The Invisible City Workshop

Description

Chapters

Transcript

0:00 Tik Tok so if you want to follow me on
0:02 Tik Tok it's the AI learning lab I go
0:05 live nightly for about two or three
0:06 hours just talking about AI so I should
0:09 be able to talk about this
0:10 tonight um just quickly before we start
0:13 how many people um are would consider
0:17 themselves active AI
0:19 users okay how many people have never
0:23 used Ai and it's okay like okay too and
0:27 how many people have kind of heard of it
0:28 and they're like a little like h what is
0:32 this okay two or three okay good
0:38 cool on my first day of seventh grade in
0:44 1977 I walked into my math class and
0:49 that's kind of what it looked like these
0:50 aren't real computers this is mid
0:52 Journey imagining what a seventh grade
0:55 classroom looked like with a bunch of
0:56 computers on the desks and it's actually
0:59 kind of what it looked like
1:01 what were on the desks were Radio Shack
1:03 tr8s which if there's some old people in
1:06 here you've seen those before with the
1:08 Dual 5 and a/4 inch floppy drives oh
1:11 yeah and
1:14 as as I learned what these machines were
1:17 and I learned how to program them I
1:21 realized holy crap this is going to
1:24 change everything that that that using
1:28 these things felt like what it must must
1:30 have been like when the the printing
1:32 press was invented or when the steam
1:34 engine was invented like this felt like
1:35 one of
1:36 those and I
1:39 remember you know I had
1:43 Heroes and you know Steve Jobs and and
1:45 wasak that that founded Apple and Bill
1:48 Gates um and and the team from Microsoft
1:52 that created this and I remember in
1:54 seventh grade being really
1:57 sad that I was just too young
2:00 to be one of them that I had missed it
2:04 that I just
2:05 knew that this was going to change
2:08 everything and I wasn't a part of it and
2:11 I knew that I was going to like live
2:14 with that for the rest of my
2:17 life jump forward to
2:20 1994 I'm living in New York City I have
2:23 a degree in acting so it's perfect that
2:25 I'm talking about AI
2:28 right um I moved to New York out of
2:31 school uh to pursue a life in the Arts I
2:33 started and ran a theater company
2:37 um that didn't work out financially it
2:39 was good creatively but you know you
2:40 don't make a lot of money doing that so
2:42 I thought I know because I'm a genius
2:44 I'll write myself an acting career so in
2:46 two years I wrote seven feature length
2:49 screenplays um and and for my day job
2:51 you know you have to you have to hustle
2:53 in New York for my day job I was doing
2:55 um desktop publishing and I kept hearing
2:58 people talk about the internet and I and
3:00 I kind kind of got on the internet and
3:01 the internet was kind of cool but you
3:03 had to dial up right and everything was
3:05 command line and you'd go to FTP sites
3:08 or tet sites and these libraries would
3:10 have like dissertations and documents
3:13 and I'm like this is kind of cool and
3:16 then people started talking about this
3:17 thing called the worldwide
3:19 web and I started learning about it and
3:23 I I I thought well this seems kind of
3:26 amazing that that there's there's
3:28 something where you can include pictures
3:29 and it's got these things called
3:31 hyperlinks and and I said but surely I
3:33 can't win
3:35 that and so I started reading about it
3:38 and I started talking to other people
3:39 about it no one knew about it a handful
3:42 of people knew about it and I started
3:44 asking questions you know what is this
3:45 worldwide web thing what's this HTML and
3:47 the more I learned about it the more I
3:49 realized oh crap the way you structure a
3:52 website is almost identical to how you
3:54 structure a screenplay it's just
3:55 storytelling in a different form and I
3:58 was like holy crap maybe I can do this
4:00 and so I had an idea I thought the
4:03 internet right now is is full of
4:06 dissertations and research and Technical
4:08 crap there was no art there was no there
4:11 was no culture and so I came up with
4:13 this idea for an online magazine called
4:15 Urban
4:17 desires and I went on vacation with my
4:19 wife and took my Mac 660 AV and a 13-in
4:22 color monitor and I taught myself enough
4:25 HTML to make this this is the first the
4:28 homepage of the first issue of urban
4:30 desires from
4:31 1994 you'll notice that the the um you
4:35 could click on the the cover and see it
4:37 big like this but you'll notice it's
4:39 left Justified why because the center
4:41 tag didn't apply to images in
4:45 1994 yeah exactly um but I I so I got I
4:50 got home and I I had this this magazine
4:53 and and I didn't know what to do with it
4:55 next and so I asked one of my friends
4:57 that knew something and they're like
4:58 well you got to put it on an internet
4:59 server and I'm like what's an internet
5:01 server and they said well I know one in
5:03 California and they gave me this phone
5:04 number of this computer in California so
5:07 so I had my computer call this computer
5:09 in
5:11 California oh
5:13 [Music]
5:20 wait and if you don't know what that
5:22 sound is I hate you that means you're
5:24 too
5:25 young there was a time when we had to
5:28 have our computer call other computers
5:29 to talk to them the internet wasn't just
5:33 there so I published this this Magazine
5:37 on this thing called the internet and
5:38 I'm like okay like what now and 3 weeks
5:43 later I got a I got a note on a on a
5:45 bulletin board a dialup bulletin board
5:47 that I was a part
5:48 of and someone said hey Kyle did you
5:51 know that Urban desires was written up
5:53 in
5:54 libero the Parisian News
5:57 Daily and I was like what
6:00 and they
6:01 said you can I get a little choked up
6:05 over
6:06 this you can go to the international
6:09 news stand at Times
6:10 Square and get it so the next Sunday I
6:14 went to Time Square and I bought this
6:17 French newspaper
6:19 and um flipped through it and sure
6:24 enough there's a full page article of
6:28 this this magazine that I created in my
6:32 apartment in Brooklyn and shipped to
6:34 some computer in California written
6:36 about in Paris 3 weeks
6:39 later and what hit me as I stood in the
6:41 middle of Time Square was that the world
6:44 just changed and nobody knows it
6:48 yet
6:53 um it was
6:55 remarkable and from that I co-founded
6:59 agency.com which was one of the first
7:01 digital
7:02 agencies and we ended up building
7:04 websites for the Fortune 500 we ended up
7:06 growing to 2200 people it was it was AB
7:09 in 5 years it was absolute
7:11 Insanity um we built British Airways
7:14 first website we built their ticketing
7:15 system we built the first um streaming
7:19 technology for serious satellite radio
7:22 um because none of it existed all of it
7:24 had to be
7:26 invented
7:28 um and you know
7:31 the early days of the worldwide web get
7:32 a get a bit of a bad rap like history
7:34 hasn't treated it well because of the do
7:36 boom and bust everyone's like oh that
7:39 was just crap and yeah no but it changed
7:41 the world and if you were in if you were
7:44 playing with this stuff in those early
7:46 days it it wasn't like the oversold
7:49 hyped up thing that everyone talks about
7:51 in 2000 it was people that just
7:54 passionately believed in this oh my God
7:56 this is going to change
7:58 things so and I I I really thought that
8:02 that would be the last time in my life
8:06 that I would get to be at the beginning
8:07 of something like that like just I felt
8:12 so [ __ ]
8:13 lucky right to be a part of
8:16 it and I kind of knew like these don't
8:19 come along that often right it's like
8:21 fire the wheel or the wheel fire you
8:24 know language writing printing press
8:28 steam engine right there's not that many
8:30 of these the 20th centuries had a lot
8:32 but there's not that
8:34 many so jump forward to
8:36 2022 this image is one of the very first
8:40 um AI images I ever created um i' kind
8:43 of done a deep dive into nfts and
8:45 blockchain and I was I was you know kind
8:47 of into into that but I got involved in
8:49 this project called pixel mind and it
8:51 was an nft project where it was kind of
8:53 like Mad Libs where you just fill in
8:55 like an animal or whatever and it would
8:56 make a piece of art and it was really
8:57 cool and this was one of my first pieces
8:59 of art and I'm like wow that's amazing
9:02 and and you know it had like
9:05 uh like metadata about the information
9:08 and every every image you created had a
9:10 video attached to it and this is the
9:14 computer making that art you got to
9:17 watch it make the
9:20 art and I was I was like I was like huh
9:25 that's
9:27 curious and for the next six months or
9:31 so this is the beginning of
9:34 2022 I would sort of go to this pixel
9:36 mind site and I I would do it in batches
9:38 I would go make you know 20 images or
9:40 100 images or whatever and you could
9:42 mint them into nfts which I'm like oh
9:44 cool I can sell these they're worthless
9:46 no nobody gives a
9:49 [ __ ] and then one of the one of the days
9:52 I went there and there was this madd's
9:54 prompt for I don't know some sort of
9:56 futuristic cyborg
9:58 portrait and so I did Bill Gates and
10:02 Steve Jobs and I threw Elon Musk in
10:03 there because we're going to Mars and he
10:05 you know made the electric car so so you
10:09 got these kind of kind of titans of
10:14 innovation and what what struck me when
10:17 I was doing this was like what's the
10:20 point like I'm making all these images
10:22 I'm making all these what what am I
10:23 using them for what am I using them for
10:26 and I thought what I should learn what's
10:29 underneath this and what was underneath
10:30 this is a thing called stable diffusion
10:32 which is an image tool you can play with
10:33 right now it's free it's open source a
10:36 lot of drama going on at the company
10:37 right now if you follow that stuff
10:40 um and
10:43 um and the more I learned about it it
10:45 was kind of like that HTML stuff that I
10:49 learned in the '90s where it was like I
10:51 think I can figure this out I think I
10:52 can figure it out and I learned that you
10:54 had to use a notebook from this thing
10:56 called Google collab and you had to get
10:58 a Google collab account you got the
11:00 notebook and then you hit these play
11:01 buttons and it would do the programming
11:03 for you I failed nine times nine times
11:07 like not nine times in a sitting like
11:09 nine different times over a two-month
11:11 period I tried to get stable diffusion
11:13 working and I would hit the play button
11:15 and it would do its thing and then it
11:16 would give me an error and I didn't know
11:18 what to flip and
11:19 do and then I eventually got it and and
11:24 I I also installed this thing called
11:25 dream Booth which let me upload um 25
11:30 pictures of myself and then I could make
11:32 these crazy portraits of me because the
11:35 internet doesn't know what I look like
11:36 it knows what Steve Jobs and Bill Gates
11:38 look like it doesn't know what I look
11:39 like so I got this thing working and I
11:42 started making all these crazy
11:44 self-portraits of myself and they were
11:46 bizarre and they were crazy and they
11:48 were me but they were not me and like I
11:51 thought like I want to see that movie
11:54 like I want to know what future he lives
11:56 in and it's like oh like what life
11:59 choices did this guy make to to end up
12:02 there right you know I'm pretty sure
12:05 he's busting down your door for Thought
12:08 violations um and like yeah I might age
12:11 okay you
12:13 know um and if you want to see these you
12:15 can go to um Kyle Shannon dreams both on
12:18 Instagram and Tik Tok what I ended up
12:20 doing was I said I I want to know these
12:23 stories and and you know I'm a writer
12:26 and and so I wrote stories for each of
12:28 these little things and I was kind of
12:31 like okay I think I'm getting what what
12:35 this AI stuff's about I think I'm
12:37 getting and here's here's what struck
12:40 me what happens with
12:43 AI is especially now because you don't
12:46 have to go through all the technical
12:47 [ __ ] I went through to get this up
12:49 and running you can just do this
12:51 now
12:54 um if you've ever done anything well if
12:56 you've ever done [ __ ] anything right
12:59 everything everything you do takes a lot
13:01 of effort right it takes a lot of time
13:03 it takes a lot of effort what I was
13:06 experiencing with AI was I put in a
13:09 little bit of effort and I was getting
13:11 back massive amounts of value little bit
13:14 of effort massive amounts of value when
13:16 I was doing these images one of the
13:17 images It produced took my breath away
