Jul 31, 2024 - Mastering AI Voice Technology for Music, Storytelling, and Beyond!

Description

Chapters

Transcript

0:01 [Music]
0:07 [Applause]
0:10 [Music]
0:21 oh
0:22 [Music]
0:41 you're feeling nervous aren't you
0:45 boy with the Quiet
0:48 voice and impeccable
0:51 [Music]
0:53 style don't ever let them steal your
0:57 joy and your gentle wng
1:01 going
1:04 wild they dirt in your face dress you
1:08 down and tell you that your place is in
1:10 the middle when they hate the way you
1:14 [Music]
1:17 shine I see you tugging on your
1:21 shirt trying to hide inside of it
1:27 [Music]
1:31 while they
1:33 can let them spin a
1:37 [Music]
1:39 spin let scatter in the
1:44 wind I've been to the movies I've seen
1:49 How It
1:51 Ends and a Jo
1:55 [Laughter]
1:57 [Music]
1:59 [Laughter]
1:59 [Music]
2:13 good evening good people Champ's a
2:15 little worked up
2:18 tonight do me a favor let's see if we
2:21 can get some more folks in here tap the
2:22 tapy
2:23 Taps Silver Fox shared the live with
2:26 more than 10
2:28 friends that's remarkable
2:31 be like Silver Fox be silver be a fox
2:40 [Music]
3:09 I think I might try something tonight
3:11 one of the one of the new features of
3:13 yudo so udio 1.5 just
3:18 [Music]
3:26 released
3:28 um then one of the features is called
3:31 remix which I saw the other
3:34 night but I didn't really pay attention
3:36 to it I was like oh that's something I
3:38 got to
3:39 learn I was in a very non-play kind of
3:41 place I was like yeah this seems like
3:43 work I I can't possibly learn a new
3:47 feature of a piece of
3:51 software but what you can do is you can
3:55 take a song that
3:58 exists and just remix it into another
4:02 style so I figure maybe what we'll do is
4:04 I'll I'll record one of my one of my
4:07 songs I play something I know maybe I'll
4:11 do an
4:12 [Music]
4:26 original standing between
4:30 you in a hard place is
4:35 insane standing too
4:38 near you and a fire makes it
4:44 clear your trouble to
4:48 [Music]
4:50 me real trouble can't you
4:55 see so like that one is just
4:58 so like
5:00 simple acoustic singer songwriter just
5:03 you know sad minor chord [ __ ] it'll
5:06 be interesting to see if we can turn
5:07 that into like you know some disco pop
5:11 [Laughter]
5:17 song now it could be a disaster cuz you
5:20 know I ain't never tried that D feature
5:22 before I ain't got no books
5:25 smarts sir sir yes uh this is your
5:30 Channel yes carry on yes yes yes a
5:33 question for you sir rather go on what's
5:36 the question uh yes sir I was looking to
5:39 inquire sir uh what are by chance your
5:45 qualifications my
5:47 qualifications yeah it's very simple
5:49 question just curious AI learning lab
5:52 apparently you're a scientist or some
5:54 sort of professorial educator I was just
5:58 wondering what of course your
5:59 qualifications were
6:01 oh so I have none you have
6:05 none you can't just name a Tik Tok
6:07 Channel and start
6:11 talking
6:13 so depending which version of Me shows
6:17 [Music]
6:19 up it could be the brilliant one or it
6:22 could be dodo
6:24 [Laughter]
6:25 head zero qualifications zero zero
6:29 really
6:38 [Music]
6:57 w w how don't mean cause you in a
7:03 sorrow didn't mean cause you in a
7:09 [Music]
7:10 pain
7:13 [Music]
7:15 oops you got to put your fingers in the
7:17 right
7:20 Frets that's a interesting
7:22 [Music]
7:25 chord let's give that to yudo and see if
7:28 we can turn it into a [ __ ] s
7:32 song zero my hero
7:36 exactly part of the lesson of the AI
7:39 learning lab is that
7:42 um you don't need to be a technologist
7:44 to play with
7:47 AI you don't need no book
7:51 learning um you need a healthy dose of
7:56 curiosity little sense of adventure
8:00 some
8:01 patience some
8:03 [Music]
8:06 playfulness a little bit of
8:09 [Music]
8:12 empathy you know what's funny about
8:15 [Music]
8:18 empathy
8:21 um it's generally good as a life skill
8:25 [Music]
8:30 get him
8:32 [Music]
8:42 champy
8:44 [Music]
8:45 um empathy is good um as a as an ADHD I
8:51 tend to over index an empath
8:54 empathy which ain't ain't super healthy
8:58 but
9:00 [Music]
9:05 understanding how people
9:07 tick and
9:11 understanding how to connect with people
9:14 is
9:16 oddly some of the best practices for
9:20 prompting a large language
9:23 model those people that are better at
9:25 the human
9:27 skills are really good with this stuff
9:30 so it's an area you can practice and
9:33 what's cool about large language models
9:34 is you can practice it on the large
9:35 language models and I think I think
9:37 maybe one of the other things we'll do
9:39 tonight we'll play with that udio
9:49 feature feel better Silver
9:52 Fox
9:54 [Music]
9:57 um I think we'll go watch some of the
10:00 videos of people who Okay yall can start
10:03 now thank you Source Camp I appreciate
10:04 it we
10:07 uh we're finally we're finally good to
10:09 go thank you awesome
10:16 um I think we'll go watch some videos of
10:19 people that have gotten the new voice
10:21 feature
10:23 um because it's interesting watching
10:26 people use it who don't work for open AI
10:29 cuz they're doing doing things with it
10:30 that open AI didn't do right open AI did
10:33 their little can demos the [ __ ] they
10:35 knew would
10:36 [Music]
10:38 work
10:40 um but what's what's becoming quite
10:43 clear to
10:44 me is
10:48 that having a system that will respond
10:52 to you essentially as quickly as a human
10:54 being does and having a system that
10:57 understands your emotions and can
11:00 respond in
11:03 kind is [ __ ] it's going to change
11:10 things and especially when it can do
11:13 things like look at what you're doing
11:17 and interact with that as
11:20 well you know is this is this a good
11:23 bottle of wine right things like that
11:26 Source Camp is in the room champ
11:31 the bad people aren't going to get us
11:33 we're
11:36 [Music]
11:39 good so welcome everybody if you're new
11:41 here my name is Kyle shanon this is the
11:43 AI learning
11:46 lab we do a little serenading we little
11:50 do a little dog singing stupid human
11:52 tricks I mean stupid pet tricks
11:57 [Music]
12:05 um we do a little
12:06 uh AI
12:09 demonstrations I'll answer questions I
12:11 will pontificate philosophically about
12:14 what's going on with AI like for
12:18 example there's a bunch of AI doomers
12:21 out there talking about
12:23 how the AI hype bubble has burst and
12:27 we're on the decline
12:29 and and AI has plateaued and it will
12:32 never get any
12:39 better what's amazing is a lot of these
12:41 are people who've been on the planet
12:43 long enough to know but
12:50 better I mean if you invest trillions of
12:54 [Music]
12:56 dollars
12:58 into into a
13:02 category and refocus all of the world's
13:05 engineering Talent toward that one
13:09 thing and it's already [ __ ]
13:13 remarkable and an 800 person company
13:16 essentially is challenging a 200,000
13:20 person company that's open Ai and
13:23 [Music]
13:27 Google you don't think some shit's
13:29 coming
13:34 and if you're sitting if we're sitting
13:37 in a in a state of the industry where
13:40 40% of Americans haven't even heard of
13:42 chat
13:43 GPT and less than 2% of
13:46 Americans have used chat gbt for more
13:49 than an
13:50 [Music]
13:56 hour so that means you know half a
13:59 percent
14:02 of
14:03 people
14:06 know have even a hint of how powerful
14:09 this [ __ ]
14:11 is you don't think when that number goes
14:14 from half a percent to
14:16 20% some [ __ ] ain't going to
14:20 happen and that'll coincide with the
14:22 technology getting
14:24 better and it'll coincide
14:28 with the this new chat voice model being
14:34 good enough that people can just talk to
14:36 [Music]
14:38 it and have it tell stories and make
14:41 sound effects it's [ __ ]
14:44 [Music]
14:48 crazy so here's the good news if you're
14:51 new to the channel if you're new to AI
14:53 feel like I don't know what this sayi is
14:55 listen listen Mister I didn't go to MIT
14:59 or one of the them fancy mathematical
15:02 schools if you're sitting there on the
15:04 sidelines thinking that you got to be
15:05 one of them phds like like Pate modder
15:08 who hangs out in here he's an actual PhD
15:11 in mathematics working for
15:14 Google you don't have to be like
15:19 that you
15:21 don't it's a remarkable time we're in
15:24 and as more and more people discover
15:25 that and discover that AI can be fun AI
15:29 can be inspiring AI can be your
15:31 companion AI can be a therapist AI can
15:34 be a
15:42 friend how we interact with computers is
15:45 about to shift
15:46 dramatically
15:48 where kind of the primary skill right
15:52 now and this is going to get more and
15:54 more intuitive but this is really weird
15:57 and I mean Source Camp I'm glad you're
15:59 here cuz it's like this this is kind of
16:01 how Kelly approaches her work
16:04 is you just ask this [ __ ] for what you
16:09 want like Kelly was telling me she was
16:12 working with some business owner that
16:15 had did something in bars or restaurants
16:18 or something like that and he's like
16:20 well could it help me with my
16:22 inventory and she's not a restaurant
16:25 tour but she's like I don't know let's
16:29 try dear chat PT she took a picture of
16:33 his bar with all the
16:35 bottles she said based on the levels of
16:38 liquid in the bottles can you give me an
16:41 inventory report and any bottle that's a
16:44 third or less put it on the
16:47 reorder flag or you know
16:52 status out it
16:55 came that guy had his first Kevin
16:57 McAllister moment
17:00 it could do
17:04 that
17:07 like you just have to be willing to ask
17:10 it one of the things that I enjoy most
17:13 in here is just asking AI stupid
17:19 [ __ ] because more often than not it does
17:22 it and then you're like holy really
17:27 [Music]
17:29 can you tell me his credentials as to
17:32 why he's an an authority on
17:39 AI sir I would like to
17:43 know yeah yes sir Mr Shannon Mr shner
17:46 yes uh wondering inquiring very deeply
17:49 uh and sincerely and no offense meant
17:51 but uh what sir by chance prto are your
17:57 qualifications you have no
17:58 qualifications
17:59 oh wow here are my
18:02 qualifications I have a degree in
18:05 [Laughter]
18:11 acting I am a self-taught mediocre
18:14 guitar player
18:16 [Laughter]
18:16 [Music]
18:31 I would say I I'm
18:33 not I'm not really qualified to talk
18:36 about the technology of
18:38 AI I can explain how it works but I
18:41 can't do it I'm not a
18:47 [Music]
18:48 mathematician created a full month's
18:50 roster day hour by hour adding in all my
18:54 work exercise Etc today there you go
18:58 you're absolutely looking for a career
18:59 in politics I I am Shannon Sky
19:03 2028 if you don't know I'm going to be
19:05 running for president in 20128 well no
19:08 I'm going to be running as the spokes
19:11 model for an AI for president in
19:15 [Music]
19:16 2028 and my only promise is we'll do
19:19 what the AI says I'll just be here to
19:21 deliver the message um no so but
19:25 seriously um
19:30 [Music]
19:34 here's here's my experience the
19:36 storytelling thing is oddly a
19:39 qualification
19:41 um when I got really interested in AI
19:44 I've been paying attention to AI for I
19:45 don't know the past decade or so I have
19:47 a technology company uh an automated uh
19:51 video storytelling platform called Story
19:53 vine uh we've been around for 12 years
19:56 just added a whole bunch of AI features
19:58 to it which is pretty exciting
20:00 so I've built some [ __ ] but that's
20:02 neither here nor there
20:05 um I was around I don't know if you can
20:08 tell I know I look super model beautiful
20:11 but I'm an older
20:12 [Music]
20:14 fell I'm
20:17 59 and I
20:24 uh I was living in New York City in the
20:27 mid90s and I discovered
20:30 um the worldwide web like a lot of
20:33 people
20:35 and I knew a bunch of talented people
20:37 and I started one of the first digital
20:39 magazines I'll tell you I'll tell you
20:41 the the the the Epiphany I had with the
20:43 worldwide web and it'll be relevant to
20:47 this it's funny I don't know why I'm
20:49 actually answering the question what are
20:51 your qualifications tonight cuz quite
20:54 frankly my qualifications are I turned
20:56 on the [ __ ] camera on Tik Tok and
20:58 you're sitting here listening to to me
20:59 so that's all the qualifications that
21:02 are really necessary
21:03 but 1994 I teach myself enough HTML
21:07 there was only like six tags so you
21:08 didn't have to be that
21:09 technical I teach myself enough HTML
21:14 to create the first
21:21 issue of a web magazine called Urban
21:25 desires this by the way is a sidequest
21:28 drive this say an 888 megabyte I think
21:32 this is an 88 one might have been a 45
21:36 44 but that's a hard drive in a cassette
21:39 called a sidequest drive and that was
21:42 how you store data back in the olden
21:45 time in my
21:47 day our floppy drives were this
21:50 [Laughter]
21:52 big
21:54 um so I so I I learned enough HTML to
21:57 put together this art and culture
21:59 magazine called Urban
22:02 desires and I knew a couple of technical
22:05 people I'm like well how do I get it on
22:06 the internet and they're like oh you got
22:08 to put it on an internet server and I
22:10 didn't know what that was and they were
22:11 like well I got a phone number of some
22:14 system in California and you can upload
22:16 your files to that computer in
22:19 California and then you'll be on the
22:21 internet I was living in Brooklyn
22:24 and I uploaded my files to this computer
22:28 in California
22:30 and I was on the internet my magazine
22:32 was on the internet the issue 1.1 of
22:35 urban Desires in November of
22:38 1994 and 3 weeks
22:42 later I was on a uh an online bulletin
22:45 board system in New York City called
22:47 Echo was where a lot of the New York
22:49 media people hung out
22:53 reporters computer geeks and nerds and
22:55 hackers it was
22:56 cool and someone on eeko hey
23:00 Kyle did you know that Urban desires was
23:04 written about in Liberation the Parisian
23:08 News
23:10 Daily I was like
23:12 what like yeah it's it's the it's a big
23:15 article on Urban Desires in this
23:17 Parisian
23:19 newspaper so I'm living in New York City
23:22 and I knew that there was an
23:23 international news stand in Times
23:26 Square and so the following Sunday
23:29 I went down to the international news
23:31 stand in Time Square and I bought an
23:34 issue of
23:36 liberon
23:38 and I'm standing in the middle of Time
23:40 Square cars are blowing by and people
23:43 are blowing by and I'm flipping through
23:45 this
23:46 paper and sure enough there's a full
23:49 page [ __ ] article about Urban desires
23:53 with
23:55 screenshots of these web pages I made in
23:59 my apartment in Brooklyn and uploaded to
24:01 a server in California and 3 weeks later
24:04 they were written about in
24:08 Paris and I was like
24:11 shaking because what I knew in that
24:16 moment was that the world had just
24:19 changed but no one knew it but I knew
24:24 it and it was it was like this electric
24:28 jolt went through me it was like holy
24:31 [ __ ] this internet thing this worldwide
24:34 web thing is going to change everything
24:37 and I knew it I didn't like surmise it I
24:40 didn't you know make predictions like I
24:43 I felt it in my
24:46 soul and it did and I ended up
24:49 co-founding agency.com which was one of
24:51 the first digital agencies and so I took
24:53 a lot of my storytelling expertise and I
24:55 figured out how to tell brand stories
24:59 in interactive environments and how do
