 different things that different people
2:23:33 are going to love different tools that
2:23:35 feels very healthy to me
2:23:36 for now yes it's a different way of
2:23:38 interacting yep your hair looks great
2:23:40 man thank you at least I have hair right
2:23:42 I mean come on not for nothing
2:23:44 my dad at this age I think was he had
2:23:47 the full Gene
2:23:49 he he had just he had this he had like
2:23:51 uh he took like these six hairs right
2:23:53 here he had them in the middle
2:23:56 oh that was bad all right
2:24:01 what is a free chat GPT platform that
2:24:04 talks other than Bing pie pie talks
2:24:09 you mean like talks like this
2:24:11 get the iOS
2:24:13 PI app pi as in the mathematical circle
2:24:17 formula circle equation no
2:24:21 what is pi it's a result it's a
2:24:24 constant it's a I don't know 3.1415
2:24:28 like that
2:24:30 that you can talk to and it talks backs
2:24:32 it talks back I was doing it in my car
2:24:34 the other day I was driving around we
2:24:35 were driving to Costco I said oh give me
2:24:37 the give me the specials for Costco for
2:24:39 the for the one we were going to and it
2:24:41 did that I asked it about the Taylor
2:24:43 Swift concert
2:24:46 um I asked it about quantum physics
2:24:50 um
2:24:51 what else did I do anyway yeah it's it's
2:24:53 fun it's fun
2:24:55 because you know
2:24:59 chat chat Bots over the past decade and
2:25:01 things like Alexa and Siri
2:25:04 have essentially trained us that these
2:25:06 Bots are [ __ ] idiots and and their
2:25:10 only purpose in life is to ask them for
2:25:12 something very specific and to get the
2:25:15 specific response I would like to order
2:25:18 Cheerios would you like the 96 ounce box
2:25:22 or the 60 ounce box uh the 96 ounce box
2:25:25 great that'll be there Wednesday like
2:25:27 they're good for that right but if you
2:25:30 say Alexa I kind of feel like crap oh I
2:25:33 don't know how to deal with that
2:25:35 crap are you looking for a laxative you
2:25:37 know it doesn't it doesn't know how to
2:25:39 deal with a lot of what we want to talk
2:25:42 to it about so when I was interacting
2:25:44 with pi
2:25:48 it's good
2:25:50 like at one point I was talking about my
2:25:52 kids dealing with Lyme disease and and
2:25:54 and it was like oh I'm really sorry to
2:25:57 hear that I hear that can be that can be
2:25:59 really challenging you know is there
2:26:01 anything I can help you with and I
2:26:03 started having a whole conversation
2:26:05 about it it was like it's like holy [ __ ]
2:26:09 and then I switched gears I said you
2:26:11 know what I wanna I want you to tell me
2:26:12 about you know quantum mechanics and how
2:26:14 that works can you do that well yeah I
2:26:16 can do that absolutely like there's no
2:26:18 judgment
2:26:19 there's deep comp competence
2:26:23 and there's a lot of empathy a lot of
2:26:25 empathetic cues it doesn't have empathy
2:26:28 but it sounds like it does and the and
2:26:31 the way they've got the voice programmed
2:26:32 in pi I did uh for me the the voice on
2:26:36 pie that was good was was Voice number
2:26:38 six let me see if I can I can do it here
2:26:40 on the desktop because it's not quite as
2:26:43 good
2:26:45 um
2:26:46 let me see
2:26:48 is this on
2:26:56 are you there
2:26:59 oops
2:27:00 unexpected error
2:27:03 seems you've been blocked
2:27:06 why
2:27:13 oh
2:27:20 um
2:27:21 [Music]
2:27:25 foreign
2:27:35 [Music]
2:27:40 how are you doing tonight
2:27:43 [Music]
2:27:48 it's not talking
2:27:50 the desktop one is wonky
2:27:54 hey there I'm doing great thanks for
2:27:56 asking how are you today
2:27:59 um
2:28:01 I'm doing good I'm wondering if you can
2:28:03 help me understand quantum mechanics
2:28:08 [Music]
2:28:11 sure I'd be happy to help quantum
2:28:13 mechanics is a fascinating field but
2:28:15 it's also complex so let's start with
2:28:17 the basics
