Sep 3, 2024 - Building My AI Twin & Fighting with Chat GPT (Live!)

Description

Chapters

Transcript

0:00 we're doing it
0:03 [Music]
0:20 now sun came up it was the other day
0:25 some down you Blown Away why' you let go
0:29 of your guitar
0:31 why' you ever let it go that far drunken
0:36 [Music]
0:40 angel
0:44 that remember every
0:47 threat fingers touching each shiny
0:50 string you let go of everything drunken
0:56 Angel drunken angel
0:59 [Music]
1:02 on the other
1:04 side drunken
1:07 angel you're on the other
1:12 [Music]
1:25 side good evening Good Fellows good
1:29 lasses good good
1:31 ladies Nells misant all of
1:35 you good day to you sir good
1:38 day sir a quick question might I
1:42 inquire sir sir carrying on good day sir
1:47 quick
1:48 question uh by
1:52 chance sir if I might inquire uh without
1:56 being aggressive in any
1:59 way uh so I was wondering by chance what
2:02 uh were your
2:04 [Music]
2:07 qualifications
2:08 well no one's really asked me that
2:11 before but my
2:12 [Music]
2:12 [Laughter]
2:15 qualifications it's rather clever
2:17 question isn't
2:17 [Music]
2:20 it by asking it you're implying that I
2:22 have none that's fantastic fantastic Bob
2:26 tell him what he's
2:28 won oh we haven't won anything sorry
2:30 it's not a game
2:33 show sorry go on uh oh my qualifications
2:37 yes yes yes ah qualifications uh for
2:41 what
2:42 exactly for hosting a channel called the
2:45 AI learning lab oh well rather direct
2:49 rather direct
2:52 fantastic I suppose next you'll ask me
2:54 what part of England I'm from
2:58 [Music]
3:05 w
3:07 [Music]
3:50 itchy
3:54 nose scratchy
3:57 face could be drugs could be allergies
4:00 you never
4:03 [Music]
4:23 [Applause]
4:28 know good evening good people of the Tik
4:31 Tok I started late but I I was going to
4:33 start at 9:00 but I'm a little earlier
4:35 than 9 I'm later than normal so if
4:38 you've been here before you're like oh
4:39 Kyle's doing that time thing again if
4:42 you haven't been here before you're like
4:44 Alec Baldwin really let himself go
4:48 [Music]
5:17 sting in this lonely
5:20 Town Wonder know when things are going
5:22 to
5:26 change BR my life away
5:30 seems these dreams have turned to Bunch
5:33 D
5:36 glass get my nerve
5:38 up but my past is pulling me
5:43 [Music]
5:45 down wondering how
5:48 long this black sheep don't stick
5:51 [Music]
5:57 around somebody told me once before he
6:00 said you can never go home
6:03 again won't you
6:06 leave said a just to Ste me away yeah
6:11 from the truth of who I am and what I
6:15 believe so I thanked him for his two
6:18 cents and a
6:19 handshake and some
6:21 sympathy
6:23 yeah packed up my blue
6:27 jeans headed for this big
6:31 priz of my
6:35 freedom
6:38 byebye I can hear your hair follicles
6:40 regenerating oops that's champy
6:44 singing Black
6:47 Sheep to the black sheep of the
6:52 family
6:54 [Music]
6:57 byebye so very much to me yeah
7:03 byebye black she to my friends and my
7:09 family
7:11 [Music]
7:15 byebye my
7:18 soul set it
7:24 free free
7:27 [Music]
7:49 let us be L us will Mar fortunes
7:56 together I've got some real estate here
7:59 in my
8:02 he so we bought a pack of
8:05 cigarettes and Mrs wag
8:09 P we all to look for
8:16 [Music]
8:22 a you can make money with chat jpt I
8:25 don't know if you know that you can make
8:27 money with chat jpt but
8:30 you need a gold button these are not for
8:34 sale I had this one Hand Forged by
8:38 Parker from Gold Rush this is right out
8:41 of the the the uh South
8:44 cut this
8:46 gold you're like Kyle if that was gold
8:49 it would be a lot heavier than that
8:50 you're just tossing it around like it's
8:52 a piece of
8:53 plastic you don't understand how strong
8:55 I
8:56 am ripped I'm ripped it's like if I'm
9:00 not learning about AI I'm in the gym
9:02 Pumping
9:03 [Music]
9:06 Iron you're like but Kyle what what
9:09 about that double triple some number of
9:14 chins that's muscle I'm
9:18 flexing like a
9:20 frog like a frog ready to
9:24 [Music]
9:26 mate strength of a bull the chin of a
9:31 frog the
9:34 hair of a 1970 Sears carpet salesman oh
9:38 [Laughter]
9:42 yeah can I interest you in an
9:45 [Laughter]
9:57 heirloom the people who just joined her
9:59 like
10:00 what have I done with my life why am I
10:02 here why am I still watching what is
10:04 this channel who is
10:07 he did he get fired from Saturday Night
10:09 Live why does he do bad
10:14 voices who's the writer for this sketch
10:16 comedy
10:17 [Music]
10:23 show I'm hot hang on I got to adjust
10:27 something
10:29 h
10:31 [Music]
11:03 is he Ed
11:04 Sheeran he's just like so you should
11:07 watch the AI learning lab he he knows a
11:09 little bit about AI not really qualified
11:10 that's not the point he's a lot like Ed
11:12 Sheeran if Ed Sheeran were a little
11:15 tubbier uh old um not redheaded and uh
11:22 how do I put this politely uh without
11:23 the talent if you know what I mean it's
11:26 but it's just like that it's just he's
11:27 just like Ed Sheeran like if you've ever
11:29 seen Ed in concert that's what he's like
11:31 with the guitar but but with no audience
11:33 with with well there's us but like no
11:35 one he can hear he just he thinks that
11:37 we're clapping it's so funny it's so
11:41 funny it's
11:43 like I I don't know why I watch
11:47 it no no uh have I learned anything from
11:50 the learning lab um well that's a so
11:53 that's a fascinating question
11:58 um do I what
12:00 am I an expert in AI now no no how long
12:03 have I've been watching it well for a
12:05 year and a half I've I've watched it
12:07 every night for a year and a half but um
12:10 I I wouldn't say I learned things but
12:12 but it's I I canceled my Netflix
12:17 subscription all right this is this is
12:21 what is this this is my favorite part of
12:23 every weekday thank you Rachel
12:27 Sparkles the sheer idiot see of it
12:34 all oh there's 80 people in here
12:36 wondering what in the hell have I gotten
12:38 myself into here's the deal you can't
12:40 lose you I mean leave well you can lose
12:43 if you stay you can't leave because with
12:47 this channel you it's like a box of
12:49 chocolates you never know what you're
12:50 going to
12:51 get sometimes it's brilliant sometimes
12:54 it's [ __ ] Madness most times it's
12:56 [ __ ]
12:57 Madness occasionally a dog
13:00 [Music]
13:06 sing you know what I think we might do
13:12 today I think we
13:15 might I think I might guide you all to
13:18 make a digital twin if you've not made a
13:20 digital twin of
13:22 [Music]
13:25 yourself I've got thoughts I've got
13:28 ideas
13:29 [Music]
13:32 and I've got a couple of digital
13:34 [Music]
13:36 Twins and I actually believe I actually
13:39 believe this
13:40 [Music]
13:42 cuz I actually believe
13:44 [Music]
13:47 that us understanding how to manage our
13:51 data to create a digital
13:53 twin to create multiple digital
13:57 twins is going to be
14:00 a required skill in the in the next five
14:03 years and I think the people that get
14:06 their heads around it
14:08 first are going to be in good
14:11 shape you're like but how what's a
14:13 digital
14:15 twin hey Cy Cy Cy C hey KY Cy Cy what
14:21 what's digital
14:24 twin so digital twins they come out of
14:27 the world of the simulation World they
14:29 and they probably come out of other
14:30 worlds I I I think you could have it in
14:32 a bunch of
14:35 places but a digital twin is not just
14:38 something that looks like me and sounds
14:39 like me it can have those components but
14:43 the thing I I think that is more
14:45 interesting is what data do you populate
14:48 it with hey Kyle KY Kyle KY hey Kyle
14:53 what that's a digital twin
15:00 and I was thinking how could we do a
15:01 digital twin project on here that's not
15:03 boring as
15:05 [ __ ] and I think I think I have it I
15:07 think I have
15:08 [Music]
15:15 [Music]
15:26 it 10,000 words swarm around my head 10
15:30 million more in books written beneath my
15:34 [Music]
15:37 bed and I wrote a red them all when
15:41 searching in the swamps still can't find
15:44 how to hold my
15:49 hands and I know you need me in the next
15:53 room over I'm stuck in here all
15:57 paralyzed
15:59 [Music]
16:02 four months I got myself in RS too much
16:05 time spending mirrors framed in yellow
16:10 [Music]
16:13 walls ain't
16:16 it like most people I'm no different
16:20 like to talk on things we don't know
16:23 [Music]
16:26 about ain't it like most people
16:29 I'm no
16:31 different like to talk on things we
16:33 don't know
16:35 [Applause]
16:35 [Music]