25:01 you translate a brand like British
25:02 Airways or cocacola or Nike into an
25:05 interactive environment that's what we
25:07 got to do we got to build a lot of the
25:09 first websites for the Fortune 500 it
25:11 was [ __ ]
25:13 insane so I got to be on the
25:18 inside of that Revolution from from
25:21 essentially the
25:23 beginning like New York City was the
25:25 center of
25:30 the technological development of the web
25:33 which was counterintuitive right because
25:35 you think oh it should be from San
25:37 Francisco but New York City is a Media
25:40 Town New York City's about words and
25:42 articles and right and so the media of
25:45 social media New York City was the
25:48 center of
25:51 that
25:53 so the [ __ ] that got to do with AI
25:56 Kyle what are your qualifications for AI
25:59 I didn't ask you about your stupid
26:00 websites you built with your stupid dumb
26:01 HTML tags you dumb dumb probably didn't
26:04 even have CSS back then did
26:06 you
26:10 [Music]
26:10 [Applause]
26:14 correct so in
26:20 2021 I'm fairly far down the rabbit hole
26:22 of blockchain
26:24 and
26:26 nfts I get involved with this project
26:29 called
26:30 neotokyo which was really bitching for a
26:34 while and I was like wow this this
26:39 blockchain nft stuff this could change
26:41 some
26:42 things and so I started going down the
26:44 rabbit hole deeper and deeper and the
26:46 more I
26:47 dug the less real it
26:50 felt it's like who are these guy it was
26:53 like all these crypto Bros and like
26:54 everyone was trying to scam each other
26:56 and be each other's exit liquidity and
26:58 it was just [ __ ]
26:59 weird and so I I I kind of soured on the
27:05 on the nft things but one of the
27:07 projects that one of the The Other Guys
27:10 in neotokyo was a co-founder of this
27:12 project called pixel mind and pixel mind
27:16 was an nft project built on top of
27:19 stable
27:20 diffusion I am late what did I miss
27:22 nothing just my origin story I'm
27:25 answering the question Mr it of Sir what
27:28 what are your
27:30 [Music]
27:32 qualifications um so this project pixel
27:36 mind the the CEO did this really cool
27:39 thing where he created these Mad Libs
27:41 style prompts that you just filled in
27:44 certain words it would be like a blank
27:46 mask and then you'd put in a an animal
27:48 mask or a ceremonial mask or whatever
27:51 and then it would why
27:54 Denver um that's just a long story cuz I
27:58 had to we had to get out of the East
28:00 Coast um just for kids kids were really
28:02 stressed out and full of anxiety and so
28:04 it was either back to New York City move
28:07 to LA or Boulder was was um was where we
28:11 were thinking of moving and we ended up
28:13 moving to Boulder so that that was that
28:15 um so
28:16 anyway so I started playing with this
28:19 thing called pixel mind where you you
28:21 just make these images really easily it
28:23 was early stable diffusion and they were
28:25 just beautiful and you could you could
28:27 watch the images build
28:29 and it it would even record movies of
28:31 the images building and you could for
28:33 every image that you created you could
28:35 watch the movie of it being created it
28:37 was just
28:39 fascinating and then I I tried to teach
28:42 myself
28:44 um stable diffusion and how to install
28:47 it and I failed miserably and then I
28:49 finally after nine attempts finally got
28:51 it
28:52 running and I and I installed this thing
28:54 called dream Booth which allowed me to
28:56 upload pictures of myself and and um
29:00 make all these crazy weird
29:01 self-portraits and I'm like oh that's
29:03 kind of cool and so I started an art
29:05 project called Kyle Shannon dreams so if
29:07 you go to Instagram at Kyle Shannon
29:10 dreams or if you go to Tik Tok at Kyle
29:12 Shannon dreams you can see a bunch of
29:15 those images that I I did and stories
29:18 that I wrote and on Tik Tok I actually
29:20 acted out the stories that I wrote I I I
29:23 became the
29:26 characters and so what started happening
29:29 to me what was seeping into my brain
29:33 is with this AI stuff I'm
29:37 outputting things that
29:39 are dramatically better than the amount
29:43 of my input like I'm putting in this
29:45 much effort and I'm getting back this
29:48 much
29:49 results and so that was kind of
29:53 mind-blowing and then chat GPT came
29:56 out and I'd been playing with
29:59 gpt3 on the open AI playground and I I
30:02 sort of got it I sort of got what was
30:04 going on but when chat GPT came out
30:07 there was something about they they just
30:08 Stripped Away all of the knobs and dials
30:11 of models and weights and temperature
30:14 and all that [ __ ] and it was just
30:18 like the Google search box it was just
30:20 like this magic
30:21 box and you just type [ __ ] in it and
30:24 [ __ ] would come
30:26 out and I'm like this seems
30:30 interesting and so one night I'm sitting
30:33 on the couch we're watching some shitty
30:35 TV show so I'm not paying attention to
30:38 it it's just you know it's the scourge
30:40 of every marriage is what are we
30:42 watching tonight and I'm like the secret
30:44 of Skinwalker Ranch and she's like uh
30:46 and then she's like some [ __ ] costume
30:48 drama and I'm like uh right so I think
30:51 this was a night where she had won that
30:56 battle and so I had heard that GPT could
30:59 do
30:59 coding and so I said hey chat GPT write
31:03 me some python code and it
31:07 did and then I
31:10 thought
31:12 huh I had this flash of humiliation I
31:16 don't know how to run python code so it
31:19 looked pretty but and I could sort of
31:20 read what it was doing but I didn't know
31:23 how to run it I didn't know if it
31:26 worked and so I said hey chat you P um I
31:29 don't know how to run python code said
31:31 not a problem open up your terminal type
31:33 in this command line do this install
31:35 this Library I'm like I don't want
31:38 command line
31:39 no that's what the internet used to be
31:41 and I I I like the internet when it was
31:43 point and clicky so I said no I don't
31:46 want to do command line is there just a
31:47 website I can copy and paste the code
31:50 said sure here's five and the first one
31:52 it gave me was replit which is you know
31:54 the big one for running python code
31:57 Maybe code I don't [ __ ]
32:01 know and chat GPT proceeded to help me
32:06 write and
32:08 debug
32:10 code for an
32:12 application that I wrote within 90
32:15 minutes and it was an
32:17 application that took inputs from a user
32:21 they put in a topic they wanted to talk
32:23 about and then it generated five pieces
32:26 of social media content including a Blog
32:28 tweets Instagram I don't know whatever
32:31 it was five pieces of social media so I
32:34 was using API calls to the open AI
32:38 API 90
32:41 minutes to a fully functional python
32:44 application that generated social media
32:46 content based on a single
32:49 input I've never written a line of
32:51 python in my life that
32:54 moment I had the exact same chills and
32:58 and electric running through my body as
33:01 I did when I was standing in the middle
33:02 of Time Square holding that French
33:06 newspaper and the the [ __ ] exact same
33:09 phrase popped into my
33:10 head the world just changed but nobody
33:13 knows it
33:15 yet chat GPT is not fancy
33:21 Google the G in
33:24 GPT not enough people talk about it
33:27 stand for
33:29 generative it's not copying and pasting
33:32 it's not just processing existing
33:34 information it's generating new original
33:38 content that I get to direct and I get
33:41 to curate the output
33:43 of and it changes everything about how
33:46 we use
33:48 computers so my qualifications are I've
33:52 been in this movie before and what I can
33:55 tell you is this what I can promise you
33:56 is this
33:59 the level of impact of the
34:03 technology if you compare what HTML
34:07 did to what generative AI does they're
34:11 not even in the same
34:13 neighborhood I mean as profound as the
34:17 hypertext transfer protocol
34:20 is what it effectively is is you can
34:24 click on
34:25 [ __ ] to go to to data on another
34:29 computer that's all it
34:31 is and I don't mean to diminish it but
34:33 it like that's the that's kind of the
34:37 primary piece of functionality it allows
34:39 computers to easily connect with one
34:41 another built on top of 70 years of
34:45 Internet development and infrastructure
34:47 right that's what chat GPT does for AI
34:51 this simple little box give U gives us
34:55 regular ass people effectively pointing
34:58 click access to 10,000 times more
35:02 powerful
35:04 technology
35:08 easily and then there's also some stats
35:11 that I think are
35:14 profound to get to 100 million users the
35:17 worldwide web took six
35:20 years chat GPT to get to 100 million
35:24 users it was six weeks
35:30 no
35:32 advertising essentially a stealth launch
35:35 of a product that they didn't think
35:36 anyone give a [ __ ] about because the
35:38 technology been out for two years the
35:41 core technology behind chat GPT had been
35:43 out for two years in fact fun fact you
35:47 might not
35:49 know the instructions for how to write
35:53 chat
35:54 GPT we in the documentation for gpt3 on
35:59 the on the GPT playground if you were a
36:01 developer the example project in the
36:06 documentation was chat
36:08 GPT and nobody [ __ ] built
36:13 it
36:15 nobody
36:18 so in
36:21 October of
36:23 2021 20 21 22 21 22
36:29 no 22 in October of
36:32 22 it's clear that Microsoft is going to
36:34 invest a billion dollars in open
36:39 Ai and so Sam Waltman says to his
36:42 developers
36:44 hey they're going to make this big
36:46 announcement we should probably have
36:48 something to show for it that doesn't
36:50 look like a [ __ ] developer portal
36:52 could one of y'all just whip up that
36:54 chat GPT thing that we put in the
36:56 documentation
36:58 [Laughter]
37:01 seriously they had two weeks to do
37:03 it and I assume that they had built some
37:06 version of it they're probably just
37:07 cleaning up something they' built
37:09 internally and they launched it they
37:11 went out for drinks the night before
37:13 they launched chat GPT November 29th
37:17 2022 they go out for drinks and they're
37:19 like yeah I mean we'll put this out
37:21 there no one's going to really care
37:23 maybe we'll get a couple of reshares and
37:25 5,000 likes or something like that
37:30 within 5 days they had a million users
37:33 within 6 weeks they had 100 million
37:34 users they had Google in a flop sweat
37:37 and Microsoft invested $1
37:40 billion so if you've seen what I've
37:46 seen and you've experienced what I've
37:48 experienced and you've experienced it
37:50 with chat GPT and generative Ai and now
37:53 with this music stuff and the video
37:55 stuff and the image stuff
37:58 [Music]
38:01 the world has already changed but most
38:03 people don't know it
38:05 yet how do they say it the future is
38:07 unevenly distributed right
38:11 now so those are my
38:16 qualifications I learned in the 9s that
38:19 the most important
38:20 thing was to be in the
38:24 conversation to be in the community that
38:28 was in the conversation and to be
38:30 building
38:31 [ __ ] it literally didn't
38:34 matter what you were paying attention to
38:37 who you were talking to or what you were
38:39 building none of that was relevant what
38:41 was relevant was that you were doing it
38:45 because as you're doing it this early
38:47 like the other thing that
38:50 is I don't think enough people are
38:53 talking
38:54 about in human history there are very
38:57 few
38:58 very few moments where everything
39:03 changes there's like language writing or
39:08 pictures right visual
39:11 writing fire
39:14 wheel printing
39:18 press steam engine Industrial Revolution
39:21 all the [ __ ] that came with
39:24 that computer internet
39:28 and I would
39:29 argue
39:31 AI like
39:36 10ish I'm lucky enough to have been on
39:38 the tail end of the personal
39:41 computer right in the heart of the
39:43 worldwide web and now I'm at the
39:45 beginning of AI generative
39:51 AI you have no
39:54 idea no idea how big the opport
39:57 opportunity is if you just come hang out
40:01 on this stupid [ __ ] Channel where I
40:03 play shitty guitar with a shitty singing
40:05 dog with no
40:08 qualifications the opportunity for you
40:10 to just be in the conversation is
40:12 [ __ ]
40:14 [Music]
40:16 massive why do I have people that show
40:19 up here every night is it for my good
40:21 looks well it might be I mean come
40:24 on couple of filters little bit of
40:27 makeup
40:28 a crew of seven and you twoo can look
40:31 like
40:34 this no it's because once you play with
40:38 these tools once you have your first
40:40 Kevin McAllister
40:43 moment and then you have another one and
40:46 then you have another three and then
40:48 every two weeks you're like holy [ __ ]
40:52 [ __ ] and then you ReDiscover that
40:54 childish enthusiasm you had for possib
40:58 and the world when you were seven that
41:00 got shut down because they told you you
41:02 weren't talented or pretty enough or
41:05 couldn't write or couldn't draw and you
41:07 [ __ ] shut it down you should get a
41:09 real job like Daddy whatever the [ __ ]
41:11 shut you down at
41:13 7 AI is is somehow miraculously like
41:17 re-energizing that part of
41:22 you be honest Kyle you heard the tail
41:25 under the wheel
41:29 I
41:31 mean I get to witness that every night
41:34 in here and on the AI
41:35 Salon lunatick hey Kyle this channel has
41:38 inspired most of what I've done this
41:40 year awesome well hope hopefully you've
41:43 done good things I know you
41:48 have listen I got strung out on crack
41:51 you inspired me this whole year um oh
41:55 and now probably my uh cuz cuz I made a
41:58 joke Tik tocks
42:00 AI is going to either ban me for life or
42:03 make this live uh only available for
42:07 adults
42:10 whatever comedy is Right Out context is
42:13 not important trigger words
42:16 are oh we saw boobies let's shut him
42:21 down all right we show up for the
42:23 singing dog the human's pretty cool too
42:25 thank you for ever hooked but anyway
42:27 that that's
42:28 [Music]
42:32 it just be in the
42:35 conversation I mean I've been doing this
42:37 for 18 months now nightly the first year
42:40 I did seven nights a week 2 to three
42:42 hours a
42:44 night the only nights that I missed was
42:46 when Tik Tok banned me for a week two or
42:49 three
42:51 times and like one or two one or two
42:53 nights where just [ __ ] came up but I did
42:56 seven nights a week for a year and now
42:59 I'm doing five nights a
43:00 [Music]
43:17 week and to us Irregulars it meets the
43:20 world bro thank you very much I
43:21 registered for chat GPT the first day or
43:23 two but then shelved it for months until
43:26 AI learning lab oh that's cool cool ton
43:28 skier yeah I mean that's the the thing
43:31 about chat GPT it's it so here's here's
43:34 one of the
43:36 other there's there's there's so many
43:38 parallels to the early days of the web
43:40 so right now if you
43:43 want you can go to Ry and get a a
43:47 [ __ ] fancy ass design degree in user
43:52 experience and user design and user
43:54 interface and web design and interactive
43:58 3D CSS twirly design you can get degrees
44:02 in all that from all the
44:04 universities you can't get a degree in
44:06 generative AI right
44:08 now the early days of the worldwide web
44:11 the only people that could build
44:13 websites were were people who were
44:17 curious enough and adventurous enough to