2:28:18 to begin can you tell me what you
2:28:20 already know about quantum mechanics
2:28:23 I know about things like superposition
2:28:25 and a cat that can be dead and alive at
2:28:28 the same time
2:28:30 [Music]
2:28:34 yep famous Schrodinger's cat thought
2:28:36 experiment That's a classic when it
2:28:38 comes to Quantum Mechanics it's actually
2:28:40 meant to be a bit tongue-in-cheek but
2:28:42 it's a useful way to introduce some of
2:28:44 the concepts
2:28:45 let's unpack that a bit shall we would
2:28:47 you like me to start with a basic
2:28:48 overview of quantum mechanics first
2:28:52 no I'm kind of bored with physics could
2:28:54 we could you tell me if there's any
2:28:58 reviews of the Taylor Swift concert from
2:29:00 last night
2:29:00 [Music]
2:29:06 there's definitely a lot of Buzz about
2:29:08 Taylor Swift's Eris tour right now
2:29:10 I can tell the concert has been a huge
2:29:12 success people are calling it an
2:29:14 unforgettable experience with plenty of
2:29:16 nostalgic moments and references to
2:29:18 Swift's classwork would you like me to
2:29:20 give you a bit more detail about the
2:29:21 concert and some of the reactions from
2:29:23 fans and critics
2:29:26 yeah that would be awesome
2:29:28 [Music]
2:29:31 I'll tell you all I know about it so
2:29:33 first I'm going to warn you that there
2:29:35 may be some mild spoilers here but in
2:29:37 the concert Swift mix
2:29:39 all right I'll turn it off so
2:29:41 I mean that that's that's haypi.com
2:29:46 h-e-y-p-i.com is where I'm just playing
2:29:48 and then the there's an IOS app called
2:29:51 pi
2:29:53 um
2:29:54 so so I mean so that one's really
2:29:56 interesting because the the way the way
2:29:58 they've got that one programmed it's
2:30:00 much less about giving you an answer so
2:30:03 so
2:30:04 um what they're talking about is
2:30:05 creating a personal assistant
2:30:08 not about a chat bot right not about a
2:30:11 large language model not about a thing
2:30:13 that gives you answers this is more a
2:30:15 thing that gives you dialogue so a lot
2:30:17 of times if you ask it for an answer it
2:30:20 will start interrogating you it'll start
2:30:22 asking you questions and so and the way
2:30:24 that it's programmed to ask the
2:30:26 questions it's very human
2:30:30 and so
2:30:32 you know I think I think as humans we
2:30:34 already
2:30:35 anthropomorphize technology as it is
2:30:38 and so when when you can truly start
2:30:41 having conversations and when when the
2:30:43 when the voice synthesis gets a little
2:30:45 less stilted like like it's getting
2:30:47 better and it's it's close it's not
2:30:49 quite there but when it gets a little
2:30:51 less stilted and and they remove that
2:30:54 that you know half a second delay
2:30:57 or whatever it is one second delay
2:31:00 when when the when the response is
2:31:02 immediate and I'll tell you when that
2:31:04 will be
2:31:06 they're getting the the uh optimization
2:31:09 of these large language models to the
2:31:11 point that you're going to be able to
2:31:13 install them on your phone
2:31:15 and and like Apple's silicon Apple's M2
2:31:19 and their R1 their reality or whatever
2:31:23 it's called the the R1 chip that that
2:31:25 reads from all the sensors it's doing
2:31:27 all the machine learning
2:31:30 computation locally on the device when
2:31:33 you have a large language model that's
2:31:35 that sits on the phone it doesn't have
2:31:37 to reach out to the internet
2:31:39 and and you can do the machine learning
2:31:41 processing like that
2:31:44 and then you've got voice synthesis that
2:31:46 sounds like a human
2:31:48 you're going to be able to have
2:31:50 incredibly powerful conversations about
2:31:53 anything it could be
2:31:56 your virtual physical therapist it could
2:31:59 be your
2:32:00 psychotherapist it could be your virtual
2:32:03 friend it could be