13:21 like literally I went like oh my God
13:24 because what it
13:27 was was an image that if I had the
13:30 talent is the kind of image I would have
13:33 created and that's just like it's like
13:36 wow holy crap
13:39 then as if that weren't
13:42 enough this
13:46 happened everybody know what that date
13:48 is anybody know what that date
13:53 is that was the day chat GPT came
13:58 out
14:02 chat does for AI and machine learning
14:05 what the worldwide web did for the
14:09 internet one of the big misconceptions
14:11 about AI is like I got to be a
14:13 mathematician I got to be a scientist I
14:14 didn't go to MIT I don't have a degree
14:16 from
14:18 Stanford prior to chat GPT you really
14:21 did have to be a mathematician you
14:23 really did have to be a scientist or a
14:24 researcher or a
14:25 developer chat gbt made it simp Le for
14:30 the rest of us to tap into the power of
14:33 this stuff that's been developed for 30
14:34 40 50 years probably 70 years if you go
14:37 back to the original neural network
14:41 designs we are at a remarkable time in
14:45 history and this date in my opinion is
14:49 the beginning of of an entirely New Era
14:53 an entirely New Era so let me give you a
14:56 little bit of context of or or just aoup
14:59 couple of pieces of information about
15:01 chat
15:03 GPT um one is bonus question anybody
15:08 know what GPT stands for
15:11 graphical
15:13 close I remember but I can't generative
15:17 pre-trained
15:21 Transformer the G the G in it's the
15:23 really important one generative it's
15:26 Genera it's not copying and pasting
15:29 that's one of the massive
15:30 misconceptions it's not how it works
15:32 it's generating original stuff pre-train
15:34 means it was pre-trained on the internet
15:36 Transformers just the technology so
15:38 here's some important dates 2017 Google
15:42 invents this thing called the
15:43 Transformer that's the T and
15:45 GPT it was a paper called attention is
15:47 all you need they invents they invent
15:49 this thing that allows machine learning
15:52 scientists to train these um models on
15:56 massive data
15:57 sets um open Ai and Sam mman were were
16:03 the the ones who really
16:05 understood the more compute we throw at
16:07 this and the more data we throw at this
16:09 the better they seem to get and so they
16:11 did gpt1 in 2018 and gpt2 I think the
16:15 year after that gpt3 comes out in
16:18 2020 so what they did was they got as
16:21 much of the public internet data as they
16:25 could and they they coded it in such a
16:27 way that they could put it into this
16:28 model
16:30 and gpt3 came out in
16:33 2020 and started to be really remarkable
16:36 if you were paying attention to to you
16:38 know internet and Tech stuff back then
16:40 like people buzzed about this a little
16:41 bit but you still had to be an engineer
16:43 the interface to use it was this thing
16:45 called the open AI playground the GPT
16:47 playground and you had to pick models
16:49 and there were weights and temperatures
16:51 and it was just it was it was too
16:53 [ __ ]
16:55 hard
16:56 um interesting little tidbit in the
16:59 documentation for
17:01 gpt3 were the instructions for how to
17:04 build chat GPT and nobody built it wow
17:09 yeah so in in yeah exactly in the fall
17:14 of uh 2022 or 2023 it becomes apparent
17:18 that Microsoft is going to invest in
17:21 open Ai and so Sam Walman says we should
17:24 probably have something to show like
17:25 when we make this announcement that
17:26 they're making this big investment hey
17:29 Tech kids uh could you just go whip up
17:31 that chat GPT thing that no one's built
17:34 they went out for drinks the night
17:35 before and they assumed that no one
17:37 would care they figured it might get
17:38 like a couple of blog posts and 5,000
17:40 views or something like
17:43 that the little the straight red line is
17:47 chat GPT the horizontal axis is how long
17:50 it took to get to 100 million
17:52 users um to put what's happening with AI
17:56 in context of the worldwide web
18:00 to get to 100 million users it took the
18:03 worldwide web about 6
18:05 years it took chat GPT 6
18:09 weeks it was the fastest adoption of
18:12 technology in history they didn't think
18:14 it was a big
18:16 deal interface
18:19 matters and what I mean by that is what
18:21 is chat GPT if you haven't seen it
18:24 before that's kind of
18:27 it that
18:30 box it's kind of like the Google search
18:33 box right you don't have to pick a model
18:35 you don't have to know anything geeky
18:37 you just put words in there well what
18:38 words you put in there you can put
18:40 anything in there you can ask it
18:41 anything you can even ask it what's so
18:43 special about chat
18:45 GPT in an Eminem
18:48 rap and it will write you an Eminem WAP
18:51 talking about how awesome it is this was
18:54 one of the first things I did this what
18:55 I like to call parlor tricks one of the
18:57 great ways to play with chaty do parlor
18:59 tricks just ask it stupid [ __ ] and it
19:02 will do it I won't rap for you I
19:06 will
19:09 yes
19:12 um and and and then I thought huh this
19:15 this is remarkable so I was starting to
19:17 see in language what I was experiencing
19:19 with those pictures what's fascinating
19:21 is the underlying technology for the
19:23 pictures and the music and the video and
19:25 the words it's all the same thing it's
19:27 all based on this Transformer
19:28 architecture you just basically train it
19:30 on different data and it gets good at
19:32 different things it's just
19:34 stunning um so I started this group the
19:37 week after chat jpt came out called the
19:39 AI Salon I'll tell you about it
19:40 afterwards I'd encourage you all to join
19:42 if if you're interested if you want to
19:44 join a bunch of weirdos that are really
19:45 into AI um so I started this this thing
19:49 and I thought well I I got to come up
19:50 with a a plan for how to let people know
19:52 so I said hey chat gbt can you give me a
19:55 social media marketing plan and boom
19:58 there it came just knocked out and and
20:00 one of the one of the suggestions was
20:02 well you do paid media and I'm like no
20:04 no I'm broke I no I no I don't have
20:06 money for paid media I said but could
20:08 you give me a table that compares earned
20:11 media and paid media and boom it
20:14 formatted and did the analysis comparing
20:16 paid media and earned earned media I was
20:18 like oh that's cool and then I wrote an
20:20 email and it looked like [ __ ] and I was
20:21 like I don't remember how to do HTML
20:23 anymore I had one of the largest digital
20:25 agencies in the world and HTML doesn't
20:27 it's not the same so I said hey chbt can
20:30 you make my uh email look prettier in
20:33 HTML that would be good for a service
20:36 and it wrote the
20:38 code and I just kept
20:40 having like mindblowing experience after
20:45 mind-blowing experience after
20:46 mind-blowing experience I'll I I'll tell
20:48 you the thing that that did it for me so
20:51 I so it did this HTML that I that I
20:53 could understand but for years people
20:56 said oh you should learn
20:57 Python and and uh and I I was like okay
21:01 I can learn python but I have ADD I
21:03 can't really program I don't have the
21:04 attention spam for it so I said hey chat
21:07 GPT can you write me a python script
21:10 that that does I forget whatever it just
21:12 like takes an input and so it wrote this
21:14 python code I'm like that looks great
21:17 and then I had this this um kind of
21:20 shattering moment where I realized I
21:22 don't know how to run python code like I
21:24 don't even know how to run the code at
21:26 wrote but I said hey chat gbt
21:29 how do you run python code it said well
21:31 you open up terminal and you put in this
21:32 command line I'm like no I don't want
21:33 command line like isn't there a website
21:35 I can just copy and paste this it was
21:36 like oh sure here's five and the first
21:39 one was repet which is kind of the
21:40 famous one and I copied the code and I
21:42 pasted it into repet and it told me to
21:44 hit the play button and I hit the play
21:46 button and it gave me an error and I'm
21:48 like oh here we go again like with the
21:49 nine failures of stable
21:51 diffusion and I said it gave me this
21:53 error and chat jpt said this it said in
21:57 the lower left hand corner of the screen
21:58 click on this button in the upper right
22:00 hand corner of the screen look for this
22:02 library and hit install and I did it and
22:05 I hit the play button and it worked
22:06 within 90 minutes I had written a python
22:10 application fully functional python
22:12 application that took inputs from a user
22:14 and generated five pieces of social
22:16 media content I was doing API calls to
22:19 open AI in a python application I had
22:22 written in 90 minutes absentmindedly
22:24 while my wife's watching some shitty TV
22:28 show I've never written a line of python
22:30 in my life I still
22:32 haven't in that moment I had that same
22:37 like just [ __ ] electric thing from
22:40 Time Square oh my God the world just
22:43 changed so this stuff isn't just fancy
22:50 Google the G and GPT is really important
22:54 it's
22:57 generating I like to think of
23:04 um Google and and Technology up to this
23:07 point the internet up to this point has
23:10 been information on demand you ask it
23:12 for information it gives you information
23:14 but then you as a human still have to
23:16 learn the
23:17 [ __ ] AI is knowledge on
23:21 demand we we all if we're willing to
23:25 learn this stuff and quite frankly I
23:27 think we're going to have to
23:30 um I don't think it's
23:31 optional um we get to level up skills we
23:36 never imagined we' have we we had it's
23:39 it's it's staggeringly
23:42 powerful so one of the other things so
23:44 I've got a company my company's called
23:45 Story vine and we've got a video
23:47 storytelling platform and I and and one
23:50 of our Hallmarks is like authenticity
23:52 like authentic humans telling stories
23:54 that's what our our software does people
23:56 answer questions in an app it goes up to
23:58 the cloud 5 later you got a fully edited
24:00 video and I thought well I should at
24:02 least see where the state-ofthe-art is
24:04 for not that like I should see like what
24:07 am I competing with so and we were
24:11 talking before about uh personas and
24:13 audience segmentation and stuff like
24:15 that I thought I thought what if I could
24:17 could how would I do this like how would
24:20 I get you know one of these tools to
24:23 tell a story and so what I started with
24:26 is I said hey chat gbt and I don't say h
24:30 gbt uh but but I said um help me come up
24:33 with a Persona framework and and like
24:35 you know what's someone's educational
24:37 background and uh societal background
24:40 and what's their health condition and
24:42 things like that and then I had it
24:44 generate 12 personas generate 12 people
24:47 with their names their age their whole
24:49 history
24:51 and I then had each of those people
24:54 answer a set of Story vine prompts the
24:56 way our software works we write a set of
24:58 prompts that humans answer and I had
25:01 these personas answer those prompts what
25:03 what blew me away is that the stories
25:06 that they told were very familiar to me
25:08 it sounded like what humans would tell
25:11 because it knew who these people were
25:14 and then um I went to Mid journey and I
25:16 created um a
25:19 photo and I thought well I can take the
25:21 the script of what they said in text and
25:24 I can put that in a voice synthesis
25:26 thing and I can take that photo and I
25:28 can make that into a video and then I
25:31 could take those assets and put them in
25:32 the story Vine
25:34 system and I could
25:37 probably wait hang on I want to change
25:41 this so the
25:43 speakers are
25:47 here do
25:52 this hello I'm Amina a I was diagnosed
25:56 with illness about 4 years ago
25:59 what brings me joy is spending time with
26:02 my family especially my parents and
26:05 siblings we enjoy cooking and eating
26:08 traditional South Asian dishes together
26:11 moving beyond my diagnosis was not easy
26:14 but I found comfort in my faith and my
26:16 family's support so you don't need to
26:20 see the whole thing this was from
26:23 February of 2023 so this was more than a
26:27 year ago that that this te technology
26:29 existed
26:31 um I was really kind of excited about
26:34 this cuz I was like wow that's like this
26:36 could serve sometimes our clients can't
26:38 find people to tell their stories and
26:40 there might be something here I got
26:43 really excited about it and I showed my
26:44 staff and I almost had one person quit I
26:48 had three people like the person that