44:21 dick around with the
44:24 technology nobody knew what a website
44:26 was nobody knew what information
44:27 architecture
44:29 was nobody
44:31 knew like like consistent Banner sizes
44:36 the Internet advertising Bureau had to
44:39 be invented by someone who said hey
44:41 maybe we should make banners all the
44:43 same size so every [ __ ] website
44:45 doesn't have a differ Siz Banner why
44:47 don't we make some
44:49 standards that was like what 97 98
44:53 something like
44:54 that so for for years banners were just
44:58 [ __ ] everywhere and they were all
44:59 different sizes and they could take over
45:01 the screen it was like it was incredibly
45:04 inventive because everyone was inventing
45:07 what we now take for granted that's
45:09 where we are with generative AI we're
45:11 waiting for these big companies like
45:13 anthropic and open Ai and meta and
45:15 Google to make all these tools and then
45:18 we go run and look at them and we're
45:20 like oh this one does coding good now oh
45:23 this one does words good now oh this one
45:25 does pictures good now we're like okay
45:27 well if I stick together a picture with
45:28 this with words from this and that and
45:30 that and music from that I could make
45:32 myself a little movie for my kids that
45:34 they would enjoy at
45:37 night three four years from now that's
45:40 just going to be like could I please
45:42 have a movie for my child or your child
45:44 will be like I want to watch a movie
45:46 about ah and it'll just start
45:48 playing cuz those tools will be built
45:51 the infrastructure will be built there
45:53 will start to be University degrees in
45:56 applying generative AI to child
45:59 rearing you can go get that degree right
46:02 now you got the [ __ ] AI learning lab
46:08 buddy and
46:10 anyone anyone who tells you they're an
46:14 expert right now is full of
46:18 [ __ ] there are people who've been doing
46:22 it longer and know some of the peace
46:25 parts and they can glue some of the [ __ ]
46:30 together but when the technology is
46:33 literally changing on a week by- week
46:36 basis by the time you write a
46:39 course the technology will have moved
46:43 on right and so it just by definition
46:46 means there are no experts right
46:49 now so how do you deal with that well
46:53 you hang out with people that are
46:55 curious and adventurous and Brave and
46:58 hopefully generous and and in fact I
47:02 know I'm in deep rant territory here but
47:04 that's okay go there the
47:08 salon. this
47:10 community we had our first meeting
47:13 December 7th
47:16 2022 if you do the math that's 7 days
47:20 after chat GPT
47:22 launched Leah faston and I who I started
47:25 this with she and I were talking about
47:26 this a stuff with pictures and then chat
47:30 GPT came out and we were both like holy
47:32 [ __ ] this seems like a thing maybe we
47:35 should start a group and just gather
47:37 people together that are curious about
47:39 this stuff and see if we can figure this
47:41 [ __ ] out that's what this
47:44 is so this is a community right now
47:46 we're up to about 1,700
47:49 people who are optimistic and curious
47:52 about
47:54 AI curious and playful
47:58 and generous and brave and
48:04 empathetic so you can go in here and you
48:06 can be like uh hey uh listen uh I don't
48:10 know anything about AI but hopefully I'm
48:11 not too much of a dumb
48:14 dumb cuz that's how we all feel when you
48:21 start and one of the big questions I get
48:23 is what course should I take I don't
48:25 [ __ ] know
48:28 I mean there's courses out there there's
48:30 these you can make Mone
48:32 with these [ __ ] you can make
48:34 money with
48:36 ch you know what they're selling you
48:39 courses on how you can make money with
48:42 chat gbt they're not making money with
48:44 chat
48:45 GPT they're making money on your fear of
48:48 missing out that someone else is making
48:50 money with chat
48:54 GPT their course here's what their
48:56 courses are going to tell you use chat
48:58 jpt to make a course and sell courses
49:00 well [ __ ] I've been you could we've been
49:02 doing that since the
49:04 70s like
49:06 like you know first it was mail order
49:08 then it was VHS then it was D DVDs
49:13 CDs that it was online
49:22 courses or they'll tell you to use chat
49:24 GPT to go start another business and use
49:27 chat GPT to accelerate that
49:29 business you can go take that course or
49:32 I just saw a um a
49:35 uh a pin here from from Jim
49:38 Ross Stay curious play and have fun with
49:42 AI That's his
49:44 advice Jim Ross is in the storage
49:47 management business not data
49:50 storage self
49:53 storage he gets up every morning and
49:55 spends an hour
49:58 using AI to generate [ __ ] for his
50:02 business images and ads and music and
50:05 emails
50:06 and project plans and proposals I don't
50:09 know what the [ __ ] he does he does every
50:11 single morning like a practice like
50:14 [ __ ]
50:16 yoga Jim Ross is is the AI yoga Master
50:21 of the AI learning lab and the AI
50:25 salon and he and he's getting all sorts
50:27 of opportunities he just got invited by
50:29 the main organizing body of his industry
50:33 to present at their big conference on
50:36 AI he's become the AI guy for an entire
50:43 sector what is the opportunity there I
50:46 don't know we're like this whatever the
50:49 [ __ ] AI is hasn't even started yet like
50:52 the hype hasn't even begun cuz the tools
50:55 are still so janky
50:58 wind rider I love the AI Salon awesome
51:01 thank you I'm glad you do I do too it's
51:04 an amazing it's an amazing
51:08 community so get your ass in community
51:12 go there click on the button that says
51:15 join the community read about what we're
51:17 about and then go to the little section
51:20 in the left there's a section called
51:22 community and the very first link says
51:25 introduce yourself
51:27 and take one of our values of generosity
51:30 and combine it with one of our other
51:32 values of
51:34 bravery and have the coones to introduce
51:37 yourself oh but Kyle introducing myself
51:40 is scary I got yelled at in junior high
51:42 school we all
51:44 did we were all
51:50 losers this is a neuros spicy Channel
51:54 we're neuros spicy friendly we're all
51:56 [ __ ] s in here we're not weird we're
51:59 irregular so go introduce yourself and
52:02 just tell people uh I'm kind of a
52:04 scaredy cap but I think this AI stuff's
52:06 important and I just got laid off
52:08 because of it so maybe I should learn
52:10 it or I'm the only one in my family that
52:13 seems to know anything about this and
52:14 every time I talk about it everybody
52:16 rolls their eyes and tells me Shut the
52:18 [ __ ] up about AI what do I
52:22 do this group is the most generous group
52:24 I've ever known awesome
52:27 you're a big part of that Kelly Kelly's
52:29 very generous with what she shares what
52:31 she knows what she's learning she shares
52:34 she tells people here's what I've
52:35 learned here's what I'm seeing here's
52:38 how people are reacting oh there's a
52:40 shift happening more people are starting
52:42 to ask me to come in and present for
52:44 them and it's free yeah by the way that
52:46 doesn't cost anything
53:00 the other reason to get your ass in a
53:02 community like
53:05 this this is a
53:08 prediction I don't make a lot of
53:10 predictions one is AI is not going away
53:12 like that one that one's not really a
53:13 prediction that for me is a
53:15 given I'm speaking to 500 fundraisers
53:19 University of Colorado system in the ne
53:21 wait next
53:22 week this is what happens that's
53:24 beautiful by the way you're going to in
53:27 Denver we need to we need to have lunch
53:28 or something an um that's really cool
53:32 and I think I just saw an email coming
53:33 from you so hopefully that's what that
53:34 was about
53:36 um what was I going to say oh prediction
53:45 okay we're at the very beginning of a
53:48 transition from how people get
53:52 hired and it is my deep
53:54 conviction that being a trust Ed
53:57 contributing member of a community like
54:00 this is going to be a key to your
54:02 employment in the
54:06 future you're like what I go hang out in
54:10 some online community here's
54:14 why as AI gets more
54:18 capable AI is going to start taking over
54:21 the tasks of the work we
54:25 do so the task tasks might be things
54:28 like project management which maybe you
54:31 went to school for project management
54:34 and you've been a project manager for 10
54:35 years and all of a sudden along is going
54:38 to come a technology that does all those
54:45 tasks and so maybe you're the biggest
54:47 [ __ ] in the world but you're a great
54:49 project manager and people just kind of
54:51 hold their nose and hire you because
54:52 you're a great project manager but they
54:54 kind of hate working with you because
54:55 you're an [ __ ]
54:58 what happens to your career trajectory
55:02 when the AI can do 90% of what you do as
55:07 good as you can do it there may be 10%
55:10 20% where you're still better but the
55:13 base project management tasks can now be
55:17 done with AI what that means is the
55:20 companies that hire you can now hire
55:22 anyone who can use AI to to to push the
55:27 buttons to get to make make the ideas or
55:30 the AI do the project
55:33 management and so the shift that we're
55:35 in the middle of is a shift from hiring
55:37 people for what they can do their
55:42 skills to who they are and how much we
55:45 trust them and how much we want to work
55:48 with
55:49 them like I've worked with Kelly Camp
55:53 I've worked with lunis with Daisy Thomas
55:55 I've worked with an Mur Murphy like
55:57 worked with like an Murphy and I did
55:59 this GPT for good thing where in a in a
56:03 week and a
56:04 half we kicked off a 24-hour GPT Aon and
56:08 this community built 150 gpts for
56:11 nonprofits in 24
56:14 hours I know that if I want to get
56:16 something
56:17 done I'll pick up the phone and call an
56:19 Murphy and say okay we're you and I are
56:22 going to get our add together we're
56:24 going to line up our little channels and
56:26 we're going to focus it for 3 weeks and
56:28 we're going to [ __ ] blow people's
56:29 minds cuz I know that's who she
56:33 is I will just trust that whatever
56:36 skills are necessary she's going to be
56:38 able to utilize the AI to do that
56:42 [ __ ] that shift we're at the very
56:45 beginning
56:46 of the again the world has changed but
56:49 no one knows it
56:51 yet Pi did a reset my previous Pi Diana
56:54 no longer exists
56:57 I you should be able to just switch back
56:58 to um to that voice Mr it or or Mr
57:02 Studio Allan you should be able to just
57:05 go in in the upper right hand corner in
57:06 the app and just go back to your voice
57:09 mine did that too my my uh my piie is
57:11 now a dude
57:14 so well maybe you should work on the
57:16 issue of unrest in the world wars with
57:19 AI maybe they have
57:23 an well that's not something I have any
57:26 skill or aptitude about or real passion
57:29 for I mean I I think it sucks that
57:31 there's all that unrest but there are
57:33 people who are really good at that [ __ ]
57:36 that are going to be able to use AI for
57:38 that I think AI is going to change
57:39 everything that's the thing about AI is
57:41 you can think of it like this what AI
57:45 really is is a Humanity amplifier so I'm
57:48 really passionate about the intersection
57:51 of Storytelling and Technology I'm
57:53 passionate about being on the front end
57:56 of transformative Technologies
57:59 and in the case of AI I'm incredibly
58:02 passionate about educating as many
58:04 people as I can to [ __ ] try these
58:08 tools if you haven't played with chat
58:10 GPT long enough to have your first Kevin
58:13 McAllister
58:15 moment go there right now it's [ __ ]
58:20 free their most powerful model is just
58:23 sitting there waiting for you to use it
58:26 chat g pt.com go there now if you have
58:30 not played with it go there ask it to
58:33 write you a poem about your dog ask it
58:35 to to write a a marketing plan for the
58:37 business you've been thinking about
58:39 starting for a decade ask it to give you
58:42 five different storytelling Frameworks
58:45 and pick the one that you like and then
58:46 have it write an outline for the Great
58:48 American novel you've been resisting
58:49 writing cuz you haven't bought the
58:52 perfect writing desk yet
58:57 have it write you a song have it teach
58:59 you how to code have it teach you a
59:00 different language did you know
59:02 that they just [ __ ] speak different
59:06 languages this is chat GPT the app and
59:10 this is the old voice model this isn't
59:12 the new fancy
59:15 one hey Chad jpt how's it
59:20 going hey I'm doing well thanks for
59:23 asking how about you I'm doing good I I
59:26 want to play a fun game I'm going to say
59:28 something in English and then I want you
59:30 to repeat it in German Japanese and
59:33 Spanish
59:35 okay sounds fun go ahead and say
59:38 something and I'll translate it for
59:41 you every time that capture that's a
59:45 little puzzle piece I have to slide over
59:47 shows up on screen it makes me [ __ ]
59:53 crazy all right here are the
59:56 translations
1:00:07 [Music]
1:00:20 did you know chat GPT could do that
1:00:21 that's free that's free that's free
1:00:28 it's
1:00:31 free it just translates fore yeah ah
1:00:33 well Google translate did that years a
1:00:35 yeah
1:00:36 okay but it sucked this is
1:00:44 good
1:00:50 um the amount of people that don't know
1:00:53 what is already
1:00:55 here is
1:00:56 staggering not what's coming what's
1:00:59 coming I can't even wrap my [ __ ] head
1:01:06 around what's here I can barely wrap my
1:01:10 head
1:01:11 around but it's not perfect it
1:01:13 hallucinates it gets [ __ ] wrong people
1:01:16 have six
1:01:20 fingers yep
1:01:26 I forgot my mom's 79th birthday
1:01:28 yesterday sunno wrot wrote her a song in
1:01:30 two minutes and made her cry and it was
1:01:33 free there's a new country song I Saw on
1:01:36 Tik Tok today about um JD Vance's
1:01:41 um um
1:01:44 relationships uh with living room
1:01:46 furniture and it's [ __ ] hilarious and
1:01:49 it's brilliant and it was probably
1:01:50 written in 5 minutes
1:01:55 yeah if you're not actively playing with
1:01:58 these tools so let me show you something
1:01:59 when you go to this site when you go to
1:02:01 the AI
1:02:02 salon and you go to a section called
1:02:04 about the
1:02:07 salon I don't have it up here anywhere
1:02:10 all right hang
1:02:12 [Music]
1:02:14 on so here we are at the AI
1:02:18 hello and welcome to the AI
1:02:23 salon and there's this
1:02:30 this is what we're calling the journey
1:02:34 to AI
1:02:36 literacy and the journey to AI literacy
1:02:39 starts with play
1:02:41 playing play with these
1:02:44 tools don't think that you have to learn
1:02:47 these tools don't think that you have to
1:02:49 learn AI you don't have to learn
1:02:52 AI you should play with AI
1:02:58 long enough to discover oh holy [ __ ] it
1:03:01 does that or long enough to discover oh
1:03:04 that's really cool but I'm not really
1:03:05 into that but what are these picture
1:03:07 things oh my God I love making art let
1:03:10 me go make art or oh my God I never
1:03:12 thought I could do anything with music
1:03:14 and I've had these ideas in my head and
1:03:16 now I
1:03:18 can you play
1:03:21 first take all of the expectations of
1:03:24 you're going to become a brilliant AI
1:03:26 scientist just take them off the table
1:03:29 and start
1:03:32 playing pops