2:32:04 a counselor to help you get through
2:32:06 grief it could be
2:32:08 someone helping you with anxiety like
2:32:11 whatever
2:32:12 and it's creepy and weird as it sounds
2:32:14 right now like
2:32:16 that conversation I just had about the
2:32:19 Taylor Swift concert and the Google or
2:32:23 the quantum physics stuff and things
2:32:24 like that
2:32:27 like this is where we get into weird
2:32:29 hyper personalization stuff I did I did
2:32:31 something this afternoon with it where I
2:32:34 said um
2:32:37 I said I've got to go I've got to go
2:32:40 wait in a line somewhere
2:32:43 or I was getting my car washed and I was
2:32:45 waiting to to get the like the interior
2:32:48 vacuumed out or whatever
2:32:50 and so I'm just sitting waiting and I
2:32:52 said can you come up with a game for me
2:32:55 to play where I learned something
2:32:57 and it said oh that's a great idea what
2:33:00 would you like to learn about
2:33:02 and then I came up with the whatever the
2:33:05 topic was and then she said she said
2:33:08 this is awesome she said yeah we'll do a
2:33:11 game I'll ask you questions and every
2:33:13 time you get a question right I'll give
2:33:15 you points and it's like it was like it
2:33:18 just came up with a game and it was an
2:33:20 educational game and it's like so
2:33:25 you know do we quickly move from people
2:33:27 listening to audiobooks
2:33:30 to people having a conversation about
2:33:35 the audio books or about whatever books
2:33:38 or any topic or any time in history or
2:33:41 any person they want to learn about
2:33:45 so it's coming like the the the the
2:33:49 dramatic difference between Siri and
2:33:53 that conversation I just had
2:33:56 like can't be understated or it can't be
2:33:58 overstated it's it's insane it's just
2:34:00 such it's a completely different thing
2:34:03 they're like I don't even we don't even
2:34:05 know how to do it and it's the other
2:34:07 thing is going to be this you'll you'll
2:34:08 be having these conversations it'll be
2:34:10 talking back and you're like oh could
2:34:11 you just show me a picture of that sure
2:34:13 boom up it comes
2:34:15 what would the what would that sound
2:34:16 like as a song Oh do you want to hear it
2:34:18 do you have a particular genre you like
2:34:21 more than the others yeah I like
2:34:22 Americana okay let me do an Americana
2:34:25 song about that
2:34:28 is this what you had in mind that's a
2:34:30 little too slow do you have something in
2:34:31 in three
2:34:32 that's that's a little peppier sure
2:34:34 here's that yeah that's good could you
2:34:37 save that to my music library great
2:34:38 anyway where were we right it's just
2:34:42 you just go
2:34:44 hyper personalization of media
2:34:48 who are the media creators then
2:34:51 is it the tool builders
2:34:53 is an application layer sitting on top
2:34:55 of the tool
2:34:57 I don't know I don't know people
2:35:00 I got no answers tonight I got I got
2:35:03 lots of
2:35:05 I got lots of predictions and and vision
2:35:08 I got no answers tonight I was in the
2:35:10 film industry and
2:35:11 and I want to be actor s for my advice I
2:35:14 said get a new Dream Dream Crusher I'll
2:35:18 tell you Chris it's not bad it's not bad
2:35:20 advice
2:35:21 it's not bad advice it's a it's a rough
2:35:24 business
2:35:25 but you know some people have to do it
2:35:27 like like when I decided I wanted to be
2:35:30 an actor it was It was a
2:35:34 15-year obsession
2:35:38 and at some point I think I feel like
2:35:40 I'll do it again but I'll do it for a
2:35:42 very different reason right I'll that
2:35:43 gets back I was talking earlier about
2:35:45 you either love the process or you love
2:35:47 the end product and if you love the end
2:35:49 product this stuff is awesome for you
2:35:51 and if you love the process this stuff's
2:35:53 a disaster
2:35:54 so if I think if I get back into acting