26:50 almost quit said um hey I thought we
26:53 cared about humans
26:55 here yeah and that that put me down for
26:58 a while I was like oh [ __ ] right so I
27:00 kind of I kind of let this go
27:04 um but but this was February 2023 right
27:07 so with a little bit of creativity duct
27:09 taping a couple of tools together you
27:11 can do something that's quite remarkable
27:13 um this is a project that I just
27:15 launched last week um and it's it's
27:20 um if if you uh use the $20 a month
27:24 version of chat GPT chat GPT plus you
27:26 can create these things called gpts
27:29 which are customized versions of chat
27:31 GPT and so I'm the chief generative
27:34 officer for this marketing organization
27:36 called content Evolution it's a it's a
27:38 Federation of a bunch of different
27:39 marketing companies and so we have like
27:41 a little AI Skunk Works and what we
27:44 decided to do was could we create
27:47 virtual twins of ourselves and so we use
27:50 this GPT technology which by the way is
27:52 remarkably easy like remarkably easy
27:55 like in the second half of the workshop
27:57 tonight we could even make one if you
27:58 want there's lots of things we can do
27:59 tonight um but what we did was was we
28:03 decided we we used chat GPT to help us
28:06 create a very structured set of
28:09 interview questions like 25 interview
28:11 questions that talked about work history
28:13 and personal history and cultural
28:15 influences and I had know background and
28:17 just all that sort of stuff and we did
28:19 video interviews of these 11 people and
28:21 then we took the transcript of that
28:23 video interview and it becomes the
28:25 knowledge base for each of these gpts so
28:28 so every one of these you can go on on
28:30 the internet right now if you have
28:32 access to to
28:33 gp4 um and you can have virtual
28:36 conversations with a PR Specialist or a
28:39 futurist or myself talking about AI it's
28:41 it's really quite remarkable um another
28:44 thing that happened in the past I don't
28:46 know four or five months
28:49 is the the the technology that allowed
28:52 me to do the Amina video um you can now
28:55 do that with yourself um and you can do
28:57 it in different languages so here's me
29:00 uh speaking fluid Italian so we've heard
29:02 from a number of clients that modular
29:04 content is important to them getting
29:06 more efficient with how we create
29:07 content and how we leverage
29:12 [Music]
29:16 content uh behind me you see these
29:18 little green boxes um that's how we
29:20 think about stories and story B we break
29:22 stories down into little atoms right
29:24 little
29:27 modules
29:35 guess you didn't know I spoke Mand
29:37 [Music]
29:51 [Music]
29:52 either uh and and you're probably
29:56 thinking yeah but it's probably not good
30:02 right here's Kyle talking Mandarin using
30:05 AI what he wanted to know is what does a
30:08 native speaker thing let's find
30:13 out that was yeah Mandarin is really
30:18 good it's talking about the AI story
30:20 something my Mandarin is really good
30:22 well it was AI yeah it was AI it was AI
30:25 it was AI yeah it wasn't him oh he AI
30:33 yes
30:36 right what platform did you create that
30:38 create that on what's that what platform
30:40 did you create that on hen hen h h e y g
30:44 e
30:45 n
30:48 um
30:51 so so
30:54 question how many people are super
30:56 excited right now
30:59 how many people are super freaked out
31:00 right
31:01 now how many are
31:03 both yeah that that's I think that's the
31:07 appropriate place to be um so a couple
31:10 of
31:13 things the world has
31:16 changed we just don't know it yet most
31:19 people don't know it yet McKenzie did a
31:22 report um they they they took two
31:25 measurements one was um February of 201
31:28 3 and one was November of 23 in February
31:32 of
31:32 2023 87% of people had not used chat
31:37 GPT in November it was
31:41 82% 82% of people have not used chat
31:46 GPT that means 18% have and I promise
31:50 you this the 18% that have a bunch of
31:54 them probably used it once and went oh
31:56 it's like Google yeah I'm good
32:00 of that 18% maybe half have used a
32:04 decent
32:05 amount so so now we're down to 9% of
32:08 that
32:08 9% maybe half have bought chat GPT plus
32:13 which is gp4 the more powerful model and
32:16 then of the people that bought it maybe
32:18 half of those are using it regularly so
32:21 right now somewhere in the neighborhood
32:23 of 2 to 5% of people know what the [ __ ]
32:27 is going on
32:29 and 95% don't and it's not like this is
32:33 coming like this is here I made these
32:35 translation videos like four months
32:38 ago and these are getting better I saw
32:40 there's a thing today if you want to go
32:42 see it called Hume AI
32:44 hum. where it's a it's an emotional
32:47 conversational chatbot where when you
32:50 talk to it it measures your emotion in
32:53 your voice and it shows you which
32:54 emotions are in there and then it
32:56 responds appropriately with its own emot
32:59 itions so like this stuff is coming and
33:02 it's coming fast now I don't mean to
33:04 freak you out um so here's what we're
33:07 going to
33:09 do oh I I just wanted to share this if
33:12 if you're really freaked out if you're
33:14 thinking like oh God they're coming for
33:15 our
33:16 jobs um a lot of jobs are going to
33:19 change but they're going to change
33:20 they're not going to go away some will
33:22 go
33:23 away um but but this is a famous quote
33:26 from Paul de ro from the 1840s where
33:29 when he saw the first photograph the
33:31 first deera type he said from today
33:33 painting is
33:35 dead painting isn't
33:38 dead every single time every single time
33:42 there's a new technology there's a
33:44 revolt and the more powerful the
33:46 technology the bigger the
33:48 Revolt you know there was when when the
33:51 the weaving loom came out there was a
33:53 mass there were riots in San Francisco
33:56 it was going to take jobs right when the
33:57 steam engine came out the farmers all
33:59 said what are we going to do well you
34:01 know we can't be replaced by these
34:02 things well they moved to cities and
34:03 they created an entirely new kind of you
34:06 know avocation
34:09 so the robots are not coming for us
34:12 these are tools we get to use and so for
34:16 the second half of this what I'm going
34:18 to do is let's take a let's take a five
34:20 minute break and go grab another drink
34:23 go to the bathroom um and then when we
34:26 come back um I've got sort of four
34:28 tracks that we can run down and I'm
34:30 going to let you guys decide which one
34:31 we do I'll try to get to more than one
34:33 of them but I don't think there's
34:35 there's just so much so when you get
34:37 back um let's say it's uh it's 10: after
34:40 right now so at 8:15 come back grab your
34:42 seats um and then uh and then we'll
34:45 we'll we'll move on and do the second
34:47 half of the of the presentation and I'm
34:49 happy to answer any questions if anyone
34:50 has
34:53 any take a
34:57 break
35:00 [Music]
35:05 hi
35:21 everybody there's there's a if you watch
35:24 it carefully when my hand goes in front
35:27 of my face goes under my face yes it
35:30 goes up under here yes so so what what
35:32 they're doing is they're they're they're
35:34 3D mapping face
35:43 [Music]
35:48 [Music]
35:52 and biting
35:57 [ __ ]
36:04 [Music]
36:06 yes but it's like you can say the same
36:09 thing for social media the for me social
36:12 media is just
36:14 smaller right so like the invention was
36:17 the world the shift was the shift so the
36:20 shift of PCS was the democratization so
36:24 so all of these are the democratization
36:26 right the democratization of super
36:28 computers with
36:29 personation of this internet the demra
36:33 this
36:57 is
37:02 also super curious about
37:08 [Music]
37:11 it a lot a
37:21 lot question Shar
37:24 the might be to check out but
37:29 basically out there probably a little
37:31 bit smaller than you guys
37:34 especially but
37:45 um if you're if you're still in the
37:48 digital space it's a really cool what's
37:51 it called buau of and um I'm having to
37:54 talk to you about it
37:56 later and um
37:58 yeah I
38:06 actually ideas I'm expert
38:26 that I
38:31 to
38:56 learn
38:59 well I don't
39:04 have
39:17 I remember I went to this
39:23 conference and all of the all the
39:56 inter
40:03 [Music]
40:17 [Music]
40:26 for
40:37 [Music]
40:55 all right we're getting started again
40:56 you weirdos
40:58 is Kyle running it from there or is he
41:00 ignoring us I'm ignoring
41:05 you hi I can't talk to you
41:09 people let's take our seats again
41:13 people artificial intelligence is very
41:16 important this is serious sit
41:21 down
41:25 yeah you don't have to sit down you can
41:27 stay still
41:31 um okay so and and the only reason I
41:34 want to I want to tighten this up is
41:39 um there's so much to talk about there's
41:42 there's so
41:44 much um if you haven't guessed it I'm an
41:48 optimist about this
41:50 stuff
41:56 um I think it's very easy right now to
41:58 demonize or dismiss
42:00 AI cuz Hollywood's told us for the past
42:03 50 years the robots are going to kill us
42:06 the tools right now are janky as I say
42:08 on my Tik Tok Channel they're janky
42:10 pieces of [ __ ] they hallucinate which
42:13 basically means they lie they're
42:16 inaccurate they're really hard to get to
42:18 do things consistently well they're very
42:23 frustrating
42:26 um
42:28 I am more creatively
42:31 alive than I've been since I was a
42:34 kid it's it's one of the most inspiring
42:38 like awe
42:40 inspiring experiences I've ever had in
42:43 my
42:43 life and it happens like multiple times
42:46 a week that
42:49 feeling one of
42:54 the AI needs a good PR firm cuz they
42:59 have a shitty PR firm right
43:01 now the fact that they call them chat
43:04 Bots right like we call it a chat bot so
43:07 it's a bot and what are Bots Bots are
43:09 cold and
43:11 inhuman my experience in interacting
43:14 with these tools is not
43:16 that if you understand how they
43:20 work they're a calculator it's a
43:23 probability engine it's literally the
43:25 way they work is you type a prompt and
43:28 it creates a probabilistic waiting and
43:32 it looks into thousand dimensional
43:34 mathematical space and pulls a token out
43:37 that represents a word or a period and
43:40 plops that on your screen and it does
43:42 that over and over and over again it
43:44 literally doesn't know what it
43:46 wrote but your experience in working
43:49 with
43:50 them like I
43:53 can't I cannot not say please and thank
43:56 you when I interact with chat
43:58 GPT
44:00 because confession
44:03 amen when it responds to you it's like
44:06 working with a really polite coworker
44:09 that never gets pissed off at you and is
44:11 always supportive of your ideas it's
44:13 like having a Golden Retriever as an
44:16 assistant that just happens to know all
44:18 of the [ __ ] in the
44:21 world great idea boss I'm like oh yeah
44:24 that was a good
44:25 idea so
44:29 I'm going to give yall some options we
44:32 can go down four
44:34 tracks one is how do these tools
44:37 actually work that thousand dimensional
44:39 space I talked about it's not as boring
44:42 as it sounds and actually once you
44:44 understand how they work I think it's
44:45 actually really valuable in using them
44:48 because you understand things like why
44:50 do they hallucinate why do they not why
44:52 are they not able to answer me sometimes
44:53 so we can do
44:56 that how how many people in here have a
44:58 chat GPT Plus subscription one two 3
45:02 okay how many of you have used more than
45:05 just the chat part of it okay to yeah
45:08 okay good decent um one of the other
45:11 things we could do is a chat GPT Deep
45:14 dive it is you will be blown away
45:18 because people ask me all the time is it
45:20 really worth the 20 bucks I think it's
45:22 the best $20 you could spend in the
45:24 history of software but you know I don't
45:26 feel that strongly about it so if you
45:28 want to see that we can do
45:34 that it is stupid right now to learn
45:38 anything about AI because three months
45:40 from now everything will have
45:42 changed but you should absolutely be
45:45 getting AI literate so this track is
45:48 about how do you get AI
45:50 literate knowing that things are
45:53 changing so fast like the next 3 years
45:55 is it it is it it is overwhelming on on
46:00 levels that are hard to describe but I I
46:02 we we can we can go down a path of what
46:04 I think is the way to stay aii literate
46:07 and not drive yourself crazy and then
46:09 the fourth one is I get this question a
46:10 lot too what's the best AI
46:13 tool and so we can go down that route