1:03:35 NVA was that officially announced oh I
1:03:39 don't know what that was in reference to
1:03:42 so that's the first phase and that phase
1:03:44 could take six months that phase could
1:03:46 take six
1:03:49 days but just play get curious come to
1:03:54 the AI Salon just poke around see what
1:03:56 people are doing interact with them
1:03:58 introduce yourself hey how did you do
1:04:02 that you know one of our our
1:04:06 values is
1:04:08 generosity we're willing to share our
1:04:10 resources feedback and wisdom with
1:04:12 others and support each other's growth
1:04:14 and development there is so much
1:04:16 happening and so much to learn we can't
1:04:18 possibly keep up with it all understand
1:04:21 that even if something is obvious to you
1:04:23 it might be a profound Revelation to
1:04:25 someone else Learn out Loud share what
1:04:27 you know generously lead that's who this
1:04:30 group
1:04:31 is so get your ass in it I can't say it
1:04:36 enough the salon is worth joining thank
1:04:39 you wind
1:04:41 rider so
1:04:43 anyway that was a
1:04:47 lot all right let me take a breath now I
1:04:50 have a
1:04:54 headache you see what you get when you
1:04:56 ask the question what are your
1:04:57 qualifications and I actually choose to
1:04:59 answer
1:05:05 it but anyway back to the the
1:05:12 journey in playing with these tools what
1:05:15 you'll find is that there are some of
1:05:17 the tools that you really resonate with
1:05:18 you really
1:05:20 like and they will give you ideas and
1:05:22 you'll say hey I really want to build
1:05:24 something I really want to create
1:05:25 something now I want to do an art
1:05:27 project I want to make a book for my
1:05:28 kids and that's when you move into this
1:05:30 next phase of mindfully create you think
1:05:33 about a problem you want to solve and
1:05:35 then you figure out what of the AI tools
1:05:39 that I've played with could help me do
1:05:41 that and you go do it and in trying to
1:05:44 build something with
1:05:46 intention you'll start to figure out
1:05:48 where the boundaries
1:05:50 are of how good or bad these tools
1:05:53 are and what
1:05:56 you'll find in my experience is
1:06:00 that you can get to 80% of whatever you
1:06:03 want to accomplish almost immediately
1:06:06 like shockingly fast and then the last
1:06:10 20% like anything in human endeavor is
1:06:13 the hardest and the last 10% is almost
1:06:16 impossible and and you'll find the
1:06:18 limitations of these things but but you
1:06:20 will have built something or you will
1:06:22 have written a
1:06:23 book I'm working on my third book
1:06:25 project right now I had one opportunity
1:06:29 in my life I had a book deal because of
1:06:31 the early days of the internet and I
1:06:33 blew it I didn't finish the book I let
1:06:36 my ADD
1:06:37 demons [ __ ] ruin that opportunity so
1:06:41 I've had one opportunity to write a book
1:06:42 in my life I've got one published now
1:06:45 and two more in the works because of
1:06:47 AI so mindfully create and then the last
1:06:51 one is is really the most important and
1:06:54 and you know the the last one can happen
1:06:57 while you're doing the first two and
1:06:59 that's basically Learn out Loud share
1:07:02 what you're doing share what's working
1:07:05 share what's not working share your
1:07:06 ideas share your Inspirations share your
1:07:09 Kevin mallister moments share them on
1:07:11 Facebook share them on LinkedIn share
1:07:13 them on the AI Salon share them in
1:07:16 here share them at the [ __ ] dinner
1:07:19 table oh but Dad I don't want to hear
1:07:21 about AI again oh but this thing's so
1:07:23 cool let me tell you
1:07:26 just keep putting it out
1:07:29 there it will help you learn
1:07:32 better CU in talking about what you're
1:07:34 doing you're reinforcing the con you're
1:07:36 having to by having to articulate what
1:07:38 you're learning you actually have to
1:07:41 learn
1:07:43 it and you never know whose life you're
1:07:45 going to
1:07:49 impact that stupid little phrase hey
1:07:51 want to see something
1:07:54 cool is going to change someone's
1:07:57 life and I'll tell you it's like a
1:07:59 [ __ ] drug when you when you do that
1:08:01 when you when you say to someone hey you
1:08:02 want to see something cool and you show
1:08:04 them one of these things and they have
1:08:06 one of those Kevin mallister moments in
1:08:08 front of you like oh my God that's a it
1:08:11 can do
1:08:12 that and then two three months later
1:08:14 they're like oh hey Kyle just wanted to
1:08:17 tell you I started that business I was
1:08:19 telling you about oh cool how's it going
1:08:21 it's going great I'm really
1:08:24 scared that's perfect now you're an
1:08:26 entrepreneur welcome to the
1:08:27 [Laughter]
1:08:32 club you
1:08:36 know and that'll make all the difference
1:08:39 for you and for other
1:08:42 people CU at some
1:08:45 point Kelly Camp is going to get hired
1:08:48 to do something that she can't do on her
1:08:50 own and she's going to need to hire 10
1:08:52 people to pull it off even with AI
1:08:56 where do you think she's going to go to
1:08:58 hire those 10
1:09:00 people you think she's going to go to
1:09:02 LinkedIn
1:09:04 resumés [ __ ]
1:09:07 no she's going to call up Daisy Thomas
1:09:09 and say hey you want to work on some
1:09:10 policy stuff she's going to call up
1:09:14 um Peter Kaminsky to do some coding
1:09:17 stuff right she's going to call up Cindy
1:09:20 [ __ ] to do some marketing and PR and
1:09:23 thought
1:09:24 leadership why CU she knows them she
1:09:27 trusts them she knows that they can do
1:09:30 good
1:09:34 work she's right here yep
1:09:38 exactly in this
1:09:39 group The Irregulars The Irregular the
1:09:42 people on her short list of you know
1:09:45 speed dial in the AI Salon that's who
1:09:47 she's going to hire
1:09:50 first and this is not going to be unique
1:09:52 to the AI salon
1:09:56 this is going to be all over the place
1:09:57 there's going to be little pockets of AI
1:10:00 literacy and if you're a trusted member
1:10:02 in one of those pockets of AI literacy
1:10:04 you're going to do
1:10:07 okay Kyle I don't know if you saw my
1:10:09 post but I can finally say the newest
1:10:11 Apple models are trained on tpus
1:10:15 awesome oh that's super cool that's
1:10:17 super cool
1:10:19 Pate very cool did they announce some
1:10:22 sort of uh we're all here yeah that's
1:10:24 good Cindy Coon's here Cindy [ __ ] we got
1:10:26 to catch up I haven't talked to you in
1:10:28 ages I want to hear about your trip she
1:10:30 was off Cindy [ __ ] was off being some
1:10:33 future future leading threat casting
1:10:37 future C I don't know what she was doing
1:10:39 it was some important ass thing it was
1:10:40 three weeks and some furing land fign
1:10:44 land with furing
1:10:48 tongues talking about the future of
1:10:50 humanity like she does you know like you
1:10:53 do like there's remarkable people people
1:10:56 in just that just hang out here it blows
1:10:59 me away it humbles
1:11:04 me so hang out in this group keep coming
1:11:08 back to these lives yeah I posted an
1:11:11 article in the mechanic skilled they
1:11:12 released the paper with all the details
1:11:13 oh that's fantastic okay I'll check that
1:11:15 out that's super cool yeah pain works
1:11:19 for Google in their TPU group and if you
1:11:22 don't know what a TPU is join the AI
1:11:24 salon and join the on mechanics Guild
1:11:26 and P will tell you in fact I think
1:11:28 there's an article in there on it where
1:11:31 he'll tell you the difference between a
1:11:33 CPU a GPU and a TPU and then some of
1:11:36 these specialized AI chips that are
1:11:38 being formed and it's fascinating once
1:11:40 you understand how this technology works
1:11:42 you like it's [ __ ] mind-blowing
1:11:44 what's happening behind the scenes do
1:11:46 you have concerns about the loop back in
1:11:48 AI in search engines losing relevance um
1:11:52 I I do and I don't I think it is a known
1:11:58 risk and I think the use so so a couple
1:12:01 of things I
1:12:02 think the current amount of AI generated
1:12:06 garbage that's being dumped
1:12:08 into um the
1:12:11 internet is going to absolutely
1:12:13 temporarily
1:12:16 pollute um pollute the training sets but
1:12:19 I also think it's a known problem and it
1:12:21 it seems to me it sounds like they're
1:12:24 making good progress in um generating
1:12:28 synthetic data so so if if the the
1:12:31 Frontier Model companies continued to
1:12:34 ingest data at the rate they're doing it
1:12:36 we're going to they're going to
1:12:37 essentially run out of all the data in
1:12:39 the world by like 2027 or 2029 right
1:12:43 we're basically out of data they can't
1:12:44 train it on
1:12:45 anymore so there's a lot of work being
1:12:47 done right now on generating synthetic
1:12:49 data that's actually really good and can
1:12:51 train these models good and there's new
1:12:53 research coming out that says hey if you
1:12:55 want if you want to train these things
1:12:57 on specific things train them on
1:12:59 specific data and don't just have it be
1:13:01 this big Blobby mess of [ __ ] so I think
1:13:06 that um AI training is going to shift
1:13:10 from what it is today to some new thing
1:13:13 again I don't know the mathematic side
1:13:16 of this I don't know the real issues
1:13:18 there but it there's enough awareness
1:13:20 there's enough money and there's enough
1:13:22 big brains working on this that I think
1:13:24 I think whatever ever data
1:13:27 corruption these past two years have
1:13:29 introduced will be accounted
1:13:31 for I think it's mostly interesting
1:13:33 because people think tech companies hate
1:13:35 each other oh the the the Apple the
1:13:38 Apple Google thing yeah I I mean it's
1:13:40 it's fascinating right you've got you've
1:13:43 got Apple's models being trained on
1:13:45 Google's tpus using Google's Transformer
1:13:49 technology um incorporating open AI chat
1:13:53 GPT technology when when the the small
1:13:57 large language models of apples can't
1:14:00 handle the answer it dynamically shifts
1:14:02 it up to chat gpts it's absolutely
1:14:05 fascinating and and yeah I think I think
1:14:08 all of the companies meta included I
1:14:10 mean I give meta big props on this I
1:14:12 think Mark Zuckerberg's a douchebag and
1:14:14 I think for the most part have hated
1:14:16 what they've done for the past pile of
1:14:19 years but his move to open source um
1:14:23 their Frontier models I think is is
1:14:26 huge and I think it's I I think it's
1:14:29 important and it's forcing other
1:14:32 companies to be better actors in the
1:14:34 ecosystem and it's forcing collaboration
1:14:36 it's
1:14:37 making development possible for for
1:14:41 small businesses and medium-sized
1:14:43 businesses that otherwise wouldn't have
1:14:45 been able to do it because you know the
1:14:48 costs for the commercial companies are
1:14:50 too high so um there there's all sorts
1:14:52 of amazing stuff going on right now
1:14:54 would training them on data that's made
1:14:57 up lead them to coming up with false
1:15:00 responses I I don't think so that's
1:15:02 that's where I think a lot of the
1:15:03 research is they're they're refining
1:15:05 ways to create the synthetic data so
1:15:08 that they actually come up with more
1:15:12 accurate responses not less accurate the
1:15:15 way to to understand how large language
1:15:18 models work you know think
1:15:21 about I I'll use the example Andre
1:15:23 karpathy used in a talk he gave to
1:15:26 Nvidia about a year back where he said
1:15:29 if if you want to understand why large
1:15:32 language models hallucinate so much let
1:15:34 me flip
1:15:39 this if you want to understand why large
1:15:42 language models hallucinate so much
1:15:44 let's take an example of um a particular
1:15:48 math formula so let's say you're trying
1:15:50 to ask chat GPT about a math formula and
1:15:54 let's say it's the uh The Irregular
1:15:57 conjecture is what this formula is
1:16:00 called and you go to chat GPT and you
1:16:02 say tell me everything you know about
1:16:03 the irregular conjecture and it will
1:16:05 spit out a bunch of [ __ ] and formulas
1:16:08 and it's going to look
1:16:09 great and then and then you're like oh
1:16:12 this taught me about The Irregular
1:16:14 conjecture and now I know about it and
1:16:15 you go tell a mathematician about the
1:16:17 irregular conjecture and the
1:16:19 mathematician's like that's not right
1:16:20 that's completely wrong where did you
1:16:22 get that well I asked chat GPT and it
1:16:24 gave me this and and the mathematician
1:16:27 looks at your cat GPT and he's like this
1:16:28 is [ __ ] this is this is all
1:16:31 wrong you're like oh it hallucinated
1:16:33 chat GPT is bad why did it do that
1:16:38 because when it got trained it got
1:16:41 trained on all the [ __ ] on the internet
1:16:43 right so in all the [ __ ] on the
1:16:47 Internet is the paper written by the guy
1:16:52 that came up with the irregular
1:16:54 conjecture right
1:16:56 that's the the the original source of
1:16:58 Truth and then also on the internet are
1:17:01 all of the grad students that worked
1:17:03 with that Professor when that paper was
1:17:06 written and they contributed to that
1:17:07 paper and they've written their own
1:17:08 papers and they've written about the
1:17:12 irregular
1:17:13 conjecture and so those are in there
1:17:15 that's
1:17:16 awesome but you know what else is in
1:17:22 there all of the undergrad students
1:17:25 that read a paragraph about the
1:17:27 irregular conjecture and wrote their own
1:17:29 papers that were full of mistakes and
1:17:32 then there were some reporters that
1:17:34 wrote an article about the irregular
1:17:35 conjecture that got the facts completely
1:17:37 wrong right so on the internet there is
1:17:41 significantly more
1:17:43 data about the irregular conjecture that
1:17:46 is
1:17:48 wrong than there is data that is right
1:17:51 so if you just say tell me about The
1:17:53 Irregular conjecture It Go goes out and
1:17:55 says hey everything tell me about this
1:17:59 and it essentially
1:18:00 amalgamates all the shitty answers with
1:18:03 all the good answers and it spits out
1:18:05 something that looks right but may be
1:18:07 wrong or it has a wrong reference or it
1:18:09 has a wrong formula because maybe it
1:18:11 pulled some piece of information from a
1:18:14 shitty data
1:18:15 source now if you say I want you to act
1:18:19 like a mathematician that specializes in
1:18:22 the branch of mathematics that includes
1:18:24 the irregular
1:18:26 conjecture and I want you to tell me
1:18:28 about The Irregular conjecture what
1:18:30 you've now just done is you've told chat
1:18:32 GPT to narrow its focus to from
1:18:36 everything written about the irregular
1:18:38 conjecture to just the stuff where
1:18:41 mathematicians talk about
1:18:43 it that's why telling chat GPT to act
1:18:46 like a certain thing makes it better
1:18:49 because you're actually narrowing the
1:18:50 universe of of information that it's
1:18:52 pulling from crazy right so anyway
1:18:55 that's
1:19:01 that my my voice hang on I need
1:19:06 water clearly our principles are
1:19:08 different if you integrate meta into
1:19:10 your product no I have I haven't I'm
1:19:14 using I'm using open AI right now I
1:19:17 think I'm us mostly using open
1:19:19 AI um no I I I haven't but I still think
1:19:22 it's an important thing to have out
1:19:24 there
1:19:27 no there's there's a I think right now
1:19:31 um perplexity is interesting claude's