2:35:56 I'll do it more for the process and if
2:35:58 it turns into good results great but I
2:36:00 won't give a [ __ ] which that'll be like
2:36:04 it's an awful feeling to be an actor and
2:36:06 to want to work so bad and then you just
2:36:09 can't because it is such an
2:36:11 oversaturated Market
2:36:15 do you moisturize secret dancers back
2:36:19 here talking about my dry skin again I
2:36:21 think I've got some shine I'm a little
2:36:23 moist I don't know what the hell you're
2:36:24 talking about secret dancer I don't even
2:36:27 think you dance
2:36:29 and you're not that secret
2:36:31 AI knows AI knows what's going on with
2:36:34 you
2:36:35 I love your acting telling AI to go [ __ ]
2:36:38 itself thank you hey I'm gonna tell you
2:36:42 what yes so check that Ai and it [ __ ]
2:36:45 lied to me so then I said go [ __ ]
2:36:47 yourself yeah I'll be over here I'll be
2:36:49 over here you tell me you tell me when
2:36:51 you want to get back to me
2:36:52 when you're not gonna lie to me huh this
2:36:55 is about truth this is about integrity
2:36:58 um
2:36:59 this is how we work here you coming at
2:37:02 me from the side yeah I don't go for
2:37:03 that
2:37:06 the robots are gonna kill us
2:37:08 he's selling too
2:37:10 oh secret dancer
2:37:12 good Lord get a life
2:37:17 um
2:37:18 if you're not I I gotta tell you I do a
2:37:22 lot less selling than I could
2:37:24 I'm on here two or three hours a night
2:37:26 just giving and giving and giving and
2:37:28 this is the thanks I get Shaw
2:37:31 all right
2:37:32 if you're not living on the edge you're
2:37:34 taking up too much room yeah yeah but
2:37:38 you know some people don't like living
2:37:39 on the edge he wants us to subscribe no
2:37:42 I don't I don't really care in fact I'm
2:37:44 going to take off the subscription thing
2:37:45 because I don't really give a [ __ ]
2:37:50 yeah baby all right
2:37:53 all right
2:37:56 [Music]
2:37:58 secret dancer you shouldn't subscribe
2:38:00 you shouldn't even show up for these
2:38:03 they are I'm unqualified and it's not
2:38:05 worth anything so don't subscribe any of
2:38:08 you and secret dancer you can leave
2:38:10 please
2:38:12 all right I think
2:38:13 the personalization is cool but will be
2:38:16 a novelty and people will get over it I
2:38:18 don't think so I don't think so I don't
2:38:20 think it's going to be exclusive I don't
2:38:22 think it's going to be like personalized
2:38:24 content completely subsumes
2:38:27 you know generated content I think I
2:38:29 think it's just like when film came
2:38:31 along
2:38:32 you know plays didn't go away right we
2:38:35 have we still have Broadway
2:38:36 but we have Broadway right we we don't
2:38:40 have theater as the only Entertainment
2:38:43 form other than reading and writing
2:38:45 right
2:38:47 you know films came along it's it's a
2:38:49 new medium and then video games came
2:38:51 along it's a new thing I think this idea
2:38:53 of hyper personalization comes along
2:38:54 it'll be a new thing we'll still have
2:38:57 those other forms
2:39:01 but
2:39:03 you know
2:39:05 you really hate the sad ending in
2:39:07 whatever your favorite film is
2:39:12 let me watch that film where they don't
2:39:14 all die at the end okay
2:39:20 and I think what's what's fascinating in
2:39:22 that in that
2:39:24 future
2:39:26 is it's very possible that I create a
2:39:30 version of a movie that has the happy
2:39:32 ending and you know the original had the
2:39:34 sad ending
2:39:36 and they'll probably be the equivalent
2:39:38 of Netflix of the future where I can say
2:39:42 hey that was good enough that I want to
2:39:45 sell that and I'm going to put that on
2:39:46 the Netflix of the future
2:39:49 and there's the very real possibility
2:39:51 that my version of that movie will will
2:39:54 be
2:39:56 as popular as the original or more
2:39:59 popular than the original