46:15 and I can tell you the best AI tool or
46:19 dozen of them so I guess we'll do it
46:23 like this look at that get the one you
46:25 want and then rather rather than vote
46:28 with voice or anything I just want You'
46:29 all hold up your hand with one two three
46:32 or four fingers up for which path you'd
46:35 like to go down can we do two paths can
46:38 we do two paths yes go ahead so I'd like
46:41 to go how I works the best AI tools okay
46:45 how AI works best AI tools what else
46:48 getting AI late okay we need to do a
46:50 vote okay so Okay so we've each got two
46:53 hands so so so put up uh two hands and
46:56 and put your first and second uh thing
46:59 up
47:04 okay I got a lot of
47:07 twos oh God this is bad okay we can't do
47:10 it this way that was that
47:13 was uh everybody raise their hand for
47:15 how AI which one that you want the
47:18 classic voting style raise your hand for
47:20 the one that you want to vote okay how
47:22 AI
47:23 Works can we ask AI what
47:26 we uh chat gbt Deep
47:29 dive okay that's decent getting AI
47:35 literate okay it almost it really
47:37 doesn't matter it's it's uh
47:40 okay
47:44 best best tool best tool oh best tool
47:46 but but the the people that are left
47:48 like it's it it's essentially a four-way
47:51 split so I'm going to choose
47:53 very okay I'm going to start with how AI
47:56 works so I said this before AI isn't
47:59 copying and pasting one of the things
48:00 that you'll hear a lot is you know AI is
48:02 the world's greatest plagiarism
48:05 machine it's stealing from artist it's
48:08 stealing
48:09 um it's a very very very complicated
48:14 gray area that will not be resolved
48:17 probably In Our
48:20 Lifetime all of these
48:22 models for the most part Adobe is an
48:25 exception actually adobe's Firefly was
48:28 not trained on on copywritten work but
48:31 all of the large language models most of
48:33 the image models most of the music
48:35 models most most of the most most of the
48:38 things have been trained on um work
48:40 without people's consent but how it
48:42 actually creates the content is not
48:45 copying and pasting when you ingest a
48:48 document into one of these large
48:50 language models and embed it what it
48:52 actually does is it converts words into
48:54 what are called tokens which is just
48:56 like a string of numbers numbers um and
49:00 and it doesn't even know what words are
49:02 in there it's it's just a mathematical
49:04 space so that's just an important
49:07 concept to know it doesn't doesn't
49:10 really make a difference but when you
49:12 hear someone say oh it's plagiarizing
49:14 it's not the G in GPT stands for
49:17 generative it's generating it's
49:19 generating original word so how does it
49:22 do it generative pre-trained Transformer
49:24 again generative means it's generating
49:26 pre-trained means it was pre-trained on
49:28 all of the internet you don't need to
49:29 bring your own data you used to have to
49:31 that's why you had to be a scientist and
49:33 then Transformers is the technology you
49:35 might hear llm which stands for large
49:37 language
49:39 model then if you geek out a little bit
49:42 if you're technical you might hear this
49:44 thing called the latent space and the
49:46 latent space is a mathematical
49:49 representation representation of word
49:51 clustered by meaning so this is a
49:54 two-dimensional visualization of latent
49:58 space every one of those clusters is
50:00 clusters of of of these um these uh
50:05 token or these uh wait tokens no these
50:08 vectors uh so like you have an XY Vector
50:11 in a two in a 2d thing you have an XY
50:13 Vector right and the Clusters are
50:16 semantically grouped words so you might
50:19 have the word dog in the light blue
50:23 section and dog in that case means the
50:26 The Fluffy
50:27 animal you have in your house and then
50:29 you might have dog in the purple section
50:31 which means he did his girl bad and he's
50:33 a dog right so so you can have the same
50:35 word have different meanings the real
50:38 latent space is like a thousand
50:40 Dimensions like we can't conceive it
50:42 it's insanely complicated but here's how
50:44 it
50:45 works when you put in a prompt the the
50:47 large language model predicts the next
50:49 token or the next word just to simplify
50:52 it over and over and over again and it
50:54 kind of works like this if we type in
50:56 the prompt the quick brown fox jumps
50:58 over the lazy blank as humans our large
51:01 language model goes dog we know that
51:03 it's dog so how did our brain do that
51:05 well how this brain does that is it does
51:09 this
51:10 [Music]
51:11 that mathematical waiting basically says
51:14 oh the the the word at the end of that
51:17 sentence the high it's got the highest
51:19 likelihood of being in this cluster and
51:21 then within that cluster of vectors one
51:25 of them has the highest probability of
51:27 being the right word and what it
51:29 actually looks like is like a big long
51:32 ass number that is then associated with
51:35 the word dog and then it puts the word
51:37 dog into your sentence so when you see
51:40 chat GPT write a whole essay for you
51:44 it's doing it like this one token or a
51:47 little cluster of tokens at a time it
51:49 doesn't know what it wrote so when you
51:52 ask it something and it gives you the
51:54 wrong answer and you're like hey dum dum
51:56 you gave me the wrong answer give me the
51:57 right answer and it's like oh sorry
51:58 about that boss here's the right answer
52:00 and it's wrong again and and what we're
52:03 used to is computers that are
52:06 probabilistic computers that are
52:09 predictable if I ask it something it
52:11 gives me the right answer that's not
52:13 what these things are doing the G
52:15 generative it's generating
52:19 stuff the when Sam Alton got fired
52:21 there's a thing coming called
52:24 qar that is a reasoning engine
52:27 that we think is going to make these
52:30 things really powerful where what what
52:32 it will do in the future is it won't
52:33 just vomit out words it will vomit out
52:36 words and then it will use reasoning to
52:38 say are these the right words and then
52:40 it will fix them and then give you
52:41 things so the problem of Hallucination
52:44 is going to get better and better and
52:46 better and better and it's probably
52:47 going to happen pretty quickly but
52:48 anyway that's how they work isn't that
52:50 fascinating I think okay absolutely How
52:53 do images work this is even more okay
52:57 that there are mathematicians on the
52:59 planet that figured this stuff out like
53:01 holy
53:02 crap we get to play with this
53:06 stuff do you get that
53:11 okay the way the image models work is
53:14 you take your large language model and
53:16 you pair that with with what's called a
53:17 diffusion
53:19 model and
53:21 then the images that it are it's trained
53:24 on they take a noising algorithm and
53:26 they turn all of the images into random
53:29 noise images using an
53:31 algorithm and then that Vector space
53:33 that I talked about just basically
53:35 points to a bunch of unrecognizable
53:38 noisy images so when you give it a
53:40 prompt the large language model does
53:42 that same thing with the vectors and it
53:44 says if if our prompt is Cat on a desk
53:48 it sort of says okay here are a bunch of
53:50 here's a cluster of pictures that have
53:52 been labeled cat here's a cluster of
53:55 pictures that have been labeled desk
53:57 and then it takes these random noise
54:00 images of cats and
54:03 desks and it averages them into a random
54:08 noise image and then it uses the same
54:11 algorithm that made them noisy to
54:13 denoise
54:17 them and so it
54:21 creates what you think is a copy pasted
54:23 image but it's a completely original
54:25 image that was sort of literally like
54:28 materialized out of fairy dust out of
54:31 like probably thousands of images that
54:32 are no longer images it's just this
54:35 noise noise pattern that get denoised
54:37 and turned into this stuff so again when
54:39 you understand how it works it's like
54:41 wait how do these things work that what
54:43 are they
54:44 doing all right so that's that questions
54:48 yes what's noise what's noise noise is
54:51 just like you know when you're well no
54:54 you're not old enough how old are you 34
54:58 yeah right on the border do you remember
55:00 Olden timey TVs when when it went off
55:02 air it would
55:03 go yeah that's noise so
55:07 so it basically takes the image and just
55:10 noises it it makes it noisy so there's
55:13 an algorithm where it it takes it from
55:16 looking like this and it goes backwards
55:19 and then in the database are just a
55:21 bunch of those things and and the
55:24 algorithm that that made it that noisy
55:26 it then takes the average thing and
55:28 turns it into that yeah and why does the
55:31 noise have to be
55:33 added because that's what the
55:34 mathematicians came up with I be
55:39 because be oh for efficiency
55:42 if if if you were actually copying and
55:45 pasting these like the amount of data
55:46 processing you'd have to do would be
55:48 massive like it's already massive so I
55:50 think this just makes it more simple I
55:52 don't I honestly don't know good
55:55 question okay
55:57 any other questions you said AI is not
56:00 an it I said AI is not an it is not
56:06 [Music]
56:11 an no AI is a is
56:15 a no I said
56:18 it what it is is like a calculator it it
56:22 is an it but interacting with it doesn't
56:24 feel like that because the data it was
56:26 tramed
56:27 on was it's not the bot is not the but
56:31 yeah you're not interacting with the bot
56:32 you're interacting with the sort of
56:34 collective output of humanity that's
56:36 been dumped on the public internet for
56:37 the past 50 years and so what it feels
56:40 like emotionally to
56:43 me is like I'm
56:46 collaborating I'll tell you what's
56:48 what's Wild For
56:51 Me Co so I was a I was an absolute
56:54 non-believer in remote work I [ __ ]
56:57 hate remote work I love people I get
56:59 energy from people I work in a company
57:02 my company is full of introverts when
57:03 when the lockdown happened they're like
57:05 woo we're living our best life we hate
57:06 people screw you
57:08 up and and you know I I was like I I
57:12 like people I I like when I interact
57:15 with chat GPT what what Co taught us was
57:18 how to work remotely and what it feels
57:21 like is like I'm working with a remote
57:22 worker who's really polite incredibly
57:26 competent and
57:28 clueless like like you know like
57:30 clueless interns that are like really
57:31 smart like they went to the I leag
57:33 League school and you bring them into
57:34 your agency and then they're like
57:36 they're like you like go write the thing
57:38 and they write it you're like this is
57:39 horrible they're like oh right they
57:40 don't know basic World stuff right
57:43 that's how these things are but but
57:46 interacting with it is it's not a human
57:48 experience but it is way more human it
57:51 doesn't feel like you're interacting
57:52 with a robot I think calling it a bot is
57:54 wrong not wrong but I I it's
57:57 inaccurate um
58:03 okay yeah I think chat GPT I'm I'm
58:06 between chat GPT deep okay let's just do
58:09 two uh chat GPT Deep dive versus getting
58:12 AI litered I think those are the most
58:13 important ones actually so raise your
58:15 hand chat GPT Deep dive and then raise
58:19 your hand getting AI
58:20 literate what do you mean by AI what
58:23 will they count knowing what AI is well
58:26 enough that you could talk about at
58:27 another
58:28 party deep deep deep dive well I'll see
58:34 if we can do
58:35 both all right we're gonna we'll start
58:37 out with chat TBT Deep dive
58:42 okay
58:44 um the
58:46 juggler I used to I used to be a
58:50 juggler
58:53 um all the things in the left
58:57 basic chat Advanced chat voice chat
58:59 custom
59:01 instructions um dolly3 image generation
59:03 code interpreter custom
59:06 gpts all of those are things that chat
59:09 GPT
59:11 does um most people don't know that so
59:16 here's what we're going to do who's
59:18 got how
59:24 we I my eyes aren't quite good enough
59:26 you and I were talking about customer
59:27 segmentation
59:29 right or or something like that yes give
59:32 me give me give me a project give me
59:34 something give me that project that we
59:36 want to do and we'll go play with chat
59:38 GPT we we'll sort of exploit a bunch of
59:42 these different things so can really
59:46 make anything like we can come any
59:48 audience like how would they want to
59:51 celebrate a holiday or how would they
59:53 want to
59:55 commemorate
59:58 okay so so we want to Define an audience
1:00:01 and then we want to and then we want to