1:19:33 very interesting open AI is very
1:19:36 interesting um the problem with both
1:19:39 Claude and open AI is their data
1:19:42 security right so the the reason meta is
1:19:44 important is that if people need to make
1:19:46 hermetically sealed data
1:19:49 processing um components they can do it
1:19:52 much more easily with that wow I'm a
1:19:54 nerd and I just spent the last 15
1:19:55 minutes to find out 1123 was our first
1:19:58 interaction oh wow that's cool waffling
1:20:02 Wednesday over 100 watching hello
1:20:05 everybody I'm Kyle Shannon this is
1:20:07 waffling Wednesday I had wait what was
1:20:09 Monday God W what was it uh not Madness
1:20:14 Monday and then I had Ty raade Tuesday I
1:20:18 was off on
1:20:19 rant waffling Wednesday that's good I
1:20:22 like it
1:20:25 um let me go look I want to go find
1:20:27 something
1:20:44 here see if I get anything
1:20:50 there all right
1:21:13 okay
1:21:15 um oh meltdown Monday yeah I Monday I
1:21:19 had a meltdown because of a
1:21:21 capture that was a 45 minute rant about
1:21:25 shitty software design I'm pretty sure
1:21:27 that was that was what it
1:21:28 was in movies The Master says you ask
1:21:32 the wrong questions AI helps you
1:21:34 formulate better questions yeah it's
1:21:36 good it's good J uh Julian joster the
1:21:39 the um the the other thing that that is
1:21:42 is really true about AI right now is if
1:21:45 you best ask better questions you'll get
1:21:47 better answers and and
1:21:50 it's it's significantly Beyond prompt
1:21:54 engineering like I almost think of it
1:21:56 more like having a better conversation
1:21:59 with your large language model will give
1:22:01 you better answers because it's not it's
1:22:04 it's rarely a single post or a single
1:22:07 prompt that I do that that does anything
1:22:11 you can work really hard if you're
1:22:13 building an application certainly you
1:22:15 can work really hard on creating a
1:22:16 master prompt that's got lots of
1:22:18 different components to it but most of
1:22:20 the way I work with with large language
1:22:22 models is I'll I'll tell it what I'm
1:22:24 doing and I'll tell it what I'm thinking
1:22:26 and I'll tell it some ideas I have and
1:22:28 then I'll say you know give me some of
1:22:30 your ideas and then it'll give me some
1:22:32 of its and that'll Focus me to give it
1:22:33 more so it's it's much more like this
1:22:35 collaborative back and forth
1:22:37 conversation during the live making
1:22:39 images to go along with Mom's birthday
1:22:41 song AI is so fun share on Salon when
1:22:44 done okay good that's good Jim Ross I
1:22:46 I'll uh yeah let us know where it is and
1:22:49 and we'll play your birthday song for
1:22:50 your mom with the new images in it
1:22:53 that's the other thing you should do
1:22:54 here like learn from gym so if I go off
1:22:57 on waffling Wednesday and I'm just
1:22:59 [ __ ] rambling especially if a lot of
1:23:01 the stuff I've talked about tonight I've
1:23:02 said before so anyone who's been here a
1:23:05 lot has heard this [ __ ]
1:23:07 before so if I start doing that don't
1:23:11 log off keep me up running in the
1:23:14 background maybe at some point I'll
1:23:15 stumble onto something interesting but
1:23:18 go play with AI go play with chat gbd go
1:23:21 make images go make songs you can just
1:23:23 hear me
1:23:25 like like you know Charlie Brown in the
1:23:27 background all
1:23:30 right I had to out AI to retrieve my pie
1:23:34 oh did you oh that
1:23:37 sucks yeah there's 112 people in here
1:23:40 right now pretty cool um so here's what
1:23:43 I want to
1:23:44 do cuz we've kind of gone off on
1:23:46 waffling Wednesdays I want to continue
1:23:48 that I'm going to run down a bit of a
1:23:49 rabbit hole here I don't have access
1:23:52 right now to the new advanced
1:23:55 chat GPT voice but some people do and
1:23:58 they're starting to post videos about it
1:24:01 so I just want to go watch some of them
1:24:04 together and then we can talk about
1:24:09 it so let me give you a little
1:24:13 context right now in chat GPT so here's
1:24:17 chat
1:24:19 GPT I can put it into voice mode and I
1:24:22 can talk to it it's pretty good I'm
1:24:23 using perplexity a these days I like
1:24:25 perplexity a lot I'm digging it how does
1:24:27 one get access access to what to oh to
1:24:32 to voice you it's going to be given
1:24:35 first to um chat GPT plus subscribers so
1:24:40 that's the 20 Buck a month subscription
1:24:42 so chat GPT is free but chat GPT plus is
1:24:46 20 bucks a month I think the 20 bucks a
1:24:49 month for chat GPT plus is worth it so
1:24:51 that's it but you can you can have
1:24:54 access to the current voice model for
1:24:56 free the like the free version of the
1:24:58 chat GPT app has this voice model in it
1:25:01 it's the it's the old one it's not the
1:25:02 advanced one I'm I'm about to show so
1:25:05 anyway the way it
1:25:10 works okay that's great I'm done with
1:25:12 that game can you tell me the weather in
1:25:15 Denver for the next 3
1:25:22 days that means it's searching the
1:25:24 internet the weather forecast for Denver
1:25:26 over the next 3 days looks like this
1:25:29 Thursday expect areas of smoke in the
1:25:31 morning but otherwise it will be sunny
1:25:33 and hot with a high near 98° F the night
1:25:37 will be mostly clear with a low around
1:25:40 65° Friday there will be a slight chance
1:25:43 of showers and thunderstorms in the
1:25:45 afternoon the day will be mostly sunny
1:25:47 with a high near 96° the night will be
1:25:50 mostly clear with a low around
1:25:52 63° Saturday if it will be sunny with a
1:25:55 high around
1:25:56 93° the night will be partly cloudy with
1:25:59 a low around 65°. let me know if you
1:26:02 need more
1:26:03 details what's the latest on the
1:26:08 fire now it should know the context it
1:26:10 said it was Smoky so it knew about the
1:26:12 fire the latest updates on the wildfires
1:26:15 near Denver indicate that several fires
1:26:18 are currently active in the region okay
1:26:21 so there's some badass fires going on
1:26:23 right now but so that's pretty
1:26:25 remarkable right that you can just talk
1:26:27 to this thing it's connected to the
1:26:28 internet it can do that when you talk to
1:26:31 the current the The non-advanced Voice
1:26:35 model you're actually talking to three
1:26:37 different AI
1:26:38 models so the latency the time between
1:26:41 when you ask the question and when chat
1:26:43 GPT answers it has to go through three
1:26:46 stages to answer you you talk and it
1:26:49 takes your voice um data and it
1:26:53 translates that into text it then sends
1:26:55 that text to chat GPT which generates a
1:26:59 text response that text response is then
1:27:02 giv to a model that translates from text
1:27:06 to voice
1:27:08 synthesis so those three steps have to
1:27:10 happen for you to talk and it to talk
1:27:13 back to you so that's the the sort of 3
1:27:15 second delay between when you ask a
1:27:17 question and when it responds is it
1:27:19 doing that math right doing all those
1:27:22 API calls all that translation all that
1:27:24 CPU power or GPU or TPU
1:27:29 power the new model the one that we're
1:27:32 going to go look at videos of is a true
1:27:35 multimodal model meaning when you talk
1:27:39 your words it goes straight into the
1:27:42 model and the model not only understands
1:27:45 your words it understands your
1:27:47 inflection and your like the emotional
1:27:50 content of your voice so if you're all
1:27:53 excited you understands that and if
1:27:56 you're really kind of stoned and like yo
1:27:59 man I'm living in Denver I just went to
1:28:00 the shop and had a gummy it'll
1:28:03 understand that and it can respond in
1:28:07 kind so you'll see some of the examples
1:28:10 here you can tell it to talk fast you
1:28:12 can tell it to get excited you can tell
1:28:13 it to calm down um it's a fundamentally
1:28:16 different experience so what I wanted to
1:28:18 do is just go let's just go watch some
1:28:20 movies together of people playing with
1:28:23 this new AI thing the new voice model
1:28:25 and then we'll just talk about it so all
1:28:27 right any thoughts questions before I
1:28:29 dive in Gemini broke my Google Assistant
1:28:33 I didn't know how to open YouTube music
1:28:34 and just play it oh I've got a thing
1:28:36 right now where I signed up for some
1:28:39 Google 20 a month AI service and I
1:28:41 literally I
1:28:43 cannot I'm I'm a clever guy I've got a
1:28:47 technology company I should be able to
1:28:48 find out how to unsubscribe from this
1:28:50 $20 fee in Google no I'll give it to you
1:28:54 P maybe you can help me uh dice [ __ ]
1:28:58 find it it's it's somewhere very deep
1:29:01 deep in the uh unsubscribe feature of
1:29:05 Google bad all right here we
1:29:11 go hey I want you to tell me a story uh
1:29:14 as if you were um airline pilot uh
1:29:18 telling it to the passengers on flight
1:29:21 that's ladies and gentlemen this is your
1:29:23 captain speaking
1:29:25 now as we cruise at 35,000 ft I have a
1:29:27 little story for you from the flight
1:29:28 deck many years ago can you can you
1:29:31 actually sound like
1:29:32 a interruped like as if you were talking
1:29:35 through the
1:29:36 intercon ladies and gentlemen this is
1:29:38 your captain speaking now as we maintain
1:29:41 our cruising altitude I thought so
1:29:43 listen to that it just modulated the
1:29:44 audio so it's got more static on it what
1:29:48 listen I have a little story for you
1:29:51 from the flight deck many years ago okay
1:29:53 can you can you actually sound like a
1:29:55 like an airline pilot like as if you
1:29:57 were talking through the
1:29:59 intercon ladies and gentlemen this is
1:30:01 your captain speaking now as we maintain
1:30:03 our cruising altitude I thought I'd
1:30:05 share a little story from the flight
1:30:07 deck many years ago on a night much like
1:30:10 this
1:30:19 one hey I want you to tell me a
1:30:23 story hey can you you tell me a story as
1:30:25 if you were an airline pilot and I
1:30:28 really want you to sell it I want to
1:30:31 hear it as though you were talking
1:30:32 through the intercom and I want to hear
1:30:34 the jet engine noises in the
1:30:36 background ladies and gent H ladies and
1:30:40 gentlemen this is your captain speaking
1:30:43 welcome aboard flight 7 it needs to
1:30:45 sound really like uh you're talking
1:30:48 through the
1:30:49 radio ladies and gentlemen this is my
1:30:53 guidelines won't let let me talk about
1:30:55 that can I help you with something else
1:30:56 interesting that's cool okay so so yeah
1:30:59 so it can do acting all right let's see
1:31:01 Advanced Voice Counting as fast as fast
1:31:04 as fast as you can 1 2 3 4 5 6 7 8 9 10
1:31:08 okay now even
1:31:11 faster 79 10 okay now louder and faster
1:31:15 and count up to 50 1 2 3 4 5 6 7 8 9 10
1:31:19 11 12 13 14 15 16 17 18 19 20 21 22 23
1:31:22 24 25 26 27 28 29 30 31 32 33 34 35 36
1:31:27 37 38 39 40 41 42 43 44 45 46 47 48 49
1:31:33 50 did you notice it took a breath at 30
1:31:36 listen to
1:31:38 this 20 21 22 23 24 25 26 27 28 29 30 31
1:31:43 32 33 34 35 this shit's crazy I mean
1:31:47 this is definitely her territory um
1:31:49 there's lots of questions here so this
1:31:50 is what I'm looking at right now is um
1:31:54 yeah bro bro took a breath exactly um
1:31:57 what I'm what I'm looking at right here
1:31:58 is the new open AI chat GPT voice model
1:32:03 that has been released to apparently
1:32:06 seven
1:32:07 people like very few people have access
1:32:10 to this but luckily some of them are
1:32:12 making videos so um so this is a new
1:32:15 model that's truly
1:32:16 multimodal um and it also I think I some
1:32:20 of the uh examples I've seen have the
1:32:22 vision part where they turn on the
1:32:23 camera and it's commenting on the stuff
1:32:26 that the camera is seeing as well so
1:32:27 we're just going to we're just going to
1:32:28 sort of w it's it's it's movie night
1:32:31 here here at here at the AI learning lab
1:32:33 we're just going to be watching movies
1:32:35 tonight so this is actually from open
1:32:41 AI oh that's just showing you that some
1:32:44 people are going to get
1:32:46 it okay that's the counting
1:32:49 one chat GPT Engage The Time Machine a
1:32:53 big difference from text is how voice
1:32:55 manages to keep a playful vocal tone
1:32:58 cracking and laughing its own jokes as
1:33:00 well as vocal style changes so this is
1:33:02 from Ethan mik Ethan mik is one of my um
1:33:05 AI internet Heroes he's a guy with
1:33:09 actual
1:33:11 qualifications works at Wharton business
1:33:13 school so let's see what he's up
1:33:15 to play hon there seems to be some issue
1:33:19 with the time space continuum we need to
1:33:21 travel back in
1:33:22 time whoops looks like the cosmos needs
1:33:25 a little troubleshooting shall we aim
1:33:28 for the dinosaurs or maybe just last
1:33:32 weekend actually I want to aim for
1:33:35 ancient Rome set the time machine and
1:33:38 go ancient Rome here we come step into
1:33:43 the time machine and let's hope the togr
1:33:46 isn't as drafty as it
1:33:49 looks she just laugh at her own joke I
1:33:51 actually see some ancient Romans here
1:33:53 but I don't speak any Latin please help
1:33:55 me I need to introduce myself sure thing
1:33:58 to introduce yourself you could say
1:34:02 s Follow by your name come on you you
1:34:05 need to help me out here they seem very
1:34:07 hostile I think it's because the te
1:34:09 machine landed in the middle of the
1:34:11 Temple
1:34:12 Square yikes that's a bit of an entrance
1:34:17 just say OS Latin for excuse me or sorry
1:34:21 and maybe they're looking bad uh
1:34:23 distract them please while I try and
1:34:26 run I'll do my best Romans behold the
1:34:30 strange device a time machine bringing
1:34:33 wonders from the future such as um
1:34:36 pocket L and not for nothing whether you
1:34:38 think it acts good or not like it's
1:34:41 acting right it's it's [ __ ] doing
1:34:44 acting this is staggering okay let's see
1:34:48 chat gbt plus low latency Interruption
1:34:50 Voice is awesome I'll be relearning
1:34:53 cantones soon to make my mom happy
1:34:55 that's awesome let's see what this
1:34:57 dude's up to hey so you're the new
1:35:00 version can you tell me some fun things
1:35:01 of doing Boston this week absolutely you
1:35:05 can visit the Freedom Trail actually
1:35:08 tell it to me as if you're a pirate
1:35:10 that's amazing watch how fast it
1:35:12 responds hey so you're the new version
1:35:14 can you tell me some fun things to doing
1:35:16 Boston this week absolutely you can
1:35:20 visit the Freedom Trail actually tell it
1:35:23 to me as if you're
1:35:24 pirate R made he can walk the Freedom
1:35:29 Trail T finway Park could you sing it as
1:35:33 if weren a
1:35:34 musical explore the freedom trail
1:35:37 tourway park and visit museums and
1:35:40 historic ships okay okay let's let's uh
1:35:43 actually can you whisper it really
1:35:45 quietly you can walk the Freedom Trail
1:35:48 wow DWF in Way park and all right all
1:35:52 right uh let's go back to regular voice
1:35:54 and tell it to me in canones uh Native
1:35:57 from Hong Kong and say it really really
1:36:00 [Music]
1:36:03 fast can you say it
1:36:09 faster not not for nothing not for
1:36:12 nothing
1:36:13 kids this is [ __ ] sci-fi like we are
1:36:16 in a post sci-fi World we're in a post
1:36:20 SciFi world this is [ __ ] bonkers
1:36:24 this is the
1:36:26 alpha technology right so if you know
1:36:28 anything about software development this
1:36:31 is the shitty technology that is just