2:40:01 and then and then what happens then
2:40:03 right so do the original creators get 90
2:40:05 of their revenue and I get 10 do I get
2:40:09 50 50 do I get nothing
2:40:12 because it was just a brain fart and I
2:40:14 put it out there I don't know
2:40:17 why is chat EBT developed to only 2021
2:40:20 with its content the way they build the
2:40:22 large language models is a very
2:40:25 labor-intensive
2:40:26 uh manual process where they have to tag
2:40:29 a lot of the content and the data we
2:40:31 have to then do this thing called
2:40:33 embedding which is very expensive where
2:40:35 they turn all the
2:40:36 all the words from all the content into
2:40:38 these vectors these numbers you know
2:40:41 semantic
2:40:43 language clouds
2:40:46 um
2:40:47 and and they have to basically just lock
2:40:49 it down at some point and say okay you
2:40:51 know this model cuts off at this point
2:40:54 so Claude
2:40:57 um is a more recent model so it cuts off
2:40:59 sometime in 2023
2:41:01 but if you use Bing Bing is gpt4's base
2:41:06 large language model that that's that
2:41:09 stops at 2021 but because it's connected
2:41:12 to the internet
2:41:14 if you ask it about Taylor Swift it'll
2:41:16 basically go do a search on the internet
2:41:18 get the latest data and it'll augment
2:41:21 you know on top of that large large
2:41:24 language model
2:41:25 but that's why it does it because it's
2:41:27 just it's a very labor-intensive process
2:41:29 to make that base model
2:41:31 do you offer a list of your top favorite
2:41:34 AI apps yeah I sit beside it every night
2:41:38 these are them right now so
2:41:40 um
2:41:42 the things I pay for
2:41:45 chat GPT
2:41:48 I'm gonna get rid of Poe but I pay for
2:41:50 Poe Claude is free Bard is free
2:41:54 I pay for mid-journey
2:41:57 I pay for zapier which is an automation
2:41:59 tool
2:42:00 and I pay for
2:42:02 Google collab in case I want to geek out
2:42:09 and I might pay for one other but it's
2:42:11 basically a large language model an
2:42:13 image generation tool
2:42:15 an automation tool
2:42:19 and a and a virtual machine Google
2:42:22 collabs like a virtual environment where
2:42:24 I've got access to gpus if I want to run
2:42:27 models
2:42:28 you know in a kind of in a standalone
2:42:31 way
2:42:32 um
2:42:34 and that's it right now like I'm not
2:42:36 paying for a lot because this stuff's
2:42:38 moving too fast like all the stuff at
2:42:41 Future pedia there's probably you know
2:42:43 there's probably 30 or 40 amazing Tools
2:42:47 in there but you have to sift through
2:42:49 38 or 3 800 or 4 000 to find those 30 or
2:42:54 40 amazing tools
2:42:57 Runway ml is really interesting for
2:42:59 video
2:43:02 um if you just need loopy kind of
2:43:04 background music
2:43:06 um for
2:43:08 projects then sound draw is interesting
2:43:11 it's like it's like GarageBand except
2:43:13 you just describe a genre and it makes
2:43:15 you 20 or 30 songs
2:43:17 and then you can you know mix and match
2:43:19 the clips and the sounds and you know
2:43:21 it's
2:43:23 so that's interesting I don't know
2:43:26 that's about it
2:43:27 like like these things are so deep you
2:43:29 know I could I could easily spend the
2:43:31 next decade trying to figure this [ __ ]
2:43:32 out like like what they actually mean
2:43:35 what what they actually do code
2:43:37 interpreter at chat GPT is is so
2:43:41 mind-blowing and I like I I am so
2:43:46 um under educated about
2:43:50 [Music]
2:43:50 um
2:43:51 what I can even do with it or like what
2:43:53 it what it even makes possible
2:43:56 other than to know holy [ __ ] it's going
2:43:59 to change everything so I need to get
2:44:01 much more educated about code
2:44:03 interpreter and what it makes possible
2:44:05 and that's just going to be through
2:44:06 experimentation so I just and the
2:44:08 plugins like and now that there's five