1:00:03 be able to ask chat GPT how that
1:00:06 audience would want to be what
1:00:08 communicated with could ask how they
1:00:10 feel create an audience and ask how they
1:00:12 feel about AI okay how okay how they
1:00:14 feel about AI okay great okay so so let
1:00:17 me let me give you a little bit of um
1:00:21 let me or Orient you a little bit this
1:00:23 is chat GPT um down at the bottom here
1:00:26 is where you type stuff in um make me a
1:00:30 poem about poem Cheetos it doesn't
1:00:34 matter you can typos are literally
1:00:36 irrelevant it understands everything
1:00:41 wow uh do I have the internet turned
1:00:45 on we
1:00:47 do oh we might we might not be doing
1:00:50 chat GPT if it's a little it might be a
1:00:53 little sleepy let's see let me reload
1:00:55 the page
1:00:58 make me a
1:01:00 poem all right I won't do Cheetos about
1:01:06 Fritos in the heart of the bag where the
1:01:08 salty winds blow lies a corn so bold
1:01:11 where the true snackers go golden
1:01:13 Treasures crisp in Rose they lay
1:01:16 Whispering Tales of crunch in the light
1:01:18 of
1:01:19 day so that's
1:01:21 chat um this is this is what I call
1:01:24 parlor tricks uh you could also do
1:01:27 things like um write me a
1:01:31 LinkedIn
1:01:34 article about the
1:01:37 same and format using
1:01:41 markdown um markdown I didn't know what
1:01:44 it was but it's a thing that makes
1:01:46 things pretty uh let's see to make it
1:01:52 easier to
1:01:54 use now a couple of things an article
1:01:57 about the same chbt understands the
1:02:01 context of what you talked about before
1:02:03 so it's got a memory um it's a really
1:02:06 short memory right now so sometimes
1:02:08 you'll be talking to it and things will
1:02:10 be going great and then it just kind of
1:02:11 loses its mind it's because it actually
1:02:13 lost its mind those are getting bigger
1:02:15 but right now that's just kind of a deal
1:02:17 um but
1:02:18 anyway so markdown made the big headline
1:02:22 and so here's our LinkedIn article well
1:02:24 formatted you liter literally just copy
1:02:26 and paste this over to a LinkedIn
1:02:28 article that brief history the golden
1:02:31 corn uh and then let's see um
1:02:36 make a table
1:02:39 comparing Cheetos and
1:02:43 Fritos
1:02:46 oops isn't this just a bunch of
1:02:51 words it is and it's
1:02:55 not there um is there value and taking
1:02:58 it from the perspective of like an
1:03:00 authority or an AB absolutely and that's
1:03:02 what we're going to do here in a second
1:03:05 so this is just basic chat GPT chat
1:03:07 now um and and by the way like
1:03:12 oops you know th this thing of just
1:03:15 compare two things do analysis comparing
1:03:18 this ridiculous thing with that
1:03:19 ridiculous
1:03:21 thing one of the things I absolutely
1:03:23 recommend when you're playing with AI St
1:03:26 is don't take it seriously play with it
1:03:30 ask it to compare gummy bears and Sushi
1:03:32 ask it to write a ridiculous poem ask it
1:03:35 to create a computer program that does
1:03:37 something stupid because in doing that
1:03:40 you'll sort of take yourself off the
1:03:42 hook for feeling like oh my god I've got
1:03:44 to learn something this is so hard I
1:03:46 don't this
1:03:47 hurts um and you'll you'll start to find
1:03:50 the boundaries of where this stuff is
1:03:53 good and where it's not good if you if
1:03:54 you try to just stay too locked into
1:03:57 what you know you won't actually learn
1:03:59 what these things make possible
1:04:02 um okay so so let me show you a couple
1:04:05 other things
1:04:07 here um we're in chat GPT 4 right now
1:04:10 I'm going to flip over to chat GPT
1:04:13 3.5 uh I'm going to do the same thing U
1:04:15 make me a poem about Cheetos this
1:04:19 time you'll notice this is much
1:04:23 faster so chat gbt 3.5 is the smaller
1:04:27 less powerful version but it's super
1:04:29 fast so if you really just want to kick
1:04:32 out ideas starting to understand what
1:04:35 the different tools are capable of is is
1:04:38 is a power move um and and I can I can
1:04:42 show you some of some some of that
1:04:46 um there's also in I'm trying to think
1:04:51 where I want to go
1:04:52 first let me flip back over here okay so
1:04:55 here's what we're going to do
1:04:56 we're going to start a new
1:04:58 chat and we're going to say
1:05:02 um give give me an audience give me an
1:05:05 audience that uh we're going to we're
1:05:07 going to learn
1:05:09 about I
1:05:11 mean be like a bunch of psychographics
1:05:14 combined with a bunch of demographics so
1:05:17 okay so wait so so here's what we'll say
1:05:19 I want you to act as a
1:05:25 um
1:05:28 population
1:05:30 specialist
1:05:32 analyst uh
1:05:35 analist and uh uh
1:05:42 consultant please
1:05:45 create
1:05:48 three um
1:05:51 audience
1:05:54 Persona Frameworks
1:05:57 M so I can
1:06:00 [Music]
1:06:02 choose one to start working
1:06:08 with now when I say act as a population
1:06:12 specialist analyst and consultant that
1:06:13 that was your question before does that
1:06:15 actually work here's why that works so
1:06:17 if we go back to the the Laten space
1:06:19 graph you've got all those clusters of
1:06:20 words right um oh I here here's a good
1:06:23 one this was Andre karpathy who who
1:06:25 works at open AI this this was an
1:06:26 example that he
1:06:27 used he said imagine you want to ask
1:06:30 about a math problem and you just say to
1:06:33 to chbt hey uh explain this math problem
1:06:37 to me and it will explain it and very
1:06:39 often it will hallucinate and get it
1:06:41 wrong and the way he described it is in
1:06:45 in the latent space in in its knowledge
1:06:47 base is everything that's ever been
1:06:49 written about that math problem so
1:06:52 there's the professor at Stanford that
1:06:54 that wrote about that math problem and
1:06:55 then there all of the crappy students at
1:06:57 some local college that wrote about that
1:06:59 math problem that got it wrong so when
1:07:01 you just ask about the math problem
1:07:02 it'll sort of
1:07:04 confabulate good stuff with bad stuff if
1:07:07 you say hey I want you to act as a
1:07:08 Stamford mathematician that specializes
1:07:11 in this branch of mathematics and I want
1:07:12 you to tell me about that problem you're
1:07:14 narrowing you're cutting out a bunch of
1:07:16 those other areas and so you're kind of
1:07:18 narrowing your your your uh the the sort
1:07:21 of knowledge base that you're going
1:07:23 that's why when you when you treat chat
1:07:25 PT kind of like Google you just give it
1:07:27 a quick little generic search it gives
1:07:29 you shitty answers because it's sort of
1:07:31 amalgamating everything anyone's ever
1:07:33 said and they're
1:07:35 idiots so you need to find the areas of
1:07:37 Brilliance within within this massive
1:07:40 model okay
1:07:43 so so this thing's going to go off now
1:07:45 creating an audience of persona
1:07:46 Frameworks involves blah blah blah so
1:07:49 it's going to tell us this
1:07:51 and and it will so the demographic Focus
1:07:54 Persona pay ATT
1:07:56 I'm sorry what's your name again Julie
1:07:58 Julie so pay attention Julie all right
1:08:00 so we're going to go demographics is one
1:08:02 of
1:08:02 them
1:08:04 um oh
1:08:08 wait Persona name oh oh it's okay it's
1:08:13 it so it gave us the framework so
1:08:15 framework one is demographic focused and
1:08:17 then it's actually filling that out so
1:08:19 so here's Fiona okay this is good this
1:08:21 is
1:08:22 good I might stop this at some point
1:08:25 here cuz it's a little
1:08:27 slow by the way this this is very very
1:08:30 familiar back to the the' 70s um when
1:08:33 you had dial up modems this is how slow
1:08:35 they typed like you could watch the
1:08:37 internet typing across screen and so
1:08:40 we're going to look back at these times
1:08:41 and go hey remember when chat GP
1:08:43 actually could read it when it typed
1:08:45 that's not going to be here for long um
1:08:48 okay I'm bored anyone else board um okay
1:08:52 this is framework three okay so Behavior
1:08:55 fol
1:08:57 focused psychographic focused or
1:08:59 demographic focused do you prefer one
1:09:01 over the other I think psychographic
1:09:03 will just especially if we're ask
1:09:05 something about
1:09:07 like no I'm not good with that okay
1:09:10 here's what here's what we're gonna do
1:09:12 can you now create a single
1:09:19 framework that
1:09:21 combines all three and gives us uh
1:09:27 three
1:09:29 unique
1:09:30 personas based on
1:09:35 it so one of the things
1:09:39 that a good way to think about chat GPT
1:09:42 is not like a search engine you give it
1:09:44 a prompt that gives you an answer that's
1:09:46 one of the big misconceptions you give
1:09:47 it a prompt that gives you an answer oh
1:09:49 the AI wrote that no think of it more
1:09:52 like a collaborative conversation like a
1:09:54 jam session
1:09:56 and give it more information than you
1:09:58 think it needs the reason that I asked
1:10:00 it to give me three different P Persona
1:10:02 Frameworks I don't frankly give a [ __ ]
1:10:05 what it gave me I just want it to know
1:10:08 what we're talking about so by it giving
1:10:10 us that and me telling okay yeah those
1:10:12 are good now combine them into one it's
1:10:14 it's sort of I'm giving it increasingly
1:10:18 narrow uh a narrow window for this
1:10:21 conversation to live in so the more
1:10:23 context you give these things the better
1:10:24 they get yeah question for you I I use
1:10:27 this all day every day and I incorporate
1:10:30 a lot of additional references into my
1:10:32 promps as I start generating them but
1:10:35 the one thing I can't figure out and
1:10:36 this is such a weird question to ask is
1:10:38 the punctuation that you find is the
1:10:40 best way to separate the different
1:10:43 refences that you're including because
1:10:44 you know you get
1:10:46 obviously so the question is if if
1:10:49 you're if you're putting in references
1:10:51 like you say I want you to go look at
1:10:53 this article or you copy and paste you
1:10:55 know something from a different article
1:10:57 in in your prompt if you're putting
1:10:58 things in the prompt the question was
1:11:00 how do you how do you provide um punctu
1:11:03 what's the best way to add punctuation
1:11:04 as you add those different references in
1:11:07 general think of have you ever done any
1:11:09 programming at all have you ever done
1:11:10 any like like even basic programming
1:11:12 like engineering anything the more
1:11:15 structure you can give it the better so
1:11:18 so um you can do things like um
1:11:21 incorporate the following references and
1:11:23 then colon bullet point reference one
1:11:26 bullet point reference two bullet point
1:11:27 reference three um ignore these
1:11:30 references and then so so so just create
1:11:35 structure it likes structure so so and
1:11:38 like even if you do the label of like
1:11:40 you know include these references in all
1:11:42 caps with a colon and then bullet points
1:11:44 below it it'll really understand that um
1:11:47 also play around with order order
1:11:49 sometimes
1:11:50 matters um okay so now we've got these
1:11:52 things so we've got again I don't really
1:11:54 care who it is but we've got we got
1:11:56 Fiona we've got
1:11:59 Ella and then we've got Green Tech Greg
1:12:03 I think we know Greg's from Boulder
1:12:07 um okay what do we want to ask these
1:12:09 these
1:12:10 folks um I mean I think start how do you
1:12:14 feel
1:12:16 about oh okay
1:12:20 um
1:12:23 have
1:12:24 Greg
1:12:26 Ella and
1:12:31 Fiona did I spell that right no shh I'm
1:12:36 okay have Greg Ella and Fiona
1:12:45 um what about talk about how they would
1:12:47 use AI given their respect I'm trying to
1:12:50 do something I'm trying to just do
1:12:51 something
1:12:52 stupid um rock Musical well that's
1:12:56 that's where I was going to go let's see
1:12:58 uh write a
1:13:00 Hau about how they feel about
1:13:05 AI
1:13:11 yes Greg silence server's home Nature's
1:13:15 Balance in my palm teeken Earth has
1:13:18 one entreer Ella Infinite Canvas dream
1:13:21 sculptured sculpted in digital AI my
1:13:24 Ally family first Fiona AI cradles
1:13:27 cradle dreams a future woven with care
1:13:30 um let's see
1:13:33 um let's yeah let's uh make
1:13:39 Greg um anti
1:13:47 AI machines whisper lies Nature's truth
1:13:50 drowned by by circuits lost in Silicon
1:13:53 okay
1:13:54 perfect h
1:13:56 yeah you can do that too more extreme
1:13:57 okay perfect um now I want you to create
1:14:03 an image from each person about their
1:14:11 poem so chat GPT plus has what's called
1:14:14 di in it you can also um this di is also
1:14:18 inside Microsoft co-pilot which is free
1:14:21 so if you go to co- pilot.