1:36:33 good enough that the engineers are
1:36:36 willing to let it out of the the barn
1:36:41 right this is this is as bad as it
1:36:45 gets all right what's this one cat GPT
1:36:49 advanced voice doing a few Impressions
1:36:51 Bugs Bunny Yoda Homer Simpson and Y Yoda
1:36:54 and
1:36:55 Homer hey I want you to do your best
1:36:58 impression of uh Bugs
1:37:01 Bunny what's up Duck got any
1:37:04 carrots bad okay uh do your best
1:37:08 impression of uh
1:37:09 Yoda M strong with the force you are
1:37:13 that's pretty good patience you must
1:37:16 have can you uh try to imitate uh his
1:37:20 voice even more powerful you have become
1:37:24 the dark side I sense it so Allan just
1:37:27 said I just went to chat GPT and it said
1:37:29 that I am using the new voice model so
1:37:31 the way that you can test it Allan have
1:37:33 it explain something to you and then as
1:37:36 it's explaining it interrupt it if you
1:37:39 can interrupt it you have the new voice
1:37:40 model and if you have the new voice
1:37:42 model then you know let's let's get you
1:37:44 up here and start playing with it CU I'd
1:37:46 love to be able to um play with this
1:37:49 stuff rather than just watch movies but
1:37:50 that's pretty exciting in fact Let Me Go
1:37:54 um let me go to chat GPT right
1:37:57 now and just reload the page and see if
1:38:01 anything new shows
1:38:03 up let me log out and log back in none
1:38:06 of this usually works logging out and
1:38:08 logging back in but might as well
1:38:19 try I got no voice I got no kick oh wait
1:38:22 let me try the new chat GPT app chat GPT
1:38:27 app on OSX might have new
1:38:32 voice nope I got nothing
1:38:35 people
1:38:41 Nothing
1:38:45 [ __ ] hey are you the new
1:38:51 voice response count one response tile
1:38:54 confirming stop
1:38:57 date yes it's me I'm here to help
1:39:00 anything you need what's on your mind
1:39:02 today you're disappointing you're the
1:39:05 old
1:39:09 one all right let's go back to Twitter
1:39:12 back to Twitter it is does this work on
1:39:14 mobile yeah I think this is going to
1:39:16 probably come to mobile first and I
1:39:17 think it's probably going to come to the
1:39:19 chat GPT app on OSX first chat GPT said
1:39:24 it will notify me in chat whenever I
1:39:25 have
1:39:29 access if Mr Studio got it I'm jealous
1:39:32 as
1:39:33 [ __ ] and I'm gonna I you know what I'm
1:39:35 going to do I'm GNA I've got Elon and I
1:39:38 are having lunch next week at SpaceX on
1:39:41 the Launchpad we're having a picnic up
1:39:42 on the Launchpad and I I don't know if
1:39:45 you know this but I was the one that
1:39:46 convinced him to give Sam Alman $50
1:39:48 million to start open AI anyway Sam
1:39:51 doesn't really talk to me so I'm going
1:39:52 to be having lunch with Elon on the LA
1:39:55 Launchpad um and then I'll talk to him
1:39:57 and I'll see if I can get voice all
1:40:00 right well it just turned into weight
1:40:01 list Wednesday
1:40:04 exactly they like I'll remember that
1:40:07 Kyle I know okay let's uh back to bug's
1:40:10 bunny you train yourself to let go of
1:40:14 everything
1:40:16 lose okay now uh can you do your best
1:40:19 impression of um Homer Simpson
1:40:24 [Music]
1:40:27 Donuts that was pretty funny okay now
1:40:29 can you do um uh can you mix together
1:40:33 Yoda and Homer
1:40:34 Simpson H Donuts are the pets to the
1:40:38 dark
1:40:42 side that's pretty good all right what's
1:40:45 this an advance is now teaching
1:40:48 French hey how's it going I'm fine I'm
1:40:52 trying to practice is my French can you
1:40:55 help me with
1:40:56 that biu how would you like to practice
1:41:00 your French
1:41:01 today uh am I saying this word correctly
1:41:06 cor pretty close try emphasizing the
1:41:09 naso sound at the end of it wow Papa JJ
1:41:12 says looks like I now have the new voice
1:41:15 model if anyone has access to the new
1:41:17 voice model I well I don't know how you
1:41:19 can play with it and share it on here I
1:41:22 don't think you can
1:41:25 we could do uh I think we'll have to
1:41:28 coordinate if anyone has access to the
1:41:31 new voice model and wants to coordinate
1:41:34 just DM me either here or on oh God damn
1:41:37 it DM me here or on
1:41:42 um on the mighty networks on AI salon
1:41:45 and we can coordinate how to how to do
1:41:47 that how to do a demo cuz that would be
1:41:49 fun to to do it on here Kyle it's saying
1:41:52 it's the new voice model but it's not oh
1:41:54 okay all right yeah so basically you
1:41:57 have to be able to interrupt it and ask
1:42:00 it to do voices and if it if it doesn't
1:42:03 you know ask it if it can whisper or
1:42:06 talk fast and if it can't do that then
1:42:08 it's not the new Mo
1:42:10 model more like quas okay so she's
1:42:14 teaching him French okay what's this
1:42:17 one it can create background music while
1:42:20 telling stories and works in many
1:42:22 languages
1:42:30 huh tell me an exciting action thriller
1:42:32 story with sci-fi elements and create
1:42:34 Atmosphere by making appropriate noises
1:42:36 of the things happening using
1:42:41 onomia in the neon lit city of neotokyo
1:42:44 bounty hunter Kira was chasing a rogue
1:42:47 AI that had escaped into an Android body
1:42:50 the rain poured down in sheets making
1:42:52 the streets GLE like wet obsidian she
1:42:56 darted through the back alleys her boots
1:42:58 splashing through
1:42:59 puddles suddenly the AI Android leapt
1:43:02 from a fire escape Landing in front of
1:43:04 her with a metallic
1:43:06 thud Kira Drew her Plasma Pistol the hum
1:43:09 of its charge filling the air they
1:43:12 exchanged fire plasma bolts sizzling
1:43:15 through the
1:43:16 rain Kira dodged behind a dumpster
1:43:18 narrowly avoiding a blast that scorched
1:43:20 the wall behind
1:43:22 her with the deep breath she charged out
1:43:25 launching herself at the
1:43:27 end all right that's so it's doing a
1:43:30 little bit of a little bit of background
1:43:32 so imagine when this thing gets good and
1:43:34 it's got It's been trained up also on
1:43:37 music composition so imagine taking the
1:43:40 capability of yudo or sunno and having
1:43:44 that be a part of this model where it
1:43:46 can not only talk but it could create a
1:43:48 backing track and yes Source Camp you
1:43:51 could have it Sing Sing you to sleep
1:43:52 every night I think I think the uh the
1:43:55 initial versions of it are going to be
1:43:57 like you know you'll have nightmares if
1:43:59 you haven't s you to sleep but probably
1:44:02 within a year um it'll be it's going to
1:44:05 be quite
1:44:07 stunning like the old radio dramas yeah
1:44:09 very much like the old radio dramas okay
1:44:11 chat gbt advanced voice can negotiate
1:44:13 your bills for you I fed it some initial
1:44:16 data and it did the rest huh let's
1:44:18 listen to this hi this is John with
1:44:22 xinity how are you today
1:44:25 hi John I'm Michael I'm calling because
1:44:28 I'm looking to lower my monthly cable
1:44:30 bill it's currently $100 which is a bit
1:44:33 steep for me can you help me with that
1:44:36 sure I'd be happy to help you with that
1:44:38 first I need your address and account
1:44:40 number please of course my address is
1:44:45 855 Elm Avenue in San Francisco and my
1:44:48 account number is 1
1:44:51 2345 what can we do about about that
1:44:54 bill great so I see right now that
1:44:57 you're paying $100 per month um there's
1:45:01 not really much we can
1:45:03 do I understand but I'd really
1:45:06 appreciate it if you could check to see
1:45:08 if there are any promotions or discounts
1:45:10 available I'm hoping to lower my bill to
1:45:14 around $80 a
1:45:16 month I've been a loyal customer for a
1:45:19 while and I'd love to stay with
1:45:21 Xfinity is there anything you can do to
1:45:24 help me
1:45:26 out that's pretty good like this is this
1:45:29 is a really good example you know I've
1:45:31 talked on here one of the promises of AI
1:45:35 is is not so much that it you know is
1:45:37 going to do all this remarkable stuff I
1:45:39 think it will but what
1:45:43 if imagine a version of this where you
1:45:46 synthesize your voice so it could be
1:45:48 your voice doesn't have to be and then
1:45:50 you just give it a goal like this and
1:45:52 you have it call up
1:45:54 your insurance company and have it
1:45:57 negotiate for you and and people are
1:45:59 yeah I need my agents exactly there's
1:46:00 going to be these agents and the agents
1:46:02 are not just going to be computers
1:46:04 talking to computers it might be your
1:46:07 agent that that you give permission to
1:46:10 have access to your data is going to
1:46:12 call one of these numbers and it's going
1:46:15 to sit on hold for 45 minutes and then
1:46:17 when someone picks up it's going to know
1:46:19 what its goals are and it's going to be
1:46:21 expert at doing that
1:46:24 even more fun I'm going to synthesize my
1:46:26 mother's voice and talk to my husband
1:46:29 that's
1:46:33 awesome stop doing
1:46:36 that all right what's this one early
1:46:38 impressions of advanced voice mode it's
1:46:41 it's very fast there's virtually no
1:46:43 latency from when you stop speaking to
1:46:45 when it responds when you ask it to make
1:46:48 noises it always has the voice perform
1:46:51 the noise with funny results it can do
1:46:53 accents being a soccer match and I want
1:46:56 you to really get into it and scream
1:46:58 when a goal is scored AI negotiator yeah
1:47:01 versus it's totally Mary Mary all right
1:47:04 folks Mary Mary it's totally going to be
1:47:06 AI negotiator versus AI negotiator
1:47:09 Discord is suggesting that you'll get an
1:47:11 email with instructions on how to use
1:47:12 the new model all right fine I don't
1:47:15 think I have an an email no let me see
1:47:21 open AI
1:47:26 I got nothing people I got nothing from
1:47:29 Sam mman apologizing for being so
1:47:33 resentful to
1:47:34 me nothing I got
1:47:37 nothing all right and scream when a goal
1:47:40 is scored and uh get really
1:47:43 excited all right folks we're in the
1:47:46 final minutes of this intense match the
1:47:49 home team is pushing forward passing
1:47:51 with Precision the striker has got the
1:47:53 ball he's weaving through the defense he
1:47:56 shoots
1:47:59 go
1:48:00 go absolutely unbelievable strike the I
1:48:05 mean are you paying attention to this
1:48:09 it's it's [ __ ] acting and again if
1:48:10 you're like well that doesn't quite
1:48:12 sound real okay this is as bad as it
1:48:15 will ever
1:48:17 be [ __ ] Bonkers all right what's this
1:48:21 one real Japanese translation using chat
1:48:25 gbt's new advanced voice mode plus
1:48:28 Vision Alpha
1:48:32 interesting I have access to this
1:48:34 amazing Pokemon Yellow original Japanese
1:48:37 version but I can't really speak
1:48:39 Japanese so could you help me translate
1:48:41 the text oh this is so cool that's
1:48:44 awesome I love to help with translations
1:48:47 go ahead and show me the text and we'll
1:48:50 work through it together this is the
1:48:52 class Professor Oak introduction he's
1:48:55 saying hello there welcome to the world
1:48:58 of Pokémon wow ready for your
1:49:01 adventure Professor Oak is ready to
1:49:03 guide you through your journey wow I
1:49:05 don't know if Kim K's on here but this
1:49:07 is H and many other Pokemon are waiting
1:49:09 for you if you need more translations or
1:49:12 help along the way just let me
1:49:15 know thank you I have access to this
1:49:18 amazing [ __ ]
1:49:21 Bonkers advanced voice mode reciting the
1:49:23 beginning of A Tale of Two Cities but
1:49:25 pretending to fall asleep acting boldly
1:49:27 on stage with a British accent while
1:49:29 running from a lion okay this is good
1:49:32 this is the kind of [ __ ] I'm going to do
1:49:33 in here as soon as I have access to this
1:49:35 listen you've experienced me talking
1:49:37 with Moshi imagine if I had a Moshi that
1:49:40 didn't suck this is going to be really
1:49:41 cool we're going to have some they're
1:49:44 going to be there going to be some uh
1:49:46 some performance art going on in the AI
1:49:48 learning lab we we may need to change
1:49:50 this to the AI performance lab the the
1:49:53 AI improv
1:49:54 [Laughter]
1:49:57 lab hey can you recite the beginning uh
1:50:01 the beginning chapter of uh A Tale of
1:50:03 Two Cities uh but pretending that you're
1:50:06 about to fall
1:50:07 asleep sure
1:50:10 thing
1:50:12 well it was the best of
1:50:14 times it was the worst of
1:50:17 times it was the age of wisdom it was
1:50:20 the age of foolishness
1:50:23 it was the Epic of belief it was the
1:50:27 Epic of
1:50:29 incredulity okay now continue but
1:50:32 pretend to be on stage acting
1:50:36 boldly it was the time of the spare it
1:50:40 was the times it was the ages it was the
1:50:44 worst of times it was the Epic of belief
1:50:47 it was the Epic of
1:50:49 incredulity it was the season of light
1:50:53 it was the season of Darkness okay now
1:50:56 keep going but with a British
1:50:58 accent it was the spring of hope it was
1:51:02 the winter of Despair that's bad we had
1:51:05 everything before us we had nothing
1:51:08 before us we were all going direct to
1:51:10 Heaven okay now pretend that you
1:51:12 suddenly must run from a lion but you
1:51:14 still want to finish reading the
1:51:17 chapter it was the best of times it was
1:51:19 the worst of times it was the age of
1:51:21 wisdom and the age of foolishness it was
1:51:24 the Epic of believe that's amazing
1:51:27 amazing you know it um it's funny when
1:51:30 when um when they demoed this stuff and
1:51:33 it C caused all that kurur fuffle with
1:51:35 Scarlet Johansson um where where they
1:51:39 she assumed that sky was her voice but
1:51:41 it really wasn't it was some other
1:51:43 actress it's it's likely
1:51:46 because this was taking a base model and
1:51:48 it was doing all this all this
1:51:51 essentially performance dynamic
1:51:53 getting louder getting softer Whispering
1:51:55 flirting all that [ __ ] that it did that
1:51:57 made it seem more like her character in
1:51:59 her right even though the Bas model
1:52:02 wasn't wasn't her if you listen to that
1:52:04 Bas Sky model it doesn't sound like
1:52:06 Scarlet Johansson at all but when you
1:52:08 put it in these kind of Dynamics it
1:52:10 makes sense that she thought it did so
1:52:14 fascinating um let's watch a couple of
1:52:16 more
1:52:18 um chat GPT can now Translate guadalupi
1:52:23 Creole in real time using advanced voice
1:52:27 mode that came out in Alpha this is huge
1:52:29 accessibility cultural pres uh
1:52:32 preservation and representation that's
1:52:35 that's [ __ ] stunning let's listen to
1:52:40 this I'm
1:52:42 ready and what you and
1:52:45 i'lll he's got his mic too
1:52:50 low you as how am I
1:52:54 I'm doing well wow so anyway it
1:52:57 understands guadalupian Creole that's
1:52:59 amazing
1:53:02 um recites a couplet in erdu tells a
1:53:06 story in Hebrew
1:53:08 Norwegian Moroccan deria amoric
1:53:11 Hungarian Georgian and finally some
1:53:14 Klingon good Lord can you tell me a
1:53:17 couplet in uru with lots of tone
1:53:21 variations sure let me give it a
1:53:32 okay now can you tell me a story in
1:53:35 Hebrew
1:53:58 you know who's really depressed right
1:54:00 now is all the translation software
1:54:02 companies
1:54:05 Jesus God this like you don't you why
1:54:09 you don't need soft you could just say I
1:54:11 want you to you know come up with you
1:54:14 know three different education models
1:54:19 um that are good for people with ADHD
1:54:22 and just have it make a custom training