2:44:11 of these and there's pie the one I was
2:44:13 just talking to there's just so much
2:44:15 just in the in the chat GPT models I
2:44:18 don't think you have to dig much farther
2:44:20 than that other than the image
2:44:21 Generation stuff
2:44:24 radical personalization Choose Your Own
2:44:26 Adventure that's exactly right he's
2:44:28 selling services I'm not selling
2:44:29 services actually I've got a uh
2:44:33 a video technology company we sell
2:44:35 subscriptions but really to Pharma I
2:44:38 don't give a [ __ ] if you
2:44:39 by any of our video services
2:44:42 I think we'll see greatness out of the
2:44:46 man secret dancer needs a hug can we can
2:44:49 we all give secret dancers some love
2:44:51 I think they are uh they need a hug
2:44:55 I think we'll see greatness out of the
2:44:57 generation that grows up with AI yeah
2:45:00 for sure
2:45:01 here we go money
2:45:03 oh secret dancer he's so sweet
2:45:06 all right what do you think of the SAG
2:45:07 strike I talked a lot about that tonight
2:45:10 [Music]
2:45:11 um
2:45:12 Miss perch secret dancer what the [ __ ]
2:45:15 are you talking about
2:45:18 don't pay for free services I agree
2:45:21 don't pay for free services certainly
2:45:23 don't pay an X actor a penny
2:45:29 oh my God what services are you talking
2:45:33 about these I don't make any money off
2:45:35 these
2:45:37 that's open AI like Bings free clutch
2:45:40 free pose free bards free
2:45:43 you can pay for Poe you can pay for a
2:45:46 chat gbt but you don't have to pay for
2:45:47 any of them that's free that's free
2:45:50 I don't know what you're talking about
2:45:52 but whatever I don't give a [ __ ]
2:45:54 okay I will be ignoring secret dancer
2:45:57 from now on
2:45:58 what do you think about sag strike I
2:46:01 think it's I think they're fighting a
2:46:02 losing battle we love you Kyle thank you
2:46:04 Troy appreciate that I gotta go
2:46:08 um I thought he said stand up not shut
2:46:10 up not sure what that means is Kyle
2:46:13 Frozen artificial intelligence movie
2:46:15 2001 Dr no
2:46:17 Dr No scene reminds me of GPT we're we
2:46:20 are in uh here's an interesting thing so
2:46:23 I did it I did a table go to go to go to
2:46:25 any one of these
2:46:26 and type in make me a table of science
2:46:30 fiction movies
2:46:33 um
2:46:34 what are the major technologies that
2:46:38 that were science fiction at at a time
2:46:40 and then either give me it's here now
2:46:44 um like like
2:46:47 you know is it is it here now it'll be
2:46:49 here in one to two years it'll be here
2:46:50 in five years or it'll be here in a
2:46:52 decade and it made me a whole list of
2:46:54 movies and then and then what the
2:46:56 Technologies were and it said you know
2:46:57 what things were here or coming it's a
2:47:00 really fascinating list because it's
2:47:01 like we're sort of going through the
2:47:04 movies of our you know of our childhood
2:47:06 going yeah we can do that now we can do
2:47:08 that now we can do that now so it's
2:47:10 amazing
2:47:11 uh what site would you suggest to write
2:47:14 a book well you could play with a GPT
2:47:17 author if you're geeky and want to fire
2:47:19 up Google collab
2:47:21 um but I I would use gpt4 at open AI
2:47:26 still Frozen for me it's my internet
2:47:28 Thelonious that stinks turning off Wi-Fi
2:47:31 open up opens up new worlds for me for
2:47:34 Thelonious
2:47:36 uh good Lord all right I gotta go let's
2:47:40 see xrp is about to go parabolic
2:47:42 someone's talking xrp in my little AI
2:47:45 chat we don't talk about crypto Bros in
2:47:48 here work harder nobody cares come on
2:47:51 Kyle what site were you speaking of I'm
2:47:54 not sure cast yourself LOL
2:47:56 um
2:47:58 you don't read live comments well tigi
2:48:00 nuggy okay Tiggy Nuggie
2:48:03 [Music]
2:48:03 um
2:48:05 7.5 out of 10 on the name it's close
2:48:08 it's close 7.5 out of 10.