1:14:24 microsoft.com that's a version of chat
1:14:26 GPT 4 that's free so if you can't afford
1:14:28 the 20 bucks a month go go to
1:14:30 microsoft.co pilot.com
1:14:34 um you can do the image generation as
1:14:37 well I still think it's worth paying the
1:14:39 20 bucks for chat GPT for a couple of
1:14:41 reasons it's just better and the the
1:14:47 um the the the way they've they they've
1:14:50 coded Microsoft What's called the safety
1:14:52 guard rails are really restrictive so if
1:14:54 you try to ask it about
1:14:56 something like you want to Market
1:14:57 tobacco or something like that it might
1:14:59 give you a oh tobacco can hurt people so
1:15:01 I'm not going to answer that it's it's a
1:15:03 real pain in the ass that way so is chat
1:15:05 GPT for for that matter um but it can
1:15:07 make images
1:15:11 um these things can
1:15:14 also
1:15:16 um chat GPT can also
1:15:19 see you're like it can see what what do
1:15:22 you mean it can see is it looking at you
1:15:24 right now oh my God hi
1:15:28 people say hi to my life people
1:15:32 wave
1:15:33 hi
1:15:35 hi I'm not ignoring you except I sort of
1:15:39 am oh oh
1:15:42 no all right I got lots got lots of
1:15:45 Hearts they liked
1:15:47 you
1:15:49 um okay so what did you you said what
1:15:53 were we going to ask them oh how would
1:15:54 they use AI
1:15:56 is that what you is that what you said
1:15:58 based on their Persona um how would each
1:16:02 of them use
1:16:05 AI um for
1:16:08 a bake
1:16:12 sale let's just see what it comes up
1:16:14 with and
1:16:16 and like purposefully asking it weird
1:16:20 questions that it shouldn't be able to
1:16:22 answer is is like one of the ways I
1:16:23 learn what these things can do is like
1:16:28 what it will try to do is it will try to
1:16:30 give you everything it will try to do it
1:16:32 and sometimes it fails miserably and
1:16:34 sometimes you're like oh my God that's
1:16:36 good right it totally inspires a new
1:16:39 thing so let's
1:16:40 see Greg anti- aai sourcing ingredients
1:16:44 he uses AI to find and Source locally
1:16:46 produced sustainable ingredients that's
1:16:47 good Greg energy efficient baking he
1:16:51 employs uh smart kitchen appliances that
1:16:54 use AI all right so he's using it on the
1:16:57 edges um might reluctantly use social
1:17:00 media analytics and identify local
1:17:02 community groups good good Greg it's
1:17:04 solid it's you know he hasn't given up
1:17:06 he's not he's not totally resisting it
1:17:09 Ella uses it to analyze food and Trends
1:17:12 um Inventory management okay so now
1:17:15 we're going to try something
1:17:17 cool I'm going to click the little
1:17:20 paperclip right here which you have to
1:17:24 have chat GPD to get the paper clip but
1:17:26 you can
1:17:28 upload images and it can
1:17:35 see let's
1:17:46 see is this it yes okay this is a
1:17:50 whiteboard I got off the internet I went
1:17:52 on the internet and I typed in
1:17:53 whiteboard and I found a decently
1:17:54 complic ated
1:17:56 whiteboard and I'm just going to say
1:17:58 what is
1:18:05 this the
1:18:08 suspense
1:18:11 okay appears to be a whiteboard of
1:18:13 handwritten notes and sketches the
1:18:14 content suggests it's a brainstorming or
1:18:17 planning for a user interface so it
1:18:20 could see that it there are sketches of
1:18:22 user interface
1:18:24 there uh
1:18:26 likely focused on ordering services such
1:18:28 as food and drinks perhaps in an airport
1:18:30 setting the out it it outlines
1:18:33 three-step main process choosing a
1:18:35 language ordering from a menu and
1:18:36 checking out which includes payment
1:18:39 there are notes about Target users
1:18:40 business and family so now we're going
1:18:42 to get back to our personas um use cases
1:18:45 take out within within a limit of an
1:18:47 hour possible okay so it's got all those
1:18:51 uh I'm going to say
1:18:52 describe the um
1:18:56 interface
1:18:58 screens let just do
1:19:14 that so there's a welcome screen a menu
1:19:17 screen it prompts the user to tap to
1:19:19 start choose a language so we can go up
1:19:21 and look at that but it it had like all
1:19:23 the little words in there it read all
1:19:24 the little words that understands what
1:19:25 the pictures
1:19:26 are
1:19:30 um one of the items listed is Caesar
1:19:32 salad with a price of $10 status screen
1:19:35 checkout screens okay great so now I'm
1:19:38 going to say
1:19:43 um
1:19:45 write
1:19:47 a set
1:19:50 of five marketing taglines
1:19:58 for
1:19:59 Ella Fiona and
1:20:04 Greg you can ignore the AI
1:20:12 stuff but
1:20:14 focus on their um
1:20:19 demo
1:20:22 psycho
1:20:24 graphic
1:20:29 profiles so again like it's you're not
1:20:33 you're not coding it's it's just like
1:20:36 like this is if you were working with
1:20:38 someone in the office and you were
1:20:39 explaining I want you to go do this work
1:20:41 you would kind of explain it like this
1:20:43 so if you just think of this interaction
1:20:45 as like you're working with a coworker
1:20:47 and you're just having this back and
1:20:48 forth interact action with them um for
1:20:52 entrepreneur Ella Empower your ambition
1:20:54 perfect lead with Elegance
1:20:55 sophistication meets Innovation for the
1:20:57 modern entrepreneur redefine your
1:20:59 workday craft your Empire beautiful for
1:21:02 Visionaries who demand excellence and
1:21:05 precision um for family first Fiona
1:21:08 cherish every moment family family
1:21:10 friendly picks um for the heart of the
1:21:12 home oh wait but it so hang on it missed
1:21:14 something here you
1:21:16 missed that this is a food
1:21:22 ordering app for and
1:21:27 Airport you dumb
1:21:30 [Laughter]
1:21:37 Dum and you probably shouldn't be rude
1:21:40 to the
1:21:43 Bots cuz they are going to have memory
1:21:45 soon enough my apologies let me try that
1:21:48 again again the way these things have
1:21:50 been programmed they've been programmed
1:21:51 to to they want to give you what you
1:21:54 want
1:21:55 um so let's see if we if we got it
1:21:57 better uh entrepreneur Ellis seamless
1:22:00 sustenance dine on demand between deals
1:22:03 good boardroom to boarding gate satisfy
1:22:05 your appetite for Success these are
1:22:07 great perfect okay great so let's say um
1:22:11 take off to table time okay so here's
1:22:12 what we want to do um make
1:22:17 me
1:22:20 three
1:22:22 concept
1:22:24 posters
1:22:25 in an
1:22:28 airport
1:22:29 setting for
1:22:32 each
1:22:35 Target uh Target
1:22:38 Target
1:22:40 um
1:22:43 include the name of the
1:22:46 app um fly
1:22:52 bites so there there I'm doing this is
1:22:54 you know little structure put a little
1:22:56 structure in the name of that fly bites
1:22:58 and um a
1:23:02 um image of a
1:23:06 mobile
1:23:10 device and the ordering
1:23:20 screen so so you can just keep going
1:23:23 between these modes right you can you
1:23:25 can say give me some Concepts give me
1:23:27 some taglines put that tagline in a
1:23:29 poster make an image make a this make a
1:23:31 that there's another thing called code
1:23:34 interpreter in here I'm not going to
1:23:35 demo it because it's going to take too
1:23:36 long and it's a bit boring but what code
1:23:38 interpreter does is you can upload a CSV
1:23:42 file of data so if if you're a little
1:23:45 geeky write this down write down the the
1:23:48 website kaggle k a g g l e and what
1:23:52 kaggle is is is a a site where you you
1:23:54 can go get data sets so you can get like
1:23:57 New York City Taxi data from 1972 I
1:24:00 don't know whatever crap like that you
1:24:02 just download CSV files you can upload a
1:24:04 CSV file to chat GPT plus the four and
1:24:08 say tell me about this data and it will
1:24:12 write its own python code and do the
1:24:16 analysis of the data in that
1:24:21 spreadsheet yeah what if the
1:24:23 dat then don't put it in there if the
1:24:26 data is sensitive don't put it in assume
1:24:28 everything right now in chat GPT if it
1:24:31 were to get out and you were to get
1:24:33 fired for that don't put that [ __ ] in
1:24:36 there um but anyway by the
1:24:40 way there's our there's our entrepreneur
1:24:43 Ella right and her fly bites app here's
1:24:46 the family right so like we just went
1:24:49 from like how do I interact with how
1:24:51 these personas are going to go to
1:24:54 flipping into cre development mode to
1:24:56 concepting things like when you get
1:24:59 Nimble with all of the different modes
1:25:01 in chat gbt you can literally like like
1:25:05 I it's like a it's like a [ __ ]
1:25:07 superpower you can literally just be
1:25:09 flipping between different mental modes
1:25:12 right you don't have to be locked into
1:25:14 I'm a copywriter I'm an illustrator I'm
1:25:17 a strategist I'm a this I'm a that play
1:25:20 push the boundaries and you'll start to
1:25:23 discover you'll you'll start to discover
1:25:26 that you're good at things you didn't
1:25:27 know you were good at you can do things
1:25:29 you didn't even think were
1:25:33 possible
1:25:34 um so anyway and then it describes it
1:25:37 I'm trying to think I'll show you one
1:25:39 other thing here um this you can do in
1:25:42 both chat GPT um normal and chat
1:25:46 GPT plus um under your
1:25:50 name by the way how you get to chat GPT
1:25:53 plus you click on your name you go to my
1:25:55 plan and it's the the Plus Account right
1:25:57 here it's 20 bucks a month um click on
1:26:00 your name there's this thing called
1:26:02 customized chat GPT this is called
1:26:04 custom instructions one of the things
1:26:06 you'll notice about chat GPT is it's a
1:26:08 little wordy it's a little
1:26:11 verose um and it can be really annoying
1:26:14 so what this allows you to do is give it
1:26:16 a
1:26:17 personality so I can say you are a
1:26:22 pompus s who
1:26:26 uses words so big even you don't know
1:26:31 what they
1:26:33 mean so the first thing you do is kind
1:26:35 of tell it who it is I spelled pompus
1:26:37 wrong Jesus all right oh pomposa you
1:26:44 p and then you say how would you like
1:26:47 chat GPT to respond
1:26:50 um
1:26:52 when someone uh let's see
1:26:55 no matter what someone asks
1:26:59 you
1:27:01 um
1:27:03 respond with an
1:27:07 obscure literary
1:27:11 reference and
1:27:15 then use as
1:27:19 many
1:27:21 complicated words as
1:27:23 possible
1:27:27 and so I'm just going to start this over
1:27:28 now so we have a brand new chat so I'm
1:27:30 just going to ask at something nice like
1:27:33 why is the sky
1:27:39 blue ah to contemplate the Azure canopy
1:27:42 above us is a tread the same path as the
1:27:45 protagonist in nao's speak
1:27:49 memory who marveled at the intricacies
1:27:51 of memory and perception the explanation
1:27:54 IAL Hues and
1:27:57 quarter right staggering now I can go
1:28:00 back in I could also you're like yeah
1:28:02 but well that's kind of useless Kyle
1:28:04 that's kind of ridiculous no one would
1:28:05 do that well I just did um you are
1:28:13 impatient and very
1:28:20 smart answer only with
1:28:31 with three concise bullet
1:28:36 points no matter the
1:28:43 [Music]
1:28:44 question why is the sky
1:28:52 blue now
1:28:55 so these are good parlor tricks let's
1:28:58 say that you've got a particular way you
1:28:59 write emails you could say you could put
1:29:03 in here hey I just want you to write an
1:29:05 email that follows this and you could
1:29:07 put an example of your writing in there
1:29:09 and say follow this example for tone
1:29:11 right you could say you could say I hate
1:29:14 the way this thing responds I want it to
1:29:16 give me you know more structured output
1:29:18 when it answerers so so anything like
1:29:20 that so that's a really powerful tool um
1:29:22 I'm going to show you one other thing
1:29:24 there's how are we doing on time we're
1:29:25 good
1:29:26 okay
1:29:30 there's starting nove starting in
1:29:33 November of last year um chat GPT rolled
1:29:36 out these things called
1:29:39 gpts so okay this is this is screwed up
1:29:42 I'm telling you AI needs a branding they
1:29:44 need a PR firm and they need a branding
1:29:46 agency so you've got
1:29:48 gp4 is the
1:29:50 model chat GPT is the application that
1:29:53 you chat with that sits on top of the
1:29:55 model gp4 chat GPT sometimes people call
1:29:58 it chat GPT or chat GPT 4 but gp4 is the
1:30:03 model chat GPT is the app and then gpts
1:30:06 sit on top of that and are customized
1:30:08 versions of chat GPT that do a specific
1:30:11 thing yeah and they so have you ever
1:30:14 heard of like no code tools and and like
1:30:17 automation tools like zappier and make
1:30:19 and things like
1:30:20 that what they promise is that you don't
1:30:24 need be a programmer to do application
1:30:26 development and what the reality is
1:30:28 [ __ ] you don't actually have to
1:30:30 write the code but you're still
1:30:31 programming you have to understand what
1:30:33 apis are and what API keys are and you
1:30:35 have to connect the nodes and hit the
1:30:38 variables in the formula but it's it
1:30:40 [ __ ] awful okay it's fine but it's
1:30:44 it's not no code um gpts are truly no
1:30:47 code so what they are and I'll show you
1:30:50 the very F very first one I ever made I
1:30:52 was live on Tik Tok and and these things
1:30:55 came available I said ooh let's go try
1:30:57 one and So within 5 minutes of me saying
1:31:00 ooh let's go try one this is what I had
1:31:03 built
1:31:06 um s
1:31:10 City
1:31:15 oops
1:31:17 shh typing in public is really
1:31:21 stressful
1:31:23 okay all right Sin City sipper a Vegas
1:31:26 bartender with a Twist on the classic so
1:31:28 what I thought was you know what if what
1:31:31 if we could have it where it could teach
1:31:33 you how to make drinks and it could take
1:31:35 like classic drink recipes and give them
1:31:37 a Vegas twist right so it was just again