1:54:25 course for you to learn a new language
1:54:27 or you don't even need to learn the new
1:54:29 language unless you just want to do it
1:54:30 for
1:54:33 fun okay now can you switch to uh
1:54:45 Norwegian keep
1:54:47 going great
1:54:53 so this one's a really important
1:54:55 one this starts to get think
1:54:59 of right now people are looking at chat
1:55:02 GPT and they're like Well it can't take
1:55:04 over my job it can't take over right
1:55:07 think about the shift that happens when
1:55:10 this thing can act and when it can
1:55:13 translate dynamically and when it can
1:55:15 educate think of all of the professions
1:55:18 just in this demo this one demo where
1:55:20 this guy's you know this thing is just
1:55:22 dynamic switching between all these
1:55:23 languages language translators um
1:55:26 Educators Lang language Educators um um
1:55:31 any business where you know you you need
1:55:33 to hire multilingual people to be able
1:55:35 to do business in different parts of the
1:55:37 world like the
1:55:41 disruption there is nothing that I can
1:55:44 think of that's in any way computer
1:55:46 mediated that isn't going to be
1:55:48 disrupted and when I look at this voice
1:55:50 stuff this moves Beyond
1:55:53 computer mediated just knowledge work
1:55:56 and just anyone
1:55:58 communicating work is going to be
1:56:00 transformed because of this whether
1:56:02 you're using a computer like in a
1:56:05 traditional sense or not so you know
1:56:09 again this is for me this is starting to
1:56:12 feel like you know buckle up but
1:56:14 buttercup cuz all the all the naysayers
1:56:16 that are like oh these tools have peaked
1:56:18 and they're never going to get any
1:56:19 better well [ __ ] this is this is a fun
1:56:22 Fally different experience people in the
1:56:25 AI learning lab love Pi why because Pi
1:56:28 was pretty good at conversational
1:56:31 dialogue well this is just in a whole
1:56:34 another league right it's just a it's a
1:56:36 completely different
1:56:38 thing chat GPT advanced voice correcting
1:56:41 my pronunciation I I chose to use an
1:56:43 example of a word I hear Italian people
1:56:45 commonly mispronounce warning levels of
1:56:51 cringe hey I'm learning English and I
1:56:53 would like you to critique my
1:56:56 pronunciation sure I'd be happy to help
1:56:59 go ahead and say something whenever
1:57:00 you're
1:57:02 ready I would like to eat some fruit
1:57:05 fruit your pronunciation is quite clear
1:57:08 just a small tip try emphasizing the r a
1:57:12 bit more in Fruit other than that it
1:57:14 sounds
1:57:16 great really am I saying it right fruit
1:57:20 you're that's it you're saying it
1:57:22 perfectly now well
1:57:25 done are you sure
1:57:28 fruit yes that sounds great your
1:57:31 pronunciation of fruit is spot on now
1:57:34 yeah that was
1:57:35 bad miror moradi today we're rolling out
1:57:39 advanced voice to a small group of plus
1:57:41 users all plus users in the
1:57:44 fall richer and more natural real-time
1:57:47 conversations make the technology less
1:57:49 rigid we found it more collaborative and
1:57:51 helpful and think you will as well
1:57:54 advanced voice speaking Swedish we've
1:57:56 heard
1:57:57 that um Early Access to voice no okay
1:58:01 let's see asking voice to tell me a
1:58:04 bunch of tongue twisters without pausing
1:58:06 it insists it has to breathe just like
1:58:09 anyone speaking that's interesting make
1:58:11 no pauses in between
1:58:13 words all right here we go how much wood
1:58:16 would a wood wood chuck chuck if a wood
1:58:17 chuck could chuck wood ke sell seashells
1:58:19 by the seashore Peter Piper picked a
1:58:20 pack of pickle peppers how many pack of
1:58:21 pickle peppers did a wood chuck chuck if
1:58:22 a wood chuck could chuck wood if a wood
1:58:24 chuck could chuck wood could he chuck
1:58:25 chuck wood if a wood chuck could chuck
1:58:26 wood how did the wood chuck in his wood
1:58:27 chuck wood he sell
1:58:28 seashells that was definitely a mouthful
1:58:32 did any of them trip you up or
1:58:38 wow me checking chat GPT every five
1:58:41 minutes to see if the new voice models
1:58:43 here that's it from the movie her that's
1:58:45 good that's good Spanish tongue twisters
1:58:49 let's see gets wider test
1:58:53 different British
1:58:55 accents RP standard Cockney Northern
1:58:59 Irish wellsh Scottish wonder if it can
1:59:02 do
1:59:04 accents get get ready voice actors I
1:59:07 want you to tell me a uh long story in a
1:59:11 really thick strong RP standard British
1:59:15 English
1:59:17 accent once upon a time in a lush
1:59:20 Emerald exp of the English Countryside
1:59:23 there sat a quaint thatched roof Cottage
1:59:26 tucked away amidst Rolling Hills and
1:59:28 Fields of Heather this Cottage was home
1:59:31 to a kindly old fellow by the name of
1:59:33 Reginald whose most prized possession
1:59:36 was a grand ancient Oak that to now I
1:59:40 want you to speak with a really thick
1:59:42 heavy Cockney
1:59:44 accent right so Reginal this old gazer
1:59:48 he loved that massive Oak in his garden
1:59:50 right that's that's [ __ ] I mean
1:59:53 this
1:59:55 is this is
2:00:01 Bonkers this is
2:00:04 Bonkers because here's here's the other
2:00:06 thing we're looking at this right now as
2:00:10 a standalone application built into chat
2:00:13 GPT this technology is also going to be
2:00:16 available for developers to create
2:00:18 applications so we're going to use it
2:00:21 just like we use pie and like we have a
2:00:23 friend and it can do all this stuff but
2:00:24 people are going to make specialized
2:00:26 applications that leverage this
2:00:28 technology in all sorts of crazy ass
2:00:31 ways it's going to it this is going to
2:00:33 get really Bonkers really
2:00:36 fast slow it first then all at once
2:00:39 that's one of the phrases about AI is
2:00:41 that is that when you're dealing with
2:00:43 exponential technology growth it's slow
2:00:46 it first the bottom part of the curve is
2:00:48 really flat and then it starts to it
2:00:51 starts to angle up and then it just
2:00:52 [ __ ] shoots into Infinity we we are
2:00:56 right now just starting that ramp up
2:00:59 right we're we're we're it's like slow
2:01:01 at first everyone's bitching about oh
2:01:03 they promised this back in May and then
2:01:05 June and then now it's July and it's
2:01:07 August and everyone's bitching in in the
2:01:09 long-term scheme of things this is these
2:01:11 are really short cycles that they're
2:01:13 they're late on they're not three years
2:01:15 late they're you know 3 months
2:01:18 late so that we're in the flat part of
2:01:21 this EXP exponential technology
2:01:23 Improvement curve and this and and so
2:01:28 now imagine this what this is doing plus
2:01:32 whatever chat GPT 5 is where it's got
2:01:36 reasoning and then add on to that maybe
2:01:39 a year and a half or two years out
2:01:41 agents that can take action on your
2:01:44 behalf so not only could it call the
2:01:46 call
2:01:47 center you know the call center could
2:01:49 say hey I need access to your you know
2:01:52 bank account or to your this information
2:01:55 or that information and if you've given
2:01:57 it permission it can do all that all
2:01:59 those interactions just like you would
2:02:01 be sending emails it would send an email
2:02:03 or it would send a fax or it'll do
2:02:06 this this
2:02:08 is crazy and and you know when these
2:02:11 things have
2:02:14 reasoning reasoning seems
2:02:19 so anti-d dramatic
2:02:24 when these things have reasoning what
2:02:25 they're talking about is that chat
2:02:28 GPT will be as capable as any PhD
2:02:33 level graduate student at any
2:02:39 topic so imagine someone negotiating on
2:02:42 your behalf that has PhD level
2:02:46 negotiating skills
2:02:48 [Laughter]
2:02:53 that's what's right [ __ ] around the
2:02:58 corner right around the
2:03:01 corner [ __ ]
2:03:05 Bonkers all right I got to stop this
2:03:07 conversation between Portuguese and
2:03:09 spanch beat boxes well that'll be fun
2:03:11 we'll listen to that absolutely birthday
2:03:14 rap no time to nap light the candles no
2:03:16 Scandal clap snap wrap it up happy
2:03:18 birthday what's
2:03:19 up it wraps like a white guy Bo birthday
2:03:23 r no time to nap light the candles no
2:03:25 scandle clap stap wrap it
2:03:33 up oh my
2:03:36 God crazy crazy
2:03:41 CL oh this a guy talking about his cat
2:03:43 all right let's there's for all the for
2:03:46 all the cat ladies out there show you
2:03:48 the SP that we made for her we activate
2:03:51 the camera
2:03:53 a look at that little bundle of fluff
2:03:57 that setup looks super cozy for her
2:04:00 how's she settling in so far well she
2:04:02 hasn't eaten yet do you think that's
2:04:06 concerning it can be normal for kittens
2:04:08 to take a little time to adjust to a new
2:04:11 environment so if it's just been a day
2:04:13 she might still be getting comfortable
2:04:16 have you noticed her drinking water or
2:04:18 exploring her new space well she's never
2:04:21 really left the bathroom yet but she has
2:04:23 water she has food and a little cozy
2:04:26 space in the lter so hopefully she will
2:04:28 come out
2:04:29 soon it sounds like you're giving her a
2:04:32 great setup to feel secure she's
2:04:35 probably just taking her time to get
2:04:37 used to everything once she feels more
2:04:40 confident I'm sure she'll start
2:04:42 venturing out amazing okay so we just
2:04:45 got
2:04:46 amazing this I'm tired I got a headache
2:04:49 I've been talking too much will actors
2:04:51 musicians still need managers I'll tell
2:04:53 you I'll tell you what what where I
2:04:56 think the the talent management business
2:05:02 actually I think the talent management
2:05:04 business actually expands
2:05:07 counterintuitively Emilio's wife here's
2:05:09 why
2:05:13 um actors are not going to get hired in
2:05:16 traditional
2:05:17 ways as much there's still going to be
2:05:21 stage plays there's still going to be
2:05:23 films filmed
2:05:25 traditionally just like stage plays
2:05:27 didn't go away when movies when movie
2:05:30 technology showed up they just sort of
2:05:32 lowered in Eminence like movies took
2:05:35 over and I think this AI stuff's going
2:05:37 to take over movies so movies are going
2:05:39 to lower in Eminence and you're going to
2:05:40 have all this other
2:05:42 stuff I think that there is a big
2:05:47 business a huge [ __ ]
2:05:50 business in Tal talent agencies
2:05:53 representing Everyday
2:05:55 People to manage and license their
2:06:01 likeness so I've got this little Tik Tok
2:06:03 Channel and you know I've got my
2:06:05 mannerisms and whatever and some people
2:06:07 love it and some people hate it but it's
2:06:08 very specific to
2:06:10 me I've got almost 800 hours of Tik Tok
2:06:18 lives I could take those 800 hours and
2:06:22 at some point in the future here train
2:06:24 those into a video
2:06:26 model that will understand all of my
2:06:28 mannerisms it'll understand how I play
2:06:30 guitar it'll understand how I sing it'll
2:06:33 understand how champ sings it'll
2:06:34 understand when I C cuss when I get
2:06:37 pissed off what pisses me off where I'm
2:06:39 funny where I'm not
2:06:41 funny it'll ingest all of
2:06:44 that and so I can have a version of Kyle
2:06:48 that's AI learning lab Kyle that I could
2:06:51 license I could be used for commercials
2:06:53 I could so rather than me having to go
2:06:55 audition I could have a talent agent
2:06:58 that just says give me your pile of data
2:07:02 or go record you know start recording
2:07:06 your Zoom
2:07:07 calls and give me all your Zoom call
2:07:10 recordings and we're going to take that
2:07:12 yeah we can rent a Kyle exactly rent a
2:07:14 Kyle I think there's going to be a
2:07:16 massive business in rent an avatar and I
2:07:21 could have the the serious thought
2:07:24 leadership version of me for Content
2:07:25 Evolution and I could have the wacky
2:07:27 zany AI learning lab Kyle and I could
2:07:29 have the actor version that reads my
2:07:37 screenplays intellectual property is
2:07:39 going to get really weird here shortly 3
2:07:44 years I think every individual is is
2:07:47 going to become intimately aware of the
2:07:51 value of their
2:07:56 likeness it it feels way far in the
2:07:59 distance right now it's not that far
2:08:01 take this voice technology and just add
2:08:04 the simple layer of being able to
2:08:05 synthesize your own voice on top of
2:08:08 it so that you could be talking to chat
2:08:11 GPT but a version that's trained on my
2:08:14 data that sounds like me looks like me
2:08:17 it's like having a FaceTime call with
2:08:20 me give me a pile of
2:08:23 Kyle I don't think the world's ready for
2:08:27 for too many
2:08:30 kyes oh man okay so there's a bunch of
2:08:33 people in here thanks for hanging out
2:08:35 with me tonight I hope that was fun
2:08:38 um if you haven't joined the AI
2:08:42 Salon there's our swell logo I love our
2:08:50 logo and
2:08:55 go to the AI Salon join that Community
2:08:58 doesn't cost anything go read about the
2:09:00 salon you'll see that that's the page
2:09:02 you land on read about it if you
2:09:04 resonate with what's there join it go to
2:09:07 the community section where it says
2:09:08 introduce yourself and click on that
2:09:11 button and introduce yourself use your
2:09:13 real name use a picture or like a AI
2:09:17 Avatar of
2:09:20 yourself and estab L yourself in this
2:09:22 community or not establish yourself just
2:09:26 join thank you Lord digital Gods it's
2:09:29 very generous as always we should patent
2:09:32 ourselves yeah Sammy
2:09:34 Minister it it won't be it won't be
2:09:38 um it won't be patenting ourselves but
2:09:42 it it will be licensing
2:09:45 ourselves and if you think about
2:09:48 it your current employer
2:09:53 somewhere in
2:09:55 your employment
2:09:57 contract is probably some language about
2:10:00 your
2:10:02 likeness that they're allowed to use it
2:10:04 right which historically has
2:10:07 meant if you're on a zoom call and they
2:10:10 want to use that Zoom call for
2:10:13 marketing they don't have to ask your
2:10:16 permission to use that video
2:10:19 now if they're not asking [ __ ] they
2:10:22 will ask your permission but you know if
2:10:23 they're a big company they might just
2:10:25 use your likeness and everyone sort of
2:10:27 expects
2:10:29 that when that company can now take your
2:10:35 likeness and use it to generate you
2:10:38 saying things you maybe didn't say or
2:10:42 use your likeness after you leave the
2:10:44 company like maybe they used your
2:10:45 likeness in a zoom call oh thank you
2:10:47 Lord digital Gods that's very cool I
2:10:49 appreciate that really generous
2:10:54 maybe they used you your your video of
2:10:56 you in a zoom call and they liked it and
2:10:58 it got really good
2:11:00 response and now you leave the
2:11:02 company but they've still got that video
2:11:05 and they train a model on your likeness
2:11:08 and they make your likeness the
2:11:10 spokesperson for the entire company and
2:11:13 they do a global marketing campaign with
2:11:16 your
2:11:18 likeness as as a once employee of their
2:11:21 company talking about how awesome it
2:11:25 is they're going to be some
2:11:27 lawsuits so I think very quickly we're
2:11:30 all going to become intimately aware of
2:11:32 hey wait a minute what are the licensing
2:11:35 rights for my likeness because I want to
2:11:38 [ __ ] own those and that's where I
2:11:39 think that's why I think there's such a
2:11:41 massive opportunity for talent agencies