2:48:12 uh no I just read I read all the
2:48:14 comments so I'm usually way behind but
2:48:16 Tick Tock just jumped me to the bottom
2:48:17 and I gotta go so now I'm kind of at the
2:48:19 bottom scrolling upwards what's your
2:48:21 view on that uh what about elon's hyped
2:48:23 up xai what's your view on that um
2:48:26 he's gonna do what he wanted Sam Altman
2:48:29 to do he's going to create an open
2:48:30 source
2:48:32 well he says he's going to create an
2:48:34 open source
2:48:35 you know AGI
2:48:39 seeking company
2:48:41 um
2:48:42 we'll see if he actually does it
2:48:46 I mean I wouldn't put it past him I mean
2:48:48 he's got he's got the resources he's
2:48:50 hired some really [ __ ] smart people
2:48:52 he wants to understand the universe like
2:48:54 the imperative that he talks about that
2:48:56 this is Elon Musk with his xai company
2:48:58 the the imperative that he talks about
2:49:00 about understanding the universe is a
2:49:02 really interesting imperative for the AI
2:49:05 but uh
2:49:07 we'll see we'll see he's got a lot of
2:49:10 competition man I mean I mean Chachi PT
2:49:13 doing what it did and sort of unleashing
2:49:15 the floodgates of these competitors this
2:49:18 quickly plus pi plus meta plus we
2:49:22 haven't heard anything from Apple yet
2:49:23 plus xai I mean we're gonna have at
2:49:26 least a dozen
2:49:29 increasingly powerful multi-modal
2:49:33 tools that just do all this [ __ ]
2:49:39 and and and then they're all going to
2:49:41 get incorporated into all the tools
2:49:43 we're already using so we're going to
2:49:44 have tools outside of what we do day to
2:49:46 day and it's it's going to start to
2:49:48 infiltrate all the stuff that we do so
2:49:50 it's going to be everywhere
2:49:55 um
2:49:56 [Music]
2:49:57 where can I find this site I just came
2:49:59 up on you live I don't know which site
2:50:02 you're talking about oh that site was Pi
2:50:05 heypi.com or Pi dot a i
2:50:09 was the one where I was talking to the
2:50:11 bot that was Pi that's inflections uh
2:50:15 service it's pretty good between my AAA
2:50:18 membership and AI leveled up assistant
2:50:20 sounds like the best boyfriend ever
2:50:22 that's awesome holy [ __ ] my hand and my
2:50:26 face are hurting from the Kevin
2:50:27 McAllister moments every night here
2:50:32 all right all right everybody where can
2:50:35 I find pie it's not p-i-e it's p i dot a
2:50:39 I or hey pie
2:50:42 h-e-y-p-i.com or no dot a wait
2:50:45 haypi.com and then Pi dot AI
2:50:50 all right everybody I gotta go it's been
2:50:52 fun hanging out with you secret dancer I
2:50:55 hope you get a hug
2:50:56 I love you too
2:50:59 um
2:51:00 being shot of these things if you're if
2:51:03 you just got here
2:51:05 grab some of you know grab these URLs go
2:51:08 start playing with them
2:51:11 that's official chat that's Microsoft
2:51:13 it's getting increasingly powerful
2:51:15 claude.i AI is a really interesting
2:51:18 competitor to chat
2:51:20 gpt4po.com you can make Bots and play
2:51:23 with six different models there and then
2:51:26 Bard is
2:51:28 pulling up the pulling up the rear but
2:51:31 but getting really interesting like
2:51:32 they're incorporating uh links out to
2:51:35 the internet now in a pretty organic
2:51:38 looking way that's way better than I
2:51:40 thought it would be so all of these are
2:51:41 good prompts.chat will teach you how to
2:51:43 prompt future pedia we'll let you find
2:51:46 all sorts of rabbit holes to run
2:51:47 yourself down peace out everybody hope
2:51:50 you had a good time tonight
2:51:52 foreign