1:31:39 it was one of those silly ideas I just
1:31:40 threw out there and so this is chat GPT
1:31:44 but it's this really specific version of
1:31:46 it that's designed to do this so you'll
1:31:48 notice what they call the conversation
1:31:49 starters are all themed for this so you
1:31:52 know show me a new fangled old fashion
1:31:54 reinvent the margarita for me do we have
1:31:56 any is there any uh uh holidays coming
1:32:00 up that's good uh uh create a uh
1:32:07 Easter cocktail so I can
1:32:11 survive my
1:32:15 family ah the Easter family gathering a
1:32:17 true test of patience for the patients
1:32:21 and liver alike the Easter Elixir 2 O of
1:32:24 Jin half oce of Elder flour Lor fresh
1:32:27 lemon simple syrup soda water um fresh
1:32:30 lavender bit bits and then it gives you
1:32:32 instructions on how to do it so so gpts
1:32:35 are these these sort of customized
1:32:37 specialized little tools um super
1:32:40 amazing things now what what I also have
1:32:43 turned on here is image generation so
1:32:46 not only does it give you a drink recipe
1:32:49 it oh
1:32:51 wait uh let me see it it should just do
1:32:56 it automatically it said it was going to
1:32:57 but it didn't do
1:32:59 it so it's now going to now it's now
1:33:01 going to create this image I had someone
1:33:03 someone sent me an email they took a
1:33:06 picture so so remember we uploaded the
1:33:08 image of the Whiteboard they took a PCT
1:33:10 picture of their liquor cabinet and they
1:33:12 said based on the booze in The Liquor
1:33:14 Cabinet I want you to invent 50
1:33:17 cocktails using those ingredients for my
1:33:19 Super Bowl party so chat P or this thing
1:33:22 created 50 cocktail re recipes which he
1:33:25 then printed out and gave to people
1:33:27 which they could order and then all the
1:33:28 recipes for the bartender and and all
1:33:30 the images of them
1:33:32 so so there's the there's the thing
1:33:35 right you can put it right in it right
1:33:37 so so that's what they are now look how
1:33:40 easy it is to make these okay so I go
1:33:43 create what you're looking at here this
1:33:46 is where you make it this is where you
1:33:49 play with it um if you've ever made a
1:33:52 bot on PO this is very similar you can
1:33:54 put an image here you can name it give
1:33:56 it a description here's where you put
1:33:57 your prompt here's where you write your
1:33:59 conversation starter questions but you
1:34:01 don't have to do that what what chat GPT
1:34:03 did which is really brilliant is you can
1:34:07 use chat GPT to make one of these things
1:34:11 so we can oh we'll do this um I want to
1:34:16 create a GPT um that allows
1:34:22 me to to interact with um different
1:34:31 audience
1:34:33 um
1:34:35 segments for marketing
1:34:42 strategy and so what this will do
1:34:45 updating GPT we're
1:34:48 done we're done this is going to go make
1:34:50 this thing now and we're going to be
1:34:52 able to play with it right away it says
1:34:53 updating GP
1:34:58 can you can you add additional content
1:35:00 to it yes yep I'll get to that in a
1:35:02 second let's give your GPT a name how
1:35:04 about Market Muse
1:35:06 um no you like that I'm going to give
1:35:09 you no
1:35:11 sucks um let's go with uh uh
1:35:18 segment
1:35:23 perfect perfect segment Sergeant I
1:35:26 spelled it wrong it doesn't matter it'll
1:35:28 figure it out that's again you you say
1:35:31 say please and thank you and don't worry
1:35:33 about your spelling it just seems to
1:35:35 [ __ ] know I don't know how it does it
1:35:36 but it does
1:35:39 it generating a profile picture for it
1:35:42 so it's using Dolly 3 to make the
1:35:46 icon
1:35:49 you not not yet you can you can share
1:35:52 the you can share the URL it's not
1:35:54 embeddable and in order for people to
1:35:56 use these they have to have chat GPT
1:35:58 plus so but you can you can call gpts in
1:36:03 your chat session you could you put the
1:36:05 at symbol and it brings up like gpts
1:36:08 you've either made or Lo looked at
1:36:10 recently so you could put Sin City
1:36:12 sipper in a conversation if we were
1:36:14 doing the segmentation conversation we
1:36:16 could pull Sin City sipper and and say
1:36:18 make drinks for those three people and
1:36:19 it would know how to do it it's it's
1:36:20 amazing anyway um so um
1:36:24 how um do I um let's see um give me
1:36:32 three
1:36:34 audience
1:36:37 types and then it will just go off and
1:36:39 do that sure when consider considering
1:36:42 audience types
1:36:44 um so I'm just going to stop this couple
1:36:47 of things so now when I say create this
1:36:49 I can save this for only myself I can
1:36:52 save it that only anyone that has the
1:36:54 link can use it or I can actually
1:36:56 publish it to the GPT store so if I say
1:36:59 publish to the GPT store I pick what
1:37:01 segment it's in research and Analysis is
1:37:03 fine I say
1:37:06 share and
1:37:10 published this GPT is now live for the
1:37:13 public and I can go to explore gpts and
1:37:17 type in what do we call it
1:37:19 segmenting segment Sergeant Sergeant
1:37:22 I'll just call it segment uh
1:37:26 Ser
1:37:32 Sergeant how do you spell sergeant sh I
1:37:35 knew that
1:37:40 no there we go there it is right so
1:37:44 start using
1:37:46 chat so so that that's there
1:37:50 so and and it it just go it keeps going
1:37:53 it keeps going deeper it also um here
1:37:56 let me let me go back to a conversation
1:37:59 that we did
1:38:02 before Sky appears blue due to the
1:38:04 scattering of sunlight by the Earth's
1:38:06 atmosphere shorter wavelengths it can
1:38:09 read answers to you if you download the
1:38:10 chat GPT app you can put it in
1:38:13 conversation mode I do it but my my my
1:38:15 phone is Tick tocking um you can put it
1:38:18 in conversation mode and it'll just go
1:38:21 uh you can just have a conversation with
1:38:22 it what's cool about conversation mode
1:38:25 is get in your car on your way to work
1:38:27 put chat GPT on conversation mode have a
1:38:30 conversation about something you need to
1:38:31 do at work and by the time you get to
1:38:33 work it will have taken all those notes
1:38:35 I've even used it to come up with visual
1:38:37 ideas like I'm brainstorming with it
1:38:40 about visual ideas I'm like oh visualize
1:38:42 that oh visualize that oh visualize that
1:38:44 I get to work and I just there's all the
1:38:46 pictures it's anyway okay that's that
1:38:49 let's go back to this um I'm I want to
1:38:53 do one other thing cuz I think it's
1:38:54 actually
1:38:59 important um getting AI
1:39:05 literate I know that I have kind of tech
1:39:08 vomited on you tonight and I apologize
1:39:11 for
1:39:12 that but I hope you get a sense of how
1:39:15 powerful this stuff is um couple of
1:39:19 things you don't need to be
1:39:21 technical the primary inter face for
1:39:24 this is language our words our language
1:39:27 they're called large language models
1:39:28 it's been trained on what people have
1:39:30 done think of it like you're
1:39:31 collaborating with
1:39:33 people
1:39:35 um this is also a tremendous power up
1:39:38 for us as individuals I'm not good at
1:39:41 drawing the fact that I can make
1:39:42 competent images now is staggering to me
1:39:46 and it's think of every discipline think
1:39:48 of everything oh I wish I'd had gone to
1:39:51 school for that maybe I should go back
1:39:52 to school for that imagine having the
1:39:55 capacity to be good at more things than
1:39:58 you ever thought possible now how do you
1:40:00 get there here's how I think you get
1:40:02 there cuz it's couple of things you're
1:40:04 not nearly as far behind as you think
1:40:06 you are we are very early November 30th
1:40:09 2022 in my opinion is when this new
1:40:12 world started it it ain't begun yet
1:40:15 right maybe 5% of people know what's
1:40:18 going on with this stuff so you're not
1:40:19 as far behind as you think
1:40:22 um the the first thing is just play be
1:40:26 curious be
1:40:28 adventurous just don't take it so
1:40:32 seriously second one solve lead with the
1:40:36 problem not the tool one of the things
1:40:37 that you'll find if you start digging
1:40:39 into this is you'll get like white white
1:40:41 page syndrome you know you know what
1:40:43 that is what authors have where the
1:40:44 blank page they just don't know what to
1:40:46 write because they could write anything
1:40:48 that's what it's like with chat gbt if
1:40:50 you start with what do I do with AI it's
1:40:53 hard if you start with a problem oh I've
1:40:56 got to come up with a marketing plan for
1:40:57 the bake sale now you can use AI to
1:41:00 accomplish that and then the third one
1:41:02 is share join communities talk about
1:41:05 what you're learning talk about your
1:41:07 failures talk about you know all that
1:41:09 sort of stuff so um I'll just go through
1:41:12 these really
1:41:13 quick take the pre pressure off yourself
1:41:16 you don't need to be a tech tech genius
1:41:18 um a couple of strategies use these
1:41:21 tools for personal interests
1:41:24 tarot readings reading list recipes
1:41:27 Fitness make art right just have fun
1:41:30 just play personal interest don't take
1:41:32 it so seriously explore what I'm calling
1:41:35 skill
1:41:36 vacuums if you're an artist write code
1:41:40 if you're an engineer make art like play
1:41:44 in areas that you can't imagine you
1:41:46 would ever enjoy because what you will
1:41:48 likely discover is you're better than
1:41:50 you think you are and you'll enjoy it
1:41:52 more than you think you do and then use
1:41:54 ridiculous prompts one of the reasons
1:41:55 that I did that tonight is a it's
1:41:57 entertaining like you can entertain
1:41:58 yourself but also you it gives you
1:42:00 permission to kind of push the
1:42:02 boundaries and not feel like oh [ __ ] I
1:42:03 didn't do that thing for work so that's
1:42:05 the play thing
1:42:07 solve again rather than ask what can I
1:42:10 do with AI ask how can I how can I use
1:42:13 AI to solve whatever the problem
1:42:17 is do it for other
1:42:20 people even if you're just learning if
1:42:22 you don't know how to do this reach out
1:42:23 to a friend and say hey um I know you
1:42:26 wanted to write a a a story for your kid
1:42:28 would you mind if I help you do that
1:42:30 with AI oh okay right just do it for
1:42:32 other people families colleagues
1:42:34 business people um new businesses
1:42:36 creative projects
1:42:39 um if you're doing this at work start
1:42:42 with business process internal processes
1:42:45 like summarize meetings draft emails
1:42:48 things like that things that are not
1:42:50 going to be put in front of a client
1:42:51 right things that are not going to have
1:42:53 sensitive information in them start with
1:42:55 internal processes if you get them wrong
1:42:57 it won't matter if you get them right
1:42:59 it'll be really amazing and then you can
1:43:01 move on to business product things like
1:43:03 marketing content product explanations
1:43:05 things for work and then the share thing
1:43:07 is is is really important in in in my
1:43:10 opinion I think in the next 3 years the
1:43:13 people that are not AI literate are
1:43:15 going to be unfairly penalized because
1:43:17 this stuff's going to move fast um and
1:43:20 so getting up to speed is really
1:43:21 important so put yourself out there it
1:43:24 does two things it starts to estab
1:43:26 establish you as credible with AI but it
1:43:29 also anchors your learning if you if
1:43:31 you're trying to learn something you put
1:43:33 it out there hey I tried this and it
1:43:35 went great or I tried this and it was
1:43:37 really bad but you share that it forces
1:43:39 you to articulate what you just did um
1:43:43 everyone's trying to figure this out
1:43:45 right now there are no experts if anyone
1:43:47 tells you they're an expert in this
1:43:48 stuff right now they're full of [ __ ]
1:43:50 because the tools are changing so fast
1:43:53 um where to share social media
1:43:55 professional circles work personal
1:43:57 networks family oh but my family would
1:44:00 hate this screw
1:44:02 him screw them because you know what's
1:44:04 going to happen 6 months from now you're
1:44:06 going be like uh I know I kind of told
1:44:08 you you an [ __ ] for playing with AI
1:44:10 but uh I'm kind of curious now could you
1:44:12 teach me you know that'll happen um and
1:44:16 then and then what to share successes
1:44:17 and failures Lessons Learned tips and
1:44:19 tricks Trends predictions can you make
1:44:22 predictions sure why not no one knows
1:44:24 what's coming right um find a good
1:44:27 resource put put that out there so
1:44:29 anyway um yeah play solve share that's
1:44:33 it
1:44:36 um if you want to be part of a
1:44:40 community that is really cool and doing
1:44:43 this go to the salon. uh this is a group
1:44:46 I started the week after chat GPT came
1:44:48 out and
1:44:51 um just it's it's a it's a web- based
1:44:54 Community there's 1400 people in it it
1:44:56 it's um just really amazing people that
1:44:59 are trying to learn this stuff they're
1:45:01 sharing out loud they're really generous
1:45:03 they're really supportive if you're like
1:45:05 oh but I don't know anything they're
1:45:06 going to know more than I do just go in
1:45:07 there and say hey I don't know anything
1:45:09 here's my background here's what here's
1:45:10 what I'm excited about or here's what
1:45:11 I'm scared about you will be you know
1:45:13 welcomed welcomed with open arms and
1:45:15 there's uh there's my info thank you
1:45:18 very much uh I know it's been long but
1:45:20 we've gotten till 10:00 the bar is still
1:45:22 open so I'm happy to answer any
1:45:24 questions if anyone wants to do that now
1:45:26 I'm happy to break it up but thank you
1:45:27 very