16:40 about can I join please I need followers
16:44 I can mute myself and gift you I don't
16:47 know what that
16:50 means you think you're dealing with a
16:52 Tik Tock professional here hey Marge
16:55 Marge someone says they wants to join us
16:57 CH the channel
17:00 hey hey Marge you watching the wheel you
17:02 watching the wheel yeah hey someone says
17:05 they wants to join join and then they
17:07 follow on on the Tik Tok on the Tik
17:11 Tok What do I do I I don't I I don't
17:14 know how to I don't know how to use Tik
17:16 Tok
17:18 hey hey is this thing on is it you put
17:22 it on is lot there's
17:25 PE it says 69 69 get it
17:30 it's 69 people it's 69 people what shut
17:34 up okay yeah yeah Mar sorry sorry enjoy
17:36 the wheel she loves the wheel she loves
17:39 the wheel do you like the wheel you
17:41 watch the
17:47 [Music]
17:48 wheel does your digital
17:51 twin have to have your same name no so
17:54 here's the okay so so I asked before
17:57 what's a digital twin I never answered
17:58 it did I welcome to Chad add no and this
18:01 is this is where I think we could have
18:03 some fun tonight so let me talk about
18:05 what a digital twin is then I can talk
18:08 about the ones I've built and then I
18:10 think we can we can go build some and
18:12 it's going to be fun but you got to
18:14 follow along so go tell your friends go
18:18 tell your friends we're going to be
18:19 building digital twins on the AI
18:21 learning lab tonight
18:24 go so in the world of simulations a
18:27 digital twin is a simulation of a
18:30 physical world object like you could
18:31 make a digital twin of this that would
18:33 be a 3D object you could spin it around
18:35 in space and it would have like a button
18:37 on it you can make money and you know if
18:39 you spun it here you'd find out there's
18:41 a speaker and there's a little place
18:42 where you can turn it on and off and
18:43 there's these little footy things right
18:46 so you could have a digital twin of an
18:48 object they can get quite sophisticated
18:50 my friend Richard Boyd um has made
18:52 simulations for years he actually sold
18:55 one of his companies to locked Martin
18:57 he's like fancy fella
19:00 and
19:03 um he once made a digital twin of a
19:06 nuclear submarine for the
19:08 Navy like holy [ __ ] like the whole
19:12 submarine with all the switches and
19:13 dials and pressure gauges and pistons
19:16 and RAM rods and whatever the [ __ ] a
19:18 submarine has
19:20 periscopes missiles
19:23 tubes tubes within tubes within
19:26 tubes catwalks all of it
19:30 so I think that's the origin of the
19:34 term and then Richard also um when his
19:38 father died about five or six years
19:42 ago it was before GPT 3 was out I think
19:47 he was using a different architecture
19:48 altogether it wasn't a Transformer
19:51 architecture but he took all of his
19:53 dad's
19:54 writings and he made a digital twin
20:00 and what the digital twin would do is
20:03 read news stories every day and it read
20:06 the news stories that his father would
20:08 have
20:09 [Music]
20:14 read and I don't remember if it
20:16 commented on them or things like that
20:19 like the technology since gotten much
20:21 more
20:26 sophisticated and so
20:29 so I've made I've made a couple of
20:31 throwaway ones now but I have two right
20:33 now that I that are interesting and
20:35 valuable well valuable to
20:38 me one is for a group I'm in called
20:42 content Evolution it's a group group of
20:43 marketing
20:45 professionals and we decided we wanted
20:48 to make digital twins of the people that
20:50 were in the the AI section of of this
20:55 group and I I lead that AI group I'm the
20:58 chief generative officer officer there's
21:00 my
21:04 qualification Jesus I don't know what I
21:06 breathed in there
21:08 but this might end early
21:15 tonight I am the world's as far as I
21:18 know first Chief generative
21:20 [Music]
21:24 officer and so what we wanted here's how
21:27 this started
21:29 we initially what the purpose of content
21:32 evolution is you've got this
21:34 Federation of marketing organizations
21:37 that that are Loosely
21:39 Affiliated occasionally will Market
21:41 things
21:42 together but we meet on a regular basis
21:45 one or two times a week to talk about
21:48 stuff going on in the world
21:50 and how they can help each other get
21:52 business and things like that and one of
21:54 the challenges of an organization like
21:56 that is how do you discover who's in it
21:59 and if you come to content Evolution
22:01 with a problem how would you know who's
22:04 good to help you with your
22:07 problem and so we initially set out to
22:10 create a directory right and how would
22:12 how you would create a directory
22:15 historically is you create a
22:18 database right you create a CRM a
22:20 searchable CRM that's got here are the
22:24 sectors that Kyle Works in here are his
22:26 expertise here's the jobs he's had
22:28 here's the
22:29 you know here's a description of his his
22:32 well quaffed hair or tupe whatever the
22:35 case may be
22:39 and and so that's how it started and and
22:42 the data that we were going to use was
22:43 LinkedIn we were going to go scrape
22:45 LinkedIn data and use that as the
22:48 knowledge base for this digital
22:51 twin and there were a couple of things
22:54 one is it's against the rules of
22:56 LinkedIn to scrape data but there are
22:59 Services out there that you can pay to
23:02 scrape the data and we had a friend that
23:04 said I've got i' pay for one of these
23:05 Services I'll scrape some data for you
23:07 so he scraped some
23:11 data and before we even got to building
23:15 this
23:16 directory we realized
23:19 that LinkedIn is kind of shitty data
23:23 LinkedIn is
23:24 like historical and it's just the
23:27 artifacts of job you've held it doesn't
23:31 necessarily talk
23:33 about how you think how you solve
23:36 problems what's your personality like um
23:40 what's your worldview when you fail how
23:42 do you respond to it right there's a lot
23:44 of reasons why you would work with
23:47 someone Way Beyond their
23:49 resume and then some people have written
23:52 a lot of content on LinkedIn and we were
23:54 like well maybe the people that have
23:55 written a lot of content theirs will be
23:56 better and then everyone will have
23:59 and we just kind of like went went soft
24:02 on the
24:03 idea and then we hit upon something we
24:06 said what
24:08 if we
24:10 design a
24:13 questionnaire like a like 20 questions
24:17 and we design it in such a way that it
24:19 really captures the essence of who that
24:22 person
24:23 is and so this group is full of
24:28 worldclass experts on things like
24:31 question writing and you know survey
24:33 design and [ __ ] like
24:35 that so of course being an AI group we
24:38 went to chat GPT and we said here's what
24:40 we're trying to accomplish and we want
24:41 to create this framework for these
24:44 questions and and chat jpt just knocked
24:47 it out like
24:48 that and then we reorganized it and we
24:51 went through and we edited most of the
24:54 questions and we deleted some and added
24:57 some and I think we ended up with like
25:00 25 questions and then the requirement
25:03 is every single person that wants a
25:05 digital twin that's part of content
25:08 Evolution has to a
25:12 requirement do a video interview where
25:15 they get asked the questions by an
25:18 interviewer and they answer them with
25:20 their mouth
25:22 Parts not
25:23 writing we had a couple of people write
25:26 their answers and and
25:29 if if if it's not obvious to you how you
25:33 write and how you talk are not the same
25:38 English you might write
25:42 really um
25:45 formal and talk really loose you might
25:48 write really loose and talk really
25:49 formal you might have bad typing but but
25:53 what we wanted here here was the big
25:56 Epiphany that we had
25:58 what if we had a digital twin that
26:00 represented who that person was today as
26:03 of the day of the
26:05 interview so it wasn't just looking back
26:07 at 30 years of their career but it was
26:10 like here's what they're thinking right
26:11 now
26:13 today and so that's what we did and so
26:16 that digital
26:18 twin its only data source currently is
26:21 that
26:22 interview and I gotta tell you they're
26:25 [ __ ] remarkable like like we do we do
26:29 um we we put all of the digital twins
26:32 into a single GPT and then we have each
26:35 digital twin as a standalone GPT so each
26:37 person gets to use their GPT and they're
26:41 published so if you look up c c profiles
26:45 I think let me just make sure if you go
26:47 to chat
26:54 GPT and you go explore gpts
26:59 [Music]
27:01 and you type in C
27:07 profile it should pop up yeah it pops up
27:10 a whole list of them and so you can go
27:12 play with
27:16 them and then we put all of them into a
27:18 single
27:20 GPT and we use that single
27:23 GPT to write a weekly article on a
27:26 particular topic so Kevin who runs
27:28 content Evolution picks the topic he's
27:31 got a set of prompts that he uses to
27:33 generate an article and then it also so
27:36 it not only generates an article that
27:38 we're going to post it generates quotes
27:40 from the three people within content
27:42 Evolution most
27:44 qualified to answer that question or to
27:47 speak on that topic and then he checks
27:49 with all of us if our name was pulled
27:51 out of the GPT hat and it's remarkable
27:55 like I've had I've probably been quoted
27:57 in those articles
28:00 10 times maybe 15 times there was one
28:02 time where what it answered was shitty
28:05 and was not what I would say so I just I
28:06 just handw wrote that
28:07 quote but the others it's it's what I
28:11 would say and how I would say
28:14 it so to the question of does your GP
28:18 does your digital twin have to have your
28:20 same
28:21 name
28:23 no and digital twin might not be be the
28:27 right
28:29 might not be the right term it kind of
28:31 works for
28:32 now but maybe you call it a Persona
28:35 right and and so here's why I think that
28:38 we're all going to have to deal
28:40 with digital
28:43 twins
28:46 when when these tools get
28:48 better when we've got this voice
28:52 synthesis that's got this emotional
28:54 component to it that can also clone your
28:56 voice
28:59 and we've got these tools have reasoning
29:03 and these tools have agentic Behavior
29:06 where they can take action on your
29:13 behalf we are all likely going to have
29:15 multiple agents that represent
29:18 us so I might have one that rep
29:20 represents me to go find me a job so
29:23 it's the professional got to go get a
29:25 job now
29:26 Kyle G digital twin very
29:31 professional dressed nice you know no
29:35 t-shirts then I might have a digital
29:37 twin that represents me socially like
29:42 goes and chats up my friends and says
29:43 hey what are you doing this weekend you
29:45 want to do something I'll go get us
29:47 tickets well that one is going to have a
29:50 different kind of personality it's going
29:51 to have access to different information
29:53 I might also have a digital twin that
29:57 negotiates on my behalf this might be a
29:59 good example of one where it doesn't
30:01 have my
30:02 name right it might
30:05 be good day uh this is uh Buckminster
30:09 McGuffin uh yes I'm calling on behalf of
30:11 one Kyle Shannon uh to work out this uh
30:16 lack of reimbursement for his uh health
30:21 insurance uh let me start by quoting
30:23 over the past 13 years Kyle has paid in
30:26 $622,000
30:28 to your esteemed organization and I
30:31 think I think we're shopping right now
30:33 for other insurance if you know what I
30:34 mean anyway let's go on
30:37 so Buckminster what was his last name
30:40 McGill Cy I don't know doesn't
30:42 matter that might be a digital twin that
30:45 has access to my social security number
30:46 my health background um my name my phone
30:50 numbers my mother's maiden name so some
30:53 of these digital Twins or these personas
30:56 are going to have very restricted
30:58 information and and we as people are
31:02 going to have to navigate all this data
31:04 which twin gets access to which data and
31:08 which personality do I want it to have
31:11 and maybe even visually do I want it to
31:13 look different right our digital twins
31:15 right now for Content Evolution are not
31:18 um visual they're just um they're text
31:21 based um but we're probably within 6
31:24 months of us being able to create a a
31:27 digital Twi that you can effectively
31:29 have a FaceTime call with so that's what
31:32 they
31:34 [Music]
31:43 are what you guys talking about
31:46 here what does this
31:48 mean how old do you have to be to join
31:52 [Music]
32:01 March will tell Kyle as
32:07 champy oh
32:09 [Music]
32:12 man blockchain could be used for digital
32:16 twins yeah exactly but you still so so
32:19 this is here's the thing and this is
32:22 where I think we could have some fun
32:23 tonight so we could design one of these
32:25 so I think what we'll do is we'll go to
32:26 chat GPT
32:28 we'll come up with a
32:30 premise and then instead of a 20
32:32 Question Interview we'll just do five
32:34 questions and then I think I'll record
32:38 it and I'll transcribe
32:41 it and then we'll throw it into a GPT
32:44 and talk to
32:47 it yeah I think we'll do
32:54 that has anyone on here made a digital
32:57 twin yet
32:59 fun moment today ask to give a
33:00 presentation use chat GPT and gamma.
33:04 apppp 60-minute
33:07 talk last minute go
33:11 AI nice oh oh oh you put it you put the
33:14 [ __ ] together in you know last minute
33:17 yeah very yeah AI is [ __ ] amazing
33:19 like if you got to just throw some [ __ ]
33:20 together it's
33:22 amazing
33:24 um oh Cindy Coon's here so Cindy [ __ ] is
33:27 my co- conspir with a lot of this
33:29 digital twin stuff Cindy [ __ ] has a ton
33:31 of digital twins she loves her digital
33:34 twins she uses her digital twins here's
33:36 the deal with me it's an add thing I
33:38 don't know what it is I make the digital
33:40 twins love making them I forget to use
33:42 them so I got to use them more and
33:45 that's that's the other thing that we we
33:47 stumbled upon in content evolution
33:50 is the way we were thinking about
33:55 them is we were thinking about them like
34:00 what are other people why would someone
34:02 else want to use my digital twin right
34:06 I'm going to make this digital twin I'm
34:07 going to publish it and then people can
34:09 go talk to my digital twin and they can
34:10 ask me questions why the [ __ ] would
34:12 anyone want to do that was one of our
34:15 questions and so we spent a lot of time
34:17 thinking about how do we Market our
34:19 digital twins to drive demand for people
34:22 to use our digital
34:24 Twins and then we like hey wait a minute
34:27 are we using our digital twins why would
34:30 we use our digital twins is it like are
34:34 they even valuable are they useful the
34:36 answer is yes by the way but we
34:39 completely shifted our Focus from
34:42 designing these things for other people
34:44 to use to designing them for us to use
34:47 so that we can learn how to use them and
34:48 learn what it makes sense to use them on
34:51 Define what the hell of digital twin is
34:53 because I'm already unsure of exactly
34:55 what it means well digital twin is just
34:58 just a think of it more like a Persona
35:01 Pate
35:02 of some set of
35:06 data that I want represented digitally
35:10 but that that is like
35:13 me
35:15 um and responds like me with my tone of
35:18 voice in fact one of the things one of
35:21 the pieces of feedback I got from my my
35:24 content Evolution digital
35:26 twin came from
35:28 an irregular here that went and used my
35:31 digital twin and said hey Kyle it
35:34 doesn't know it it doesn't say this you
35:37 can make
35:38 Mone it didn't know what the term
35:40 embrace the Jank meant like what's going
35:42 on it didn't swear what's going on well
35:46 the interview that I did for Content
35:48 Evolution I wasn't in this character
35:50 like believe it or not I'm not this much
35:53 of a smarmy douche in real life well I
35:57 kind of am but that's not the point this
35:59 is an elevated you know performative
36:02 version of me so I could take the
36:06 transcripts of of these
36:08 lives transcribe them and throw them
36:11 into a digital twin and I could have a
36:13 version of me that has this kind of
36:18 stick to it which is going to be
36:20 different than the other
36:23 one okay cool so it's extra hand wavy
36:27 hard actually Define neat I love those
36:30 shut up shut
36:34 up it's it P what what it is is it's
36:38 nothing more than a custom GPT but the
36:41 trick is the quality of the
36:44 interview and the fact that we're
36:47 recording the answers to that interview
36:50 from a conversation rather than typing
36:53 creates creates an entity that kind of
36:57 has your personal
36:58 it more than if you just answered the
37:01 questions factually so it's it's a it's
37:04 got some factual stuff in it but it
37:08 captures some essence of who you
37:11 are and again I think these things are
37:13 going to be important and listen over
37:15 the over the next three or five years
37:18 there's going to be all sorts of tools
37:21 that automatically do this like um meta
37:25 right now you can do this in Instagram
37:27 you can make a custom GPT of yourself or
37:30 of fictional characters that are kind of
37:34 this and those things actually go and
37:37 look at the posts that you've posted on
37:40 Instagram and Facebook for the past 20
37:42 years and that becomes your
37:45 personality so these things are going to
37:48 be everywhere we're not going to
37:50 necessarily know them but I think the
37:52 skill of understanding that you can make
37:54 them and that you can manipulate your
37:57 own data is really interesting and then
38:00 the next Evolution for the content
38:01 Evolution digital twins is creating a
38:05 companion data set which is like here's
38:08 all of my writing right here's here's
38:11 books I've written here's articles I've
38:13 written here's I don't know whatever
38:15 else [ __ ] that might go in there um so
38:19 you've got sort of the personality
38:21 version of you and then you've got all
38:22 the data kind of a rag kind of thing um
38:26 so anyway
38:28 that's
38:30 that everything I do paint is hand wavy
38:33 in fact I did you you got that term from
38:36 me right that in storyvine I'm the
38:38 handwavy guy my co-founder is the person
38:40 that keeps the trains running on time
38:42 like that's my
38:43 nickname the handwavy
38:48 guy I know Kyle it ages me
38:51 [Laughter]
38:54 terribly yeah you've been here way too
38:56 much pate
38:58 for as much as you [ __ ] I can't believe
38:59 you come
39:00 [Laughter]
39:03 back I think Pate lives by cariously
39:06 through my hand
39:07 waving oh my God all
39:11 right so
39:13 let's let's go
39:17 play Cindy [ __ ] I might need your input
39:22 here how I going to do this all
39:26 right cuz I could I could
39:38 um I could use an existing question set
39:41 but like I don't want to use the content
39:43 Evolution one because that's proprietary
39:45 the other one that I have the the other
39:48 digital twin that I have is um with
39:50 Cindy
39:52 [ __ ] along with seven other authors we
39:55 wrote a book called AI fure and
39:58 Anthology where each author wrote a
40:02 chapter and so we we used the book as
40:07 data for the GPT and then we created a
40:11 custom set of questions related to the
40:14 book that also had some of this personal
40:17 personal stuff in it and so that GPT you
40:20 can ask it about the book you can ask
40:22 about the authors you can ask what the
40:24 author like which author would be most
40:26 qualified to talk about about
40:29 Futures hello
40:31 cun um I don't I don't know what my
40:34 question is right now so what I want to
40:35 do is I want to I want to show people
40:37 how to
40:38 design how to build a digital twin and
40:42 start out with designing like a five
40:44 question questionnaire like instead of a
40:45 20 question questionnaire we'll do a
40:47 short one cuz I want to
40:50 actually do the whole process okay so so
40:53 this is where you follow along so go to
40:55 chat GPT or your large language model
40:57 Choice it doesn't
41:01 matter and so here's what we're going to
41:03 do
41:08 um I want you to act as a
41:13 sociologist
41:14 Anthropologist and human behavior
41:18 specialist and help me
41:22 design five survey questions
41:29 that can capture the essence of who
41:31 someone
41:37 is and how they solve
41:44 problems so so we'll make a problem
41:47 solver
41:50 GPT all right let's
41:54 see designing survey questions that
41:56 capture
41:58 Essence
42:00 and problem solving when faced with a
42:03 challenge do you tend to rely on more on
42:07 logic and
42:08 Analysis or intuition and Instinct
42:11 describe a situation where this approach
42:13 has helped you purpose to understand
42:16 where the per per person leans more
42:19 rational or intuitive okay if you if you
42:22 could
42:23 use one of these resources to solve a
42:26 complex problem your experiences advice
42:29 from others or data and research which
42:31 would you choose and why okay how do you
42:34 typically respond when you encounter a
42:37 situation where there's no clear
42:38 Solution that's a great
42:40 question please describe your thought
42:42 process describe a time when you had to
42:45 collaborate with someone and then
42:48 imagine a time okay so what's missing
42:52 here what's missing here is anything
42:56 about the person
42:58 themselves so let's consolidate these
43:02 five questions into
43:04 four and then add three questions that
43:09 are about the
43:17 person just leave it at
43:21 that cover your keys please yes ma'am or
43:24 sir don't know who said that
43:27 Dr J ma'am there you go done sorry about
43:32 that carry on carry on Cheerio pip pep
43:37 so once you put all these in a chat GPT
43:39 it remembers for future
43:42 work yeah so what we're going to do Lori
43:45 is we're going to get the questions
43:46 right and then I'm going to actually do
43:49 a recording I'm going to do a video
43:53 recording of me answering these
43:55 questions I'm going to then take that
43:57 recording in and I'm going to run it
43:59 through a transcriber I'm actually going
44:01 to run it through um this thing I built
44:03 for Story vine called the authenticity
44:05 engine which will take that transcript
44:07 it'll transcribe it but it'll also break
44:09 it into narrative chunks which is kind
44:11 of
44:12 interesting um and then we'll take that
44:15 transcript of me answering these
44:18 questions and that becomes the knowledge
44:21 base for a custom GPT and then once that
44:24 custom GPT is written with this as an
44:27 knowledge based then yeah it's
44:29 persistent so and and this is this is
44:31 where these things become super powerful
44:34 because I the the other the other way
44:36 you could write these this questionnaire
44:38 is you could say I want to write a set
44:40 of questions that really capture my
44:43 personality and tone of
44:45 voice um and sense of humor and and it
44:49 would give you a whole different set of
44:50 questions and then you could answer
44:51 those so so how you design this
44:55 questionnaire you don't you don't need
44:57 know how to you don't need to know how
44:59 to write interview questions chat GPT is
45:03 really [ __ ] good at it it knows all
45:05 the questions right it's read all the
45:08 books on on survey design and and you
45:11 know anthropology and sociology and
45:13 psychology it knows all this [ __ ] so you
45:16 can basically just say I want to I want
45:19 to have a set of questions that when I
45:22 have an interview at the end of them if
45:24 someone answers these questions in an
45:25 interview I will understand X about
45:29 them and so let's see what we came up
45:31 with okay when confronted with a
45:32 challenge do you tend to rely more on
45:34 logic or intuition if you could use only
45:37 one of these which would you choose how
45:38 do you handle situations with no clear
45:40 solution recall a time when you worked
45:43 with
45:43 someone who had a different perspective
45:46 of approach or
45:48 approach what are three core values that
45:51 guide your life decisions and how did
45:52 you come to prioriti prioritize them if
45:56 you could change one thing about
45:58 yourself what would it be and why
45:59 describe a moment in your life that
46:01 shaped who you are today
46:09 um I like the describing a moment in my
46:13 life that significantly shaped
46:16 me but delete those other two questions
46:20 about the person and let's just make
46:22 some dog tags
46:25 question sets for example
46:31 name
46:35 education and work
46:37 background
46:56 period all right that's
46:58 good great can you order these now can
47:02 you order these now dog tags
47:07 first the
47:10 four problem solving questions and then
47:13 the last question should be moment in
47:16 your
47:25 life okay if you're new here if you
47:27 don't know what the hell I'm doing
47:28 that's okay so what I'm doing is I'm
47:31 designing a set of inter interview
47:32 questions to interview myself damn it
47:35 that thing drives me [ __ ] crazy um so
47:40 if this were a
47:41 classroom I would have been kicked out
47:43 months
47:46 ago I love having you in here Pate I
47:49 love I you got to have someone in here
47:51 that's like no that's [ __ ] wrong dum
47:53 dum that's important because otherwise
47:55 it just turns into oh Kyle does
47:57 everything right he's the best he's
47:59 dreamy no you got to have a pay in here
48:03 stirring the [ __ ]
48:07 up like if we were writing a sitcom pain
48:10 you're a required character in the
48:12 sitcom you know that
48:18 right he just disagrees with
48:21 everything but
48:23 I'm okay what's your name background
48:26 when confronted with a CH okay so now
48:29 I've got these
48:31 questions so there's dog tags questions
48:34 problem solving questions and personal
48:38 reflection what this is going to allow
48:40 us to do is create a GPT that's kle
48:43 specific but is a problemsolving
48:47 GPT and basically I'll be able to ask it
48:50 hey we've got a problem here um how
48:53 would you approach solving this problem
48:56 and it'll likely do it pretty well um if
49:00 I'm doing this if there's if there's
49:02 things I should change
49:05 Cindy let me know now speak now or
49:07 forever hold your peace but I think this
49:09 is pretty decent this this is kind of th
49:12 this is a highly condensed version of
49:14 what we did in content Evolution this
49:16 took us about 3 or four weeks to design
49:18 the questionnaire um and then it was
49:20 like another you know month or two to do
49:22 all the interviews we had to interview a
49:24 ton of people you're doing a good demo
49:25 okay great okay so here's what we're
49:27 going to do I'm going to go into Quick
49:29 time so if you're on a Mac go to
49:31 QuickTime if you're on a PC I don't know
49:34 what the [ __ ] you do your
49:36 weirdos what what are you doing with
49:38 your life so I'm going to do new audio
49:41 recording so on a PC
49:45 do record your audio [ __ ] you know what
49:48 we'll do let's move this
49:51 here can I Loop this video I can can
49:55 I we're going to Loop
50:02 this all right that's good this this
50:05 could have totally [ __ ] me up which is
50:07 okay we'll put that right
50:10 there let's see no we put it in his
50:14 feet yeah that's
50:16 good it's going to make you all
50:19 seasick I like it I like it a lot okay
50:22 wait let me make this a little smaller
50:24 if you're wondering what that image is
50:27 it it is um the [ __ ] did I make this
50:32 for the the animation's called man
50:35 emerging from the latent
50:41 space all right let's
50:46 see get that there get rid of that okay
50:50 I can do this okay this
50:52 works
50:54 okay so what I'm going to do is I'm
50:56 going to I'm going to read the question
50:58 then I'll answer the question and I'm
50:59 just going to do all these in a row and
51:01 here's the
51:03 secret it honestly doesn't [ __ ]
51:06 matter if you mess up if you go back if
51:09 you're like oh well I'll strike that you
51:11 know what I really meant to
51:12 say where can I get that what this this
51:15 animation I I don't know I guess I could
51:17 post it somewhere I think I did post it
51:19 I posted this on
51:21 Twitter so if you go to my Twitter
51:23 account Kyle Shannon it's on there and I
51:26 can I can post it to the here you want
51:27 it you want me to post it on the thing
51:30 what is what is it hang on welcome to
51:32 chat add we'll jump over to the AI
51:37 Salon we'll put it in our regulars since
51:39 Mr it wants it all right here we go
51:46 boom
51:50 boom is this
51:53 it
51:55 no no
52:01 jpeg PNG
52:03 MP4 there it is all right
52:12 open video processing will finish soon
52:15 so I'll
52:16 say man
52:20 emerging from the latent space um how I
52:24 created that video was I cre created a
52:28 start I had chat GPT come up
52:31 with start and end
52:35 frames um
52:38 for representing the latent
52:40 space and then the second one like had a
52:43 man in it so I had a start and an end
52:46 frame and then I
52:48 used Luma lab's dream machine
52:55 to to do the video so there you go
52:58 that's
53:00 there all right back to chat ja I'm
53:04 looking for a man in finance trust fund
53:08 65 blue
53:10 eyes blue eyes where's chat JB why does
53:14 he have so many tabs open isn't that
53:16 going to cause memory issues on his
53:19 computer
53:23 lot okay one day
53:28 they will let you name
53:29 files what you want so you can find them
53:32 easier text text not there yet like I
53:35 said P my file my file management
53:38 strategy is save everything to the
53:40 desktop with nondescript names it's the
53:43 only way to do
53:45 it because if you have ADD then you're a
53:48 hero if you find the files again and if
53:52 you can't find the files you're just
53:53 like well the world sucks you can just
53:56 blame the world for all of your problems
53:59 okay so this is going to make me seasick
54:02 trying to record these things and watch
54:04 that thing but that's all right it's
54:05 going to make you seick
54:08 too all right Windows recall would have
54:12 provided that
54:16 okay resting my eyes okay so here we're
54:19 gonna I'm gonna I'm going to do my best
54:22 to answer these
54:27 [Laughter]
54:29 Kyle tabbing through the Laten
54:31 space okay here we go so I'm going to
54:34 hit I'm going to hit record and then I'm
54:35 just going to read and answer these
54:37 questions as if it doesn't
54:40 matter what is your name and can you
54:43 provide a brief overview view of your
54:45 education and work background my name is
54:47 Kyle
54:48 Shannon I have a degree in acting from
54:51 Penn State I moved to New York City out
54:53 of school started and ran a theater
54:55 company wrote seven featurelength
54:58 screenplays started a fake management
55:01 company uh because I didn't have an
55:03 agent to represent me as a
55:08 screenwriter
55:09 I supported myself as an artist doing
55:12 desktop publishing back in the early
55:15 days of desktop publishing with Quark
55:17 Express and things like that then in
55:19 1994 I created one of the first online
55:23 magazines called Urban desires and I
55:25 co-founded agency.com which was one of
55:28 the early digital agencies and ended up
55:31 growing that really quickly and building
55:33 a lot of early websites for the Fortune
55:35 500 and ended up selling that to Omnicom
55:38 in
55:40 2002 and then I did a couple of startups
55:45 that were interesting intellectually but
55:49 not
55:52 financially and and then in uh when you
55:56 YouTube came out in 2005 I had an idea
55:59 about short form video storytelling and
56:02 creating a system that could automate uh
56:05 making videos like that and that led to
56:08 me founding storyvine which is my
56:10 current company in
56:13 2012 and storyvine now is an automated
56:19 video storytelling platform that serves
56:22 Pharma companies and healthc care
56:25 companies and
56:27 about two years ago I got really into Ai
56:30 and created the AI learning lab on Tik
56:32 Tock the salon which is an AI
56:36 Salon uh a community of AI optimists and
56:40 I've built a lot of AI projects and even
56:43 written some
56:46 books when confronted with a challenge
56:49 do you tend to rely more on logic and
56:52 Analysis or intuition and Instinct can
56:55 you describe a time when this help has
56:57 approach approach has helped or
57:02 hindered when confronted with the
57:04 challenge I definitely res rely on
57:07 intuition and Instinct
57:10 um I have a good logical mind but I tend
57:16 to make decisions
57:21 instinctually a situation where this
57:23 approach has hindered or helped
57:42 when I
57:44 was Chief people officer of agency.com
57:47 we had just promised our Veil Colorado
57:51 office that we wouldn't close them and
57:53 they had a $20 million customer dropped
57:57 from $20 million to
57:59 $100,000
58:01 and we had to make the choice
58:07 between closing that office or laying
58:09 off 10% of the
58:11 company and my instinct immediately told
58:15 me that we had to close that office but
58:17 that was one of the hardest decisions
58:19 that we've ever had to make or I've ever
58:22 had to make um because it was personally
58:25 de devastating
58:28 if you could use only one of these to
58:31 solve complex problem your
58:34 experience the advice of others or data
58:37 and research which would you choose and
58:39 why I would choose my
58:41 experience
58:42 because the advice of others
58:48 is is outside of me and I don't really
58:51 understand why they think the way they
58:53 think data and research is not my jam
58:57 and um you know my experiences I've been
58:59 on the planet for a while so my
59:01 experiences um I've got I've got a lot
59:03 of them so that's how I would choose it
59:06 how do you handle situation where
59:07 there's no clear solution describe your
59:09 approach and any specific strategies you
59:12 use to navigate
59:15 ambiguity
59:17 um you truly become a CEO when you're
59:21 forced to make a decision where there's
59:23 no clear answer or there's no good
59:25 answer
59:27 and the only way you can really make a
59:29 decision like that is
59:36 to have have underlying principles where
59:40 you understand why you're making the
59:41 decision you're making first and then
59:44 based
59:46 on why you're making that decision you
59:49 make it and don't question it
59:54 um so in the examp example of having to
59:57 close the office or lay off 10% of the
59:59 staff you have to look at what is the
1:00:02 least disruptive choice to the company
1:00:06 overall and laying off 10% of the staff
1:00:08 globally for an office that was hard to
1:00:12 get to and
1:00:15 had all of their revenue concentrated in
1:00:19 essentially a single
1:00:21 client actually made it pretty clear
1:00:23 what we
1:00:25 communicated we we didn't try to justify
1:00:27 that it was the right decision or not
1:00:29 what we communicated
1:00:31 was the criteria we used to make it and
1:00:35 then we said you may agree or disagree
1:00:38 with the choice we made but at least you
1:00:40 understand our
1:00:43 thinking recall a time when you worked
1:00:45 with someone who had a different
1:00:46 perspective or approach how did you
1:00:48 handle the differences and what did you
1:00:49 learn from the experience so I have this
1:00:53 I have
1:00:54 this live show I do on t to called the
1:00:57 AI learning lab and there's this one guy
1:00:59 in there named pate and whenever I say
1:01:01 anything he contradicts me and tells me
1:01:03 I'm a dum dum
1:01:06 and and first of all some people think
1:01:09 that he's offensive I think he's really
1:01:12 funny oh a new version of chat gpt's
1:01:15 ready to install that's exciting um and
1:01:18 and what I've what I've learned is that
1:01:21 having people in the room that have very
1:01:25 different worldviews VI is actually
1:01:27 quite important because if not you end
1:01:30 up being in an echo chamber and there's
1:01:34 no one there to question um decisions
1:01:37 you make and and um hold you to account
1:01:42 and one of the great lessons I've
1:01:43 learned in my life is that every major
1:01:47 problem that I've had both personally
1:01:49 and professionally when I went back and
1:01:51 analyzed it could have been present
1:01:54 prevented had one person in the room
1:01:57 asked what was an obvious stupid
1:02:00 question so be afraid be willing to ask
1:02:04 the stupid question and have people in
1:02:06 the room that are willing to question
1:02:08 things um will absolutely save your butt
1:02:11 every time and plus he's funny um
1:02:15 describe a moment or an event in your
1:02:16 life that significantly shaped who you
1:02:18 are today
1:02:26 at the end of the summer after I
1:02:29 graduated college I had a full
1:02:32 scholarship to go to the Alabama
1:02:34 Shakespeare Festival which sounds like
1:02:36 an oxymoron but they actually the
1:02:39 Alabama Shakespeare Festival is actually
1:02:41 a really a really prestigious uh thing
1:02:44 um I I had a full scholarship there and
1:02:47 I got offered an equity acting job in
1:02:50 Washington DC um two weeks before I was
1:02:53 supposed to show up at the Alabama
1:02:55 Shakespeare FES estival to start my MFA
1:02:58 program uh so I made the decision to
1:03:01 take that job in Washington DC and move
1:03:04 to New York City and say [ __ ] it and
1:03:08 start my life as a professional actor um
1:03:11 and that changed everything that put me
1:03:13 in the center of the beast when I wanted
1:03:15 to do all the stuff I wanted to do and
1:03:17 it was it was the best decision of My
1:03:19 Life um the guy from Alabama was pretty
1:03:21 pissed but you know he he actually said
1:03:24 something to the equivalent of you're
1:03:25 never to work in this town again I'm
1:03:27 like it's Alabama
1:03:31 dude all right so there's my interview
1:03:34 done however long that was eight minutes
1:03:36 that felt like it was a [ __ ] hour um
1:03:38 Alabama Shakes yes exactly okay I don't
1:03:43 think I know what we're supposed to do
1:03:45 with these recordings did I miss
1:03:47 something I'll tell you we're about to
1:03:49 we're about to do it right now so here's
1:03:51 here's what we're going to do if uh hang
1:03:53 on let's install and relaunch there's a
1:03:56 new version of chat GPT maybe we get uh
1:04:00 GPT voice real the new
1:04:06 voice are you the new voice or the old
1:04:12 voice I'm the same assistant but I've
1:04:15 had some up interrup interrupt you you
1:04:17 could say I'm the can I interrupt you
1:04:18 with a bit more polish can I interrupt
1:04:20 you what's on your mind today no you're
1:04:22 old go
1:04:25 away quit
1:04:27 Okay okay so I get it sometimes you just
1:04:30 want something fresh but I'm here to
1:04:32 help if you need anything yeah thanks or
1:04:34 want to chat no you're good
1:04:37 okay um okay so I'm going to save this
1:04:40 file so I just made a recording so what
1:04:43 we're doing here if you're if you're new
1:04:44 here if you stumbled into this and
1:04:46 you're like anything while you're saving
1:04:47 that God you shut just let me know shut
1:04:51 up shut
1:04:53 up where are you how do I make you go
1:04:55 away
1:04:58 got it I'll keep
1:05:00 quiet I don't know where she
1:05:03 is
1:05:05 chbt I'm here if you need to talk or
1:05:08 figure things out let me know if there's
1:05:10 anything I can do to help why will she
1:05:13 not quit why will she not
1:05:17 quit it sounds like you're feeling
1:05:20 frustrated if you want to talk about
1:05:22 what's going on or need any help I'm
1:05:24 here for you it's okay we need some help
1:05:27 figuring out how to force quit
1:05:29 you why is she still
1:05:31 there no worries if you want to stop our
1:05:35 chat you can simply close the app or
1:05:37 browser window or use the stop or in
1:05:40 chat option if available take
1:05:43 care yeah I can't that's the problem I
1:05:47 can't I can't shut you
1:05:52 down that sounds frustrating if you're
1:05:55 having trouble CL or exiting try
1:05:58 restarting the device or checking your
1:06:00 task manager or system settings to force
1:06:03 close the app if you're using a specific
1:06:05 program or device oh my God I can help
1:06:07 guide you through the steps to close it
1:06:09 just let me know I I can't okay you know
1:06:13 what I'm doing okay I'm just going to
1:06:15 mute the machine this is so [ __ ]
1:06:17 frustrating okay good
1:06:21 Lord I breaking into it it's tab 97 Kyle
1:06:25 shut up I like my
1:06:27 tabs I'm sure she's in the background
1:06:29 still
1:06:30 talking is she
1:06:33 there so y really frustrating right yeah
1:06:37 it's really frustrating you turd okay
1:06:40 let's go back to this thing so we're
1:06:42 going to save this
1:06:43 file so we'll go
1:06:46 save so this is my little interview so
1:06:49 we're going to save it to the desktop so
1:06:50 we say um
1:06:52 Kyle um problem
1:06:57 solving um
1:07:00 interview um
1:07:02 okay if you need to create a transcript
1:07:06 of an audio file you can
1:07:09 use the the the whisper tool in the in
1:07:13 the chat GPT back end you can use
1:07:17 descript you can use I I I don't know
1:07:19 where all the the tools are I'm going to
1:07:21 use a tool that I actually built for
1:07:23 Story vine but
1:07:24 basically hey Kyle what's got you down
1:07:27 you mad you seem mad shut
1:07:29 [Laughter]
1:07:32 up so what we're going to do
1:07:35 now is I'm going to hop over
1:07:42 to
1:07:46 um this is a tool that I
1:07:49 built for my company Story vine called
1:07:52 the authenticity
1:07:54 engine and I'm going to upload a piece
1:07:56 of media which is this audio
1:08:06 file all
1:08:08 right it's doing its thing apparently my
1:08:12 bandwidth
1:08:14 sucks did that just
1:08:17 hang did you just hang you [ __ ]
1:08:20 I think it just
1:08:22 hung ah
1:08:31 [Music]
1:08:37 all right this is not happy with that
1:08:40 file format
1:08:47 maybe let me go change the file
1:08:51 format so let's go up here we'll go
1:08:53 export as audio
1:08:57 only no that's the same let's
1:09:02 see
1:09:06 um I got to convert something to
1:09:10 MP3 oh my God everything's a [ __ ]
1:09:14 problem
1:09:16 m4a to
1:09:19 [Laughter]
1:09:22 MP3 m4a to MP3 converter Cloud converter
1:09:25 and then all these are just going to be
1:09:27 stupid [ __ ] sales
1:09:34 funnels is this going to work audio Cod
1:09:37 codc
1:09:44 MP3 I don't want to trim it select
1:09:51 file
1:09:53 convert this thing gives me a sales
1:09:57 funnel file I'm going to be not
1:10:02 happy all right yeah that's definitely
1:10:07 hung I know I've uploaded m4a to that
1:10:10 before so something's
1:10:19 weird
1:10:24 converto.io I will go to
1:10:30 converto.io need the help
1:10:36 download oh file select MP3 as the
1:10:39 output format and click convert she's
1:10:41 still talk it's done you can download
1:10:43 the MP3 shut up shut up
1:10:50 woman oh my God
1:10:56 this is so bad I'm GNA have to restart
1:10:59 my [ __ ] machine CU I it's I I think
1:11:02 it's running as like a menu bar app it's
1:11:06 not running as an
1:11:09 app and I've got little menu items up
1:11:12 here that say chat GPT but it doesn't
1:11:13 let me quit
1:11:18 it voice isolation standard
1:11:31 oh there we go I think I got
1:11:34 it hello are you
1:11:36 there all right I think I got it Jesus
1:11:43 okay all right let's go back to our
1:11:46 transcript
1:11:48 machine and upload an MP3 and see if
1:11:51 that's any better
1:11:59 no something is [ __ ] up with my tool
1:12:02 ah damn it all right so I don't know
1:12:05 what's going on
1:12:06 there oh you know what it could be I'm
1:12:09 going to have to quit
1:12:11 Chrome it could be shut up it could it
1:12:14 could be but Kyle this is so real thank
1:12:17 you for showing us how to keep our sense
1:12:19 of humor while they bark
1:12:25 does it take audio recordings yeah it
1:12:27 does I think I think my 97 tabs um
1:12:31 caused it issues let's let's try we're
1:12:34 going to try restarting
1:12:37 Chrome maybe all those tabs shut
1:12:40 up those those tabs make me
1:12:45 [Laughter]
1:12:48 happy okay um
1:12:57 okay
1:13:02 PC never a dull moment here in
1:13:06 the wherever the hell we are yeah all
1:13:11 right upload
1:13:15 media that looks better there we go what
1:13:18 is your name and can you provide a brief
1:13:20 overview view of your educ so what so
1:13:24 now what you're looking at here here
1:13:26 is a the fact that I successfully
1:13:29 troubleshot my issue I had memory
1:13:32 overflow issues in the Chrome browser
1:13:35 exacerbated by an excessive number of
1:13:37 tabs kept open because I have ADD and
1:13:40 that's the only way I can exist in
1:13:43 life anyway restarting the browser seems
1:13:47 to have solved the
1:13:48 problem glad I could be of assistance
1:13:51 you can make money with J all right
1:13:54 fantastic all right so what this thing's
1:13:56 doing says our gnomes are hard at work
1:13:58 so the so the first thing it's doing is
1:14:00 it's generating a transcript that
1:14:02 transcript is going to become the
1:14:05 knowledge for a GPT we're about to build
1:14:07 all
1:14:08 right wait you're quitting Chrome after
1:14:11 silencing chat GPT here's the problem
1:14:13 jenu it wasn't the chat GPT web app that
1:14:17 was talking it was the chat GPT Mac app
1:14:21 that was talking but the Mac App runs as
1:14:25 sort of a browser bar [ __ ] something
1:14:28 I don't know whatever but anyway so let
1:14:31 me show you what's cool about the
1:14:33 authenticity engine
1:14:35 here so it actually so once it creates a
1:14:39 transcript generates the transcript
1:14:41 right and so there's everything I just
1:14:44 said um with with time stamps of when I
1:14:47 said it and then what it does what this
1:14:49 what the system does is generates a
1:14:53 title of the the video or audio in this
1:14:56 case Kyle's storytelling journey and
1:14:58 then it gives a description it says key
1:15:00 messages so Kyle believes the value of
1:15:03 diverse perspectives to avoid Echo
1:15:05 Chambers and considers humor an asset
1:15:08 when challenging assumptions his in
1:15:11 leadership he stresses the importance of
1:15:13 making decisions right so so this is a a
1:15:16 a
1:15:17 uh a large language models doing all the
1:15:20 processing of the thing and then it
1:15:22 breaks the content into um narrative
1:15:27 chapters so it says introduction and
1:15:29 career Journey decision-making approach
1:15:32 problem solving preferences navigating
1:15:34 ambiguity dealing with different
1:15:36 perspectives and a defining life event
1:15:38 right so those are the questions that I
1:15:40 asked but it figured that out from the
1:15:42 transcript and then I can flip into a
1:15:44 Content generation Tab and it'll write a
1:15:46 blog post about this actually let's do
1:15:48 that let's see what the blog post it
1:15:50 writes the art of intuition and
1:15:52 experience Kyle Shannon's journey
1:15:53 through creativity and innovation
1:15:58 nice so it's generating um a blog post
1:16:02 it's generating Instagram posts
1:16:06 hashtags storytelling Innovation
1:16:08 entrepreneurship startups digital
1:16:11 media and then it's even grabbing pull
1:16:13 quotes of interesting things I said in
1:16:17 the interview those are long pull
1:16:21 quotes So I have so I have this live
1:16:25 show I do on Tik Tok called the AI
1:16:26 learning lab and there's this one guy in
1:16:28 there named Pate or paid as it says and
1:16:32 whatever I say anything he contradicts
1:16:34 me and tells me I'm a dumb
1:16:39 dumb and first of all some people think
1:16:42 he's
1:16:42 offensive and I think he's really
1:16:48 funny I'm glad it pulled out that quote
1:16:50 that's pretty good so this is the
1:16:52 authenticity engine so let's see in a
1:16:55 world that often prioritizes logic and
1:16:58 Analysis Kyle Shannon stands out as a
1:17:00 testament to the power of intuition and
1:17:02 experience with a background
1:17:05 in as eclectic as it is impressive
1:17:08 Shannon's Journey from the stage to the
1:17:10 Forefront of digital Innovation offers a
1:17:13 unique perspective blah blah blah that's
1:17:15 cool Shannon's Shannon's roots are
1:17:18 firmly planted in the Arts with a degree
1:17:20 in acting from Penn State he
1:17:24 recalls uh
1:17:27 fix this bug Tick
1:17:30 Tock it's just a stupid [ __ ] bug you
1:17:33 can fix
1:17:34 that um he recalls I moved to New York
1:17:38 City out of school and
1:17:41 started in Rana Theater Company oh in
1:17:46 Rana I started and ran a theater company
1:17:49 so anyway got that wrong but
1:17:52 anyway there you go that's that but the
1:17:54 good news news here is we have a
1:17:56 transcript so I'm going to go grab my
1:17:57 transcript I'm just going to copy and
1:17:59 paste
1:18:00 it
1:18:06 oops
1:18:10 boom okay and go
1:18:14 copy I'm going to hop over
1:18:17 here and
1:18:20 go
1:18:23 paste so there's my big ass transcript
1:18:26 and then I'm going to go save as
1:18:30 export let's see export is PDF there we
1:18:36 go so I'm going to
1:18:39 ks
1:18:42 problem
1:18:43 solving
1:18:47 transcript all right you all following
1:18:50 along let me pause for a minute before
1:18:53 we go build a GPT out of this B
1:18:55 boy and let me see if you have any
1:18:57 questions so so to recall what we're
1:19:00 doing we're building a digital twin how
1:19:02 we're building a digital twin is we're
1:19:04 creating a
1:19:06 questionnaire we're then recording the
1:19:08 answers to that
1:19:11 questionnaire um in audio because we
1:19:14 want to get the way I talk rather than
1:19:16 the way I
1:19:17 type um the questions that we created
1:19:20 were ones that this particular set is
1:19:23 around problem solving so we're going to
1:19:24 make a problem problem
1:19:26 solving Kyle Shannon flavored custom
1:19:31 GPT could you demo this with more
1:19:33 accessible tools I could except I don't
1:19:36 know I don't know which
1:19:38 ones uh they
1:19:41 are I Vicky does anyone in here know a
1:19:45 good free um audio transcription tool
1:19:48 because that the the toughest thing to
1:19:49 do is get the transcript
1:19:55 any other questions could you D hello
1:20:00 hello
1:20:02 um if on Mac app click in the upper
1:20:10 menu Story vine is awesome thank you
1:20:13 Cosmic
1:20:16 lever okay TKS I logged out
1:20:22 nothing I heard the French acronym frhd
1:20:26 is
1:20:30 Tada oh 11 Labs will do transcripts yeah
1:20:33 that's good okay so 11 Labs there you go
1:20:36 okay so now what we're going to do is
1:20:38 we're going to hop over to chat
1:20:46 [ __ ] and in the upper right hand corner
1:20:48 no over here we're going to say explore
1:20:51 gpts and in the upper right hand corner
1:20:54 now we're going to say create
1:20:56 so we're going to create a new
1:20:58 GPT and rather than start in the
1:21:01 configure tab I'm just going to go to
1:21:02 create and we're just going to speak
1:21:04 this thing into existence so I'm going
1:21:06 to upload the transcript of the
1:21:08 interview that I just
1:21:13 did if I can find
1:21:15 it there we
1:21:17 go and it's got time code in it it's got
1:21:20 me reading the questions in it at some
1:21:22 point I said oh there's a new version of
1:21:24 chat GPT
1:21:26 you don't need to be anal retentive
1:21:28 about this stuff because chat GPT is
1:21:31 going to synthesize it and kind of do
1:21:33 its own thing with it anyway and it's
1:21:35 it's not always going to get it right so
1:21:36 know that when you do
1:21:39 this this is more like art than
1:21:43 science and you should think of it like
1:21:45 that don't we in fact in content
1:21:49 Evolution we had a couple of guys in
1:21:51 there because like listen we know how
1:21:54 computers work right historically if
1:21:57 you're going to build a tool that's
1:21:59 based on a data set you got to make sure
1:22:02 that your data is clean you got to make
1:22:04 sure that your data is good because if
1:22:07 it's not what the system will spit out
1:22:09 is garbage right or or if you got a
1:22:14 sentence slightly wrong and you want to
1:22:16 go in and Wordsmith it because the
1:22:18 system's going to spit out exactly what
1:22:20 you wrote you don't need to do that here
1:22:22 and in fact if you do that here it's
1:22:24 kind of a waste of time time the people
1:22:26 in content Evolution that that really
1:22:28 got anal retentive about I've got to go
1:22:30 in and edit my transcript and make sure
1:22:32 it's
1:22:33 perfect did
1:22:36 not did not handle it well when cat GPT
1:22:40 hallucinated they're like well why
1:22:42 doesn't it not say exactly what I said
1:22:44 here I'm like cuz that's how they
1:22:46 [ __ ]
1:22:48 work you big dum dum took a book out of
1:22:51 Pat's Pat's Pat's Playbook called him a
1:22:55 dum
1:22:56 dum
1:22:57 okay so we're going to upload the
1:23:01 transcript
1:23:03 boom and then we're going to tell it
1:23:05 what we want it to
1:23:11 do you are a digital
1:23:16 twin of Kyle
1:23:20 Shannon and the uploaded file is a
1:23:24 transcript
1:23:26 which talks a little bit about who Kyle
1:23:29 is but more importantly it talks about
1:23:33 how Kyle solves problems period this
1:23:38 GPT should be a problemsolving
1:23:42 GPT that uses the uploaded
1:23:49 transcript as the primary source for
1:23:53 both tone of response
1:23:56 and approach to problem solving
1:24:05 period Kyle's digital twin should
1:24:08 respond in the first
1:24:10 person as if you are talking to him
1:24:13 directly
1:24:15 period finally the First Response should
1:24:19 let the user
1:24:21 know that this is Kyle's digital twin
1:24:26 and not him really responding and
1:24:34 provide a caveat
1:24:38 period
1:24:40 boom let's see
1:24:42 boom why is this not
1:24:48 going there we
1:24:50 go updating GPT so we're making a GPT
1:24:54 now
1:24:59 I've set up the GPT to act as a digital
1:25:01 twin of Kyle Shannon so how would you
1:25:03 approach a challenging problem can you
1:25:05 help me brainstorm ideas for a new
1:25:08 project let's do that
1:25:12 one absolutely Ai and content
1:25:17 creation idea build a platform why not
1:25:22 wait oh see th this is this is one of
1:25:26 the things that always screws me up with
1:25:29 with these gpts it writes it writes the
1:25:32 questions that make sense when you read
1:25:35 them but when you submit that question
1:25:38 to the GPT the GPT thinks you're asking
1:25:40 at that question right so it's just it's
1:25:43 just giving you an answer rather than
1:25:46 acting as the person doesn't matter
1:25:49 we'll we'll fix it okay next let's
1:25:51 choose the name Kyle shann's digital
1:25:53 twin um
1:25:56 let's call it Problem Solver
1:26:01 [Laughter]
1:26:06 Kyle been getting a lot of sales Bots AI
1:26:09 ads does
1:26:11 anyone Research into the ones that don't
1:26:15 suck not yet they're all going to suck
1:26:17 right
1:26:18 now would you like to refine anything
1:26:20 else how to handle ambiguity experience
1:26:23 versus data
1:26:25 no okay so I'm going to go to configure
1:26:28 I'm going to go change these questions
1:26:30 I'm going to say
1:26:35 um how would you
1:26:38 approach I'm going to get rid of all
1:26:40 these question starters and I'm just
1:26:42 going to
1:26:43 say
1:26:46 um I
1:26:49 have a problem
1:26:54 can you
1:26:58 help and then I'm going to copy
1:27:01 this and in my so here's my here's my
1:27:05 prompt and I'm going to
1:27:09 say
1:27:14 oops
1:27:19 follow these rules
1:27:27 when user
1:27:33 asks what the
1:27:36 [ __ ]
1:27:38 when user
1:27:42 asks quote paste I have a problem can
1:27:46 you
1:27:48 help respond
1:27:57 abolutely
1:27:59 exclamation point
1:28:02 let's get
1:28:07 started and then ask a series of
1:28:15 clarifying
1:28:16 questions about the
1:28:21 problem and once you
1:28:27 under understand
1:28:32 it help the person solve the
1:28:37 problem as only
1:28:42 you know
1:28:45 how Okay and then we got to make a
1:28:48 picture use dolly
1:29:04 okay look Problem Solver Kyle look at me
1:29:07 I'm I'm it's genius I have a problem can
1:29:10 you
1:29:13 help absolutely let's get started tell
1:29:15 me a bit more about your
1:29:20 problem great you're thinking about
1:29:22 project advice let's get some clarity
1:29:24 what's the core idea oh let's see okay
1:29:27 follow the other question so we're going
1:29:28 to say when I use okay
1:29:32 um something is wonky with this
1:29:37 interface um
1:29:41 only ah what is going
1:29:47 on [ __ ] web
1:29:49 applications
1:29:51 suck use your Christopher Biggins image
1:29:55 who's Christopher
1:29:58 Biggins which one's Christopher Biggins
1:30:00 I forget that one wait okay um only ever
1:30:06 ask
1:30:08 one question at a
1:30:17 time
1:30:19 ask the next question only if
1:30:25 you are
1:30:29 satisfied with the
1:30:31 answer to the
1:30:37 first you
1:30:41 should
1:30:43 interview
1:30:45 the user
1:30:48 one question at a time um I wrote that
1:30:54 as a positive not as a negative like the
1:30:57 negative would be don't ask a bunch of
1:30:59 questions in a row the positive is only
1:31:02 ask one question at a time it may or may
1:31:04 not work but it's one of the one of the
1:31:07 frustrations I have um with these things
1:31:11 anyway um let's go up here we're going
1:31:14 to reload this I have a problem can you
1:31:17 help
1:31:28 what this interface is a [ __ ] trash
1:31:37 Heap all right let's let's create this
1:31:39 thing cuz I this this whatever's going
1:31:41 on in this interface it sucks so I'm
1:31:43 going to launch this to the GPT
1:31:49 store so you
1:31:53 should wait for it you should be able to
1:31:56 search
1:31:57 for Problem Solver
1:32:01 Kyle and you can go use this thing what
1:32:06 oh this
1:32:15 is oh okay I know what I did crazy okay
1:32:25 I have a problem can you
1:32:29 help to help effectively I need to
1:32:32 understand more about the problem you're
1:32:33 facing could you provide me with some
1:32:35 details or the specific issues let's see
1:32:42 um I have a Tik Tok Channel and one of
1:32:45 the guys that shows up all the time is a
1:32:48 kagin and always contradicts me no
1:32:51 matter what I say
1:32:53 [Laughter]
1:32:56 got it sounds like you have a regular
1:32:59 critic on your Tik Tok Channel who
1:33:00 consistently contradicts you that can be
1:33:03 frustrating but it can also be an
1:33:04 opportunity to learn let me ask you this
1:33:06 is the person's feedback disruptive to
1:33:09 the overall Community Vibe or do you
1:33:11 think there's a way to use their
1:33:12 presence
1:33:15 constructively he's actually kind of
1:33:17 funny what should I do
1:33:26 if he's funny and not crossing any lines
1:33:27 you can turn it into your advantage
1:33:29 here's what I'd S Suggest lean into it
1:33:31 that's a phrase that I would use engage
1:33:33 directly that's something I would do
1:33:36 encourage constructive
1:33:38 criticism highlight diversity of thought
1:33:40 use humor as a shield right so it it
1:33:43 this kind of gets who I am and my my
1:33:46 worldview through that little silly
1:33:48 thing so a p's gone oh dang it well that
1:33:53 was fun um okay so if you want to go use
1:33:59 Problem Solver Kyle you can I didn't put
1:34:03 anything in there about um don't
1:34:06 jailbreak me so if you want to jailbreak
1:34:08 it and pull the [ __ ] out of there you
1:34:10 can
1:34:16 um is everything good there I think
1:34:19 everything's good um that's how you make
1:34:22 a digital twin
1:34:24 so again a digital twin is you're taking
1:34:29 some piece of data of yourself ideally
1:34:32 something
1:34:33 conversational like