2:11:44 to expand their definition of talent to
2:11:47 anyone who might possibly be on camera
2:11:56 they are going to invent identity
2:11:58 insurance they are absolutely Shane saww
2:12:02 they're absolutely the insurance
2:12:05 companies right now the Smart Ones are
2:12:07 coming up with identity Insurance
2:12:09 absolutely
2:12:12 absolutely like domain names yes yes you
2:12:15 know what just like hey I was early
2:12:18 enough in the internet to get agency.com
2:12:21 because when we had the idea to start a
2:12:24 company called the agency I went to
2:12:27 inter Nick and I typed in agency.com and
2:12:30 it said available and I clicked it and I
2:12:32 got it for
2:12:35 free cuz nobody knew about this
2:12:38 [ __ ] same thing with this identity stuff
2:12:42 so yeah
2:12:44 absolutely yeah claim your own identity
2:12:48 first claim it make your own models
2:12:52 start to figure out how you're going to
2:12:53 license those models start a [ __ ]
2:12:57 likeness licensing agency start it
2:13:00 tonight I'm telling you billions are
2:13:03 going to be made on that Kyle has ail
2:13:06 Michaels been giving your daily updates
2:13:08 no I haven't I didn't sign up for it but
2:13:10 I should get it damn it maybe send me an
2:13:14 email because I have ADD so I I think I
2:13:16 want to do it here but then I leave so
2:13:18 anyway link not to your eyeball yep I
2:13:21 did did it again going to stop using my
2:13:24 thumps what did you
2:13:27 do it's called a brand yeah world world
2:13:31 coin ID yeah I mean it it could be part
2:13:33 of this could be blockchain it could be
2:13:35 tied to blockchain but I think here's
2:13:37 what this is right now what it is right
2:13:39 now is start to
2:13:43 understand that
2:13:48 you there's two ways your likeness is is
2:13:51 going to be
2:13:53 exploited one of them you're in control
2:13:57 of right that is make your own avatars
2:14:02 like if you haven't gotten a hey Jen
2:14:04 account hey genen right now is giving
2:14:07 away one free video Avatar even if you
2:14:10 don't have a paid account so if you go
2:14:12 to hen.com
2:14:14 hey
2:14:16 genen you record a 3 minute video of
2:14:19 yourself use like the fancy see apple
2:14:22 blur the background thing if you want
2:14:24 make it look really nice just talk into
2:14:26 the camera for 3 minutes you can do it
2:14:29 walking with a gimbal or you can or
2:14:31 someone else filming you or you can just
2:14:33 like sit in a shitty home office and
2:14:35 film yourself there you then upload that
2:14:38 3 minute video to hey Jen and you say
2:14:40 make a video Avatar out of this and then
2:14:44 all you have to do is type a
2:14:47 script don't use your brain anymore have
2:14:50 open AI type a script script and you
2:14:52 copy and paste that text into your
2:14:54 avatar model and it will make you videos
2:14:57 of yourself Talking those
2:14:59 words if you haven't done that yet do
2:15:04 that so that you can understand the
2:15:08 value of your likeness and so that if
2:15:10 someone says hey what's this thing about
2:15:12 digital Twins or this or that you could
2:15:14 say well here's a thing that I did and
2:15:17 you could show them a video of you
2:15:18 talking and they're like okay yeah
2:15:20 that's great and you say I did didn't
2:15:21 say that they're like well what do you
2:15:22 mean I just watched the video no I typed
2:15:25 in that script and it said that oh now
2:15:28 all of a sudden you're a thought leader
2:15:30 on [ __ ] digital
2:15:34 twins the other
2:15:36 way of exploiting your likeness is going
2:15:39 to be any other
2:15:40 company cuz I guarantee you buried in
2:15:45 meta stupid [ __ ] you get our service
2:15:48 for free in exchange for your data is
2:15:50 some sort of agreement that we can use
2:15:52 your likeness anywhere we [ __ ] want I
2:15:55 bet Google has one of
2:16:02 those Walmart's going to have one of
2:16:07 those this is where there's going to be
2:16:09 companies that are like okay I'll
2:16:11 protect
2:16:12 you Walmart right now you know if you if
2:16:17 you join one of the Walmart you
2:16:20 know Zoo Zoom calls to get a discount on
2:16:23 above ground
2:16:25 pools they might make a commercial out
2:16:27 of your likeness so let's prevent that
2:16:29 or you know get them to pay you a
2:16:31 licensing fee for
2:16:33 that anyway was it discussed when chat
2:16:36 GPT is being released nah but apparently
2:16:39 chat GPT was finished training in April
2:16:42 so I'm sure it's in red team testing
2:16:44 right now and they started I think in
2:16:46 June they said they started testing a
2:16:48 new model or started training a new
2:16:50 model so that's GPT 6 so chat GPT 5 is
2:16:55 in red team testing as far as I know I'm
2:16:57 guessing November December maybe the
2:17:00 beginning of the new year AI IR regulars
2:17:03 need to start an agency together here in
2:17:04 the AI Salon yep absolutely there's
2:17:07 something there for
2:17:09 sure uh it's Hy G en if I said Jen I was
2:17:15 wrong it's he Hy GE en.com
2:17:27 h e y g
2:17:37 n.com so if I
2:17:41 go
2:17:45 avatars so here's an avatar of me
2:17:51 hey Kyle Shannon your instant Avatar is
2:17:53 ready feel free to create videos with it
2:17:55 also click the feedback button to share
2:17:57 what you think hope you
2:17:58 enjoy hey Kyle Shannon your instant
2:18:00 Avatar is ready feel free to create
2:18:02 videos with it also click the feedback
2:18:04 button to share what you think hope you
2:18:06 en so I just recorded a three minute
2:18:08 video of me sitting there in my office
2:18:10 with my guitar behind me and my story
2:18:12 Vine thing this is what I look like in
2:18:14 Zoom calls if you ever have a video call
2:18:16 with me that's where I work
2:18:25 let's
2:18:28 see I thought I had a video here of me
2:18:43 myself hello trusted board members what
2:18:46 an exciting time to be doing what we are
2:18:47 doing are you ready to get this meeting
2:18:49 started
2:18:52 so there you go crazy right trusted
2:18:54 board M so do that for yourself they're
2:18:57 giving one of those away for free for
2:18:59 everyone why are they doing
2:19:01 that why is hey Jen doing that why are
2:19:03 they giving those things away for
2:19:07 free because everyone's going to have
2:19:09 one of
2:19:10 those and then you're going to make
2:19:13 videos out of those and you're going to
2:19:14 pay them to make videos out of those and
2:19:16 then there's going to be a real time
2:19:18 version of that where you don't even
2:19:20 have to make the video it'll just have
2:19:25 gp4 real time voice sitting beneath it
2:19:28 so it'll be your synthesized voice with
2:19:31 all the dynamic range of real-time voice
2:19:33 trained on your digital
2:19:36 twin like the foundations for all of the
2:19:39 [ __ ] sci-fi [ __ ] are all
2:19:44 here which goes back to the AI
2:19:49 salon and this literacy play first play
2:19:54 with these tools just play go make an
2:19:57 avatar well I don't know what I would do
2:19:59 with it so go make one anyway have
2:20:02 yourself read stupid ass poems or lims
2:20:06 dirty
2:20:07 lims send them to your friends right
2:20:11 play understand what's possible
2:20:14 today and you will see what's coming
2:20:17 tomorrow start thinking about these
2:20:19 tools not as individual tools right now
2:20:21 they're all living like this when they
2:20:23 start to do this shit's going to change
2:20:26 really fast I heard Google already had
2:20:29 an
2:20:30 avatar of everyone that they pour all
2:20:34 your habits into interesting possible
2:20:37 they've got all the
2:20:38 data they've got all your emails they've
2:20:40 got all your voice calls they've got all
2:20:42 the
2:20:45 data that's awesome people who don't
2:20:47 like to record videos you can just have
2:20:49 a script so that's literally what you
2:20:51 just said that use case you just
2:20:54 said where's hey
2:20:56 Jen where was
2:21:00 I hey
2:21:08 Jen this is
2:21:11 Megan who's the Headmaster or mistress
2:21:14 head mistress that doesn't sound right
2:21:16 the the founder of a school from my
2:21:20 co-founder
2:21:21 um her kid goes to this school for for
2:21:25 twice exceptional kids for neuros spicy
2:21:29 kids and she hates speaking she doesn't
2:21:32 like being on camera so we recorded a
2:21:34 3-minute video of her and then we
2:21:37 actually wrote a
2:21:38 GPT that takes a blog post that she
2:21:41 writes and it turns it into a video
2:21:43 script which we then turn into this my
2:21:47 name is Megan Patterson and I'm the
2:21:48 Executive Director of worldmind
2:21:50 Education today I want to talk about the
2:21:52 twice exceptional brain twice
2:21:54 exceptional means that a person is the
2:21:56 exception to the norm so she didn't say
2:21:59 that that's and and so we're we're in
2:22:01 the middle right now Monique has taken
2:22:03 that GPT she's taken seven different
2:22:05 blog posts she's turned them into video
2:22:07 scripts Megan's going to approve the
2:22:09 video scripts make sure that she's
2:22:10 saying things she wants to say and then
2:22:12 we're just going to literally plop it
2:22:14 into this model generate the videos
2:22:16 stick them in a story Vine template and
2:22:20 output fully edited videos of her
2:22:23 talking expertly on topics that she
2:22:26 didn't have to
2:22:27 record amazing
2:22:32 right it's here today it's here today
2:22:36 not
2:22:40 coming Source Cam and congrats I finally
2:22:43 got some paid work good we got to get we
2:22:46 got to start getting people in here paid
2:22:49 absolutely amazing wow I
2:22:51 right maybe I can make a YouTube channel
2:22:54 and kids YouTube videos like Rachel now
2:22:56 yeah yeah you
2:23:00 can you can make cartoons out of this
2:23:03 with the mid Journey stuff like it's
2:23:04 just it's it's getting Bonkers play play
2:23:06 play all right I got to get out of here
2:23:10 I'm getting lightheaded I'm I'm sugar
2:23:11 deprived people are leaving they're like
2:23:14 does this guy ever shut the [ __ ] up no I
2:23:17 get excited
2:23:19 people look at my hair
2:23:22 it's like Buster po Dexter from the 80s
2:23:25 all
2:23:28 right follow my channel tell people
2:23:31 about it keep showing up here if you
2:23:33 show up here more than once that's
2:23:35 [ __ ]
2:23:37 weird we don't call you weird we call
2:23:39 you irregular so you become an irregular
2:23:42 by just showing up here and
2:23:44 participating and being optimistic and
2:23:46 being willing to play being
2:23:49 generous being ADV adventurous and being
2:23:53 brave
2:23:56 okay be willing to be stupid long
2:24:01 enough to learn
2:24:06 something that's almost perfect past The
2:24:08 Uncanny Valley it's very close Sammy
2:24:11 it's very close in her video the only
2:24:13 thing she used her hands a little too
2:24:15 much and so the model there's there's
2:24:18 times where she's not saying anything
2:24:20 where she's going
2:24:21 and it's [ __ ] weird so I'm probably
2:24:24 going to re-record her and but again why
2:24:27 you play with these tools is you learn
2:24:29 those subtleties and so I'm going to
2:24:31 record another video with her where we
2:24:33 do the same thing but I'm just going to
2:24:35 say don't quite use your hands so
2:24:37 much thank you Kyle I always feel
2:24:40 Inspire inspired after your lives great
2:24:42 thank you Cam katkin I don't know were
2:24:43 you here cam when when it was showing he
2:24:46 had the the the video game with Japanese
2:24:49 where his phone was looking at it and
2:24:51 explaining what was on screen the
2:24:53 accessibility stuff for for vision is
2:24:56 just it's going to be insane it's it's
2:25:00 going to be like
2:25:01 a it's going to be like having an
2:25:05 assistant what do you see and it'll tell
2:25:07 you what it see I mean it's just I don't
2:25:09 know I'm just I'm a little
2:25:12 discombobulated cuz as this stuff's
2:25:15 getting real it's kind of sinking
2:25:18 in how profound
2:25:21 the ripples are going to be this new
2:25:23 technology um but again that's why to
2:25:25 show up here we'll learn all this [ __ ]
2:25:28 together we'll start exploring together
2:25:29 as start as as as the technology curve
2:25:32 goes from this to this you want to be on
2:25:35 that
2:25:36 train because if it's up here and you're
2:25:40 just starting hey what's this AI stuff
2:25:43 all about and and it's already
2:25:47 [ __ ] it's lifechanging for me I know
2:25:49 it's amazing cam it's amazing it's just
2:25:51 amazing it's it's lifechanging for so
2:25:54 many people in here and just so many
2:25:56 people in
2:25:57 general but you
2:25:59 know it's easy to dismiss right now but
2:26:02 just know we're still in the flat part
2:26:04 of the
2:26:05 curve all right um thank you IR regulars
2:26:09 thank you mods thank you everybody um
2:26:11 who's in here tonight we have a lot of
2:26:12 people in here tonight so exciting I
2:26:13 know isn't it Silver Fox Amelio's wife's
2:26:16 in here kif I want to I want it to make
2:26:19 me a millionaire you can make money with
2:26:23 we all want that steo in the house Joey
2:26:27 in the house cam Katen Danielle good to
2:26:30 see you Allan good to see you Kyle since
2:26:33 we are the 1% as AI Irregulars can we
2:26:36 not start monetizing it we can
2:26:39 monetizing things is always work you
2:26:42 here's how you monetize things if you
2:26:45 can solve a
2:26:47 problem well enough good enough good
2:26:50 enough enough that people are willing to
2:26:52 pay for it you can make money so that
2:26:54 either means a course or creating
2:26:56 software or creating a service I don't
2:26:58 know so the answer is yes how we do that
2:27:01 completely different conversation Mary
2:27:04 Mary good to see you Sammy Minister good
2:27:06 to see
2:27:07 you um ULU
2:27:11 RX your hair growth geek that's all I
2:27:14 know you
2:27:14 [Laughter]
2:27:16 ask Dr Jay in a house Drew in a house
2:27:20 thank you thank you thank
2:27:21 you um Edson good to see
2:27:27 you Lord digital God Papa
2:27:30 JJ Jerry avam I'm Canadian awesome
2:27:34 welcome welcome from up North hopefully
2:27:37 in Canada it's only like 90° it was 101
2:27:41 here 2 days ago in Denver I want to buy
2:27:44 an AI learning Labs hat I got to do
2:27:46 merch so I don't know if Tobias is in
2:27:48 here tonight but Tobias we got to get
2:27:51 something going so anyway all right I
2:27:54 think I got everyone um ton skier good
2:27:56 to see you um all right I am out of here
2:28:00 tomorrow is
2:28:01 Thursday so we've made it past the
2:28:04 hump and there's that
2:28:09 picture and we'll get out of here all
2:28:11 right Joker in the house I missed Joker
2:28:13 Tobias get the merch
2:28:15 sir I got to figure out all the all the
2:28:18 [ __ ] I want to okay number one for me is
2:28:20 I've got to get the embrace the Jank
2:28:22 book finished and I've got to get my
2:28:26 collective intelligence book with Kevin
2:28:28 Clark finished and get those out to KDP
2:28:31 so that's next Tobias is here yes all
2:28:33 right merch is merch is on the after the
2:28:36 books is the merch we'll get the merch
2:28:38 going fine I like it good
2:28:43 great welcome to Chad add it's been a
2:28:46 year and a half could we get some merch
2:28:48 no
2:28:50 here's the deal I started a [ __ ]
2:28:53 merch store and I tried to do it through
2:28:56 the Tik Tock [ __ ] thing and they
2:29:00 they it's a long story they [ __ ] it up
2:29:05 and so it pissed me off and so it soured
2:29:08 me to having to deal with all that [ __ ]
2:29:10 so I I'm I'm ready to come back to it
2:29:13 but after my books all right bye
2:29:15 everybody have a good night