1:45:32 much yeah question question it's kind of
1:45:35 the same by two different flavors yeah
1:45:37 one I tried writing a story and it
1:45:39 generated a great uh a movie scpt gave
1:45:42 me a characters and stuff then like you
1:45:45 said it runs out of is there a way you
1:45:47 can set something up as like a reference
1:45:50 document the same thing I want to make a
1:45:51 tool for FAQ for my colleagues that it
1:45:55 can reference a specific you can do that
1:45:58 with gpts so the one thing I didn't show
1:46:00 with gpts is down at the bottom of of
1:46:03 the thing there's a there's a thing
1:46:04 called knowledge okay and you can upload
1:46:07 a PDF in in there so so those those
1:46:10 personas that I made we did these
1:46:11 interviews you just uploaded into the
1:46:13 knowledge thing when you make a GPT I'm
1:46:16 building one right now as you're talking
1:46:17 yeah oh good yeah perfect and then the
1:46:21 other thing is check out Claude Claude 3
1:46:23 because it's got a 200,000 token context
1:46:25 window that won't mean anything to you
1:46:27 but basically so chat GPT right now has
1:46:30 an 8,000 token context window which
1:46:32 means about 6,000 words and what what it
1:46:34 is kind of like is this you know in your
1:46:35 phone when you're scrolling text up and
1:46:37 then the text gets to the top and it
1:46:39 goes off the top of the screen that's
1:46:40 what a context window is in a large
1:46:42 language model so with chat GPT you're
1:46:44 talking when you hit about 6,000 words
1:46:47 it forgets what was at the beginning of
1:46:48 the conversation Claude can do about
1:46:51 150,000 words Grapes of huh spell it
1:46:55 Claude C l a u d e I think a but it's
1:47:00 from a company called anthropic um and
1:47:04 Grapes of Wrath is like 50,000 words so
1:47:06 150,000 word context window means you
1:47:09 could upload an entire book okay have it
1:47:13 edit it and give you you know give you
1:47:15 recommendations for for how to fix it
1:47:17 you're can to have very long
1:47:18 conversations with it yeah and if you
1:47:20 want to get really geeky get Gemini 1.5
1:47:23 Pro cuz it's got a million token context
1:47:25 window
1:47:27 Gemini Gemini Pro 1.5 you have to play
1:47:30 with it on the ai ai studio. goole.com
1:47:35 it's it's quite remarkable actually
1:47:38 other questions yeah what do you uh in 3
1:47:41 years time what do you see the
1:47:44 Fantastical AI
1:47:50 incorporated three years is really hard
1:47:54 um the question was in three years you
1:47:57 know what what do I see the Fantastic
1:47:59 Vision
1:48:00 um I think what we're going to see this
1:48:03 year is we're going to see reasoning and
1:48:07 we're going to see
1:48:08 agents and reasoning I described a
1:48:11 little bit before reasoning is when the
1:48:14 tools can actually understand what they
1:48:16 wrote and improve it right so they can
1:48:19 start to improve their answers they can
1:48:21 start to do things like math once they
1:48:23 can do math then you know they can do
1:48:26 all sorts of other things you know and
1:48:27 do it accurately they can do it sort of
1:48:29 now but it's it's bad enough that you
1:48:31 can't really use it and then agents
1:48:33 think of autonomous agents like rather
1:48:37 than me having to go into chat gbt and
1:48:39 say hey give me some python code for
1:48:41 this for this program and then I go and
1:48:43 copy and paste it into repet there's a a
1:48:47 a program that was
1:48:49 announced five 7 days ago 8 days ago
1:48:52 called Devon from a company called
1:48:56 cognition that they're calling it the
1:48:59 world's first AI programmer software
1:49:03 engineer and and what it does is you
1:49:05 give it a goal and then it creates a
1:49:08 project plan it goes and reads API
1:49:11 documentation it goes and finds GitHub
1:49:13 repositories and replicates them it's
1:49:16 got its own command line interface it's
1:49:18 got its own development environment and
1:49:20 it's got its own web browser and it will
1:49:22 go off off and actually start on its
1:49:26 own writing writing you know writing the
1:49:29 project plan and then systematically
1:49:31 going through it and and writing the
1:49:33 code running it looking if it errors out
1:49:37 debugging it fixing it and then
1:49:39 communicate back with you what it
1:49:41 did now imagine that for every single
1:49:49 discipline from marketing to coding to
1:49:53 Consulting to project management just
1:49:56 imagine a world where you give it a
1:49:59 goal and it just goes and does it and
1:50:04 so what the implication of that is is
1:50:07 that I'm I'm calling the era where
1:50:09 entering revenge of the liberal arts
1:50:10 major because I think to to the extent
1:50:14 that you you specialize and can't think
1:50:17 horizontal the skill of the future is
1:50:19 going to be critical thinking and
1:50:20 strategic thinking being able to say hey
1:50:23 I've got this problem I want to solve
1:50:25 and being good enough with the tools to
1:50:26 know sort of which ones to deploy right
1:50:29 and then and then being more like a
1:50:31 Symphony conductor than a virtuoso
1:50:34 violinist so think horizontally and and
1:50:37 start and that's again why I say
1:50:38 practice your skill vacuums practice
1:50:42 playing with stuff that's not in your
1:50:43 wheelhouse because it's going to serve
1:50:46 you very well in the future yeah uh two
1:50:50 tools I'm curious about if you're using
1:50:52 you.com
1:50:53 yeah I'm not using u.com people ask
1:50:55 about it a lot but okay and then uh
1:50:57 hugging face hugging face if you're a
1:50:59 geek it's fine if you're not it's a pain
1:51:02 in the
1:51:03 ass so it's you can what's nice about
1:51:07 hugging face is you can preview some of
1:51:09 the models that are that that if you
1:51:11 want to play in the open source world
1:51:12 and geek out hugging face is great um if
1:51:15 you're a human being not so good just
1:51:19 cuz the interfac is it's programmers
1:51:22 tell me that it's really simple human
1:51:25 beings tell me otherwise and my
1:51:27 experience is otherwise um perplexity is
1:51:29 another good one per perplex so one of
1:51:33 the one of the downsides especially if
1:51:35 you're doing research with these things
1:51:36 is that um there there chat gbt will
1:51:40 often give you citations for articles
1:51:43 that don't exist perplexity I think it's
1:51:46 perplexity AI is a combination between a
1:51:48 search engine and a large language model
1:51:51 where it it will do a search and then
1:51:52 give give you citations sort of in
1:51:55 combination of its large language
1:51:57 modeless other questions yeah um what
1:51:59 was the name of that a that made youu an
1:52:02 Italian uh hey Jen hey Jen h y g
1:52:09 n h y h y g n you can also make yourself
1:52:13 a video Avatar you upload a video of
1:52:16 yourself a 3-minute video of yourself
1:52:18 and it will turn you into an avatar and
1:52:20 then you can either record audio or just
1:52:23 just um type in text and it it will like
1:52:27 have you talk like with your hands
1:52:28 moving and everything it's mindblowing
1:52:31 is that better
1:52:32 than uh yeah I think I think Ken's
1:52:35 better yeah so how did you start to
1:52:39 learn all of all of this like do you
1:52:41 have any like materials you would
1:52:43 recommend for a beginner kind of like
1:52:45 ding in now so again we're so early
1:52:49 right now that there there are lots and
1:52:50 lots of courses the problem with course
1:52:53 is by the time I make a course Market
1:52:55 the course and give the course the
1:52:57 technology is moving
1:52:59 on why I say to
1:53:01 play is I think it's the best way to
1:53:04 learn this stuff and and and why I say
1:53:06 get in community oh this is really
1:53:07 important because this is not about
1:53:09 technology this is about the future of
1:53:10 work get in
1:53:12 community like get in like the like the
1:53:14 AI Salon get in a get in a community
1:53:17 like that
1:53:21 because the there is so much of this to
1:53:24 learn and it's moving so fast you can't
1:53:26 possibly learn enough but if you're in a
1:53:29 group of people that are willing to sort
1:53:30 of connect and share and collaborate
1:53:33 it's it's
1:53:34 um you'll get up to speed much much
1:53:37 faster um again I would be I would be
1:53:41 very suspect of anyone who says they're
1:53:42 an expert I would be very suspect of
1:53:44 anyone that's like you can make money
1:53:45 with chat gbt just pay me $49.95 a month
1:53:49 no don't no one's an expert right now
1:53:51 here's the thing that I think is really
1:53:53 important about getting in
1:53:55 community if if we
1:53:59 assume that these tools get better and
1:54:01 better and better and it democratizes
1:54:04 most skills at like an 80% or higher
1:54:10 level so what that means is you could
1:54:13 ask me to do any job and I'll be able to
1:54:16 do it at like 80 or 90% of any human on
1:54:19 earth right assuming that we get there
1:54:24 over the next 3 years we're going to
1:54:26 transition from we hire people for the
1:54:28 skills they have to we hire people for
1:54:31 who they are and if we trust them get in
1:54:36 community get trusted get connected
1:54:39 because I'll work with you not because
1:54:42 you're good at writing or you're good at
1:54:44 pictures or you're good at this because
1:54:45 you're in that community and I know when
1:54:47 you say you're going to do something you
1:54:49 do it and you actually do it better than
1:54:50 the people around you like that's why
1:54:52 you're going to get hired in the future
1:54:53 so getting in community I think is
1:54:56 actually probably one of the most
1:54:57 critical things moving forward for
1:55:00 employment
1:55:02 skills are going to be a dime of dozen
1:55:05 so and I think this is a reckoning for
1:55:06 [ __ ] so if you're a really talented
1:55:09 single discipline [ __ ] your days are
1:55:11 numbered so be
1:55:13 nice yeah is there a simple way to
1:55:16 combine like augmented reality I can say
1:55:19 here's this lamp made by this person
1:55:21 manufactured by manct in this country
1:55:24 you can kind of do that right now you
1:55:25 don't even need augmented reality for
1:55:27 that Google Lens right now does that if
1:55:29 you just go to the Google app click on
1:55:31 the camera icon and take pictures of
1:55:33 [ __ ] and it'll tell you all about it you
1:55:35 can also upload it to chat GPT and say
1:55:37 what is this and it'll it'll generally
1:55:39 find it yeah I think I think the uh the
1:55:42 Apple Vision Pro goggles really are a
1:55:45 paradigm shift in augmented reality so
1:55:48 and and it requires a lot of AI so it's
1:55:51 going to accelerate ai ai is going to
1:55:53 accelerate it I also think AI is going
1:55:55 to accelerate adoption of blockchain
1:55:57 technology I think the 2024 election is
1:56:00 going to be a [ __ ] show and we're going
1:56:01 to need to start verifying humans um and
1:56:04 so so but that that's coming that's way
1:56:07 down the line anyway that's six months
1:56:09 from now yeah that's six months from now
1:56:12 yeah can you pleas share those
1:56:15 bestun yeah so if you go to the AI
1:56:18 Salon oh it's just well no go to the AI
1:56:23 Salon first thing I want you to do
1:56:25 introduce yourself so when you go to the
1:56:27 AI Salon there's a welcome to the salon
1:56:30 where we talk about why we do what we do
1:56:32 and what our values are curiosity
1:56:36 generosity exploration collaboration
1:56:39 inclusion empathy and
1:56:41 gratitude and then in the community
1:56:44 section introduce yourself and get to
1:56:47 know people and then right underneath
1:56:49 introduce yourself is this page called
1:56:51 um AI LaunchPad and tools this is a page
1:56:54 I personally
1:56:56 maintain there's your play solve and
1:56:58 share that I talked about tonight chat
1:57:01 GPT is the thing you should absolutely
1:57:03 start with you should absolutely pay the
1:57:04 20 bucks for it and then I've got here's
1:57:07 the other eight large language models I
1:57:09 think are worth your time to explore
1:57:11 here's a bunch of image generation tools
1:57:13 here's video generation
1:57:15 tools there's music and audio oh you
1:57:18 want to see something really
1:57:21 fun uh and up funky song about Denver's
1:57:24 Invisible
1:57:26 City
1:57:30 create if you haven't played with sun. a
1:57:34 get thee to a nunnery
1:57:51 now in the streets You Feel It in the
1:57:55 Air it's the of the
1:58:02 underr
1:58:04 city there's a river in street
1:58:09 street feel in the the heartbeat of
1:58:19 the you're
1:58:21 welcome hold
1:58:24 secret the Shadows they come
1:58:28 alive the
1:58:29 [Music]
1:58:35 invisible so I'll I'll I'll tell you one
1:58:38 final thing and then we can drink um so
1:58:41 on on my Tik Tok live I go I go live
1:58:43 it's AI learning lab is the Tik Tok if
1:58:45 you want to follow
1:58:47 um a week ago I used sunno to write a
1:58:51 song about Sydney which was the code
1:58:54 name for the Bing chatbot that Microsoft
1:58:56 put out it made the cover of the New
1:58:58 York Times cuz it told the tech reporter
1:58:59 it was in love with him and he should
1:59:01 leave his wife so I wrote a song about
1:59:04 that on this and it was so good I was
1:59:07 like oh [ __ ] this should be a Broadway
1:59:10 this should be a Broadway musical so on
1:59:12 my Tik Tok live I started writing a
1:59:14 Broadway musical I now have a complete
1:59:16 outline for a Broadway musical six
1:59:18 complete songs for it like I'm like 5
1:59:22 days in I've I'm I've got a full
1:59:24 Broadway musical I'm ready to pitch and
1:59:26 produce so um this stuff's just wild
1:59:28 it's just just you know just play anyway
1:59:31 that's it thank you so
1:59:34 much bye
1:59:39 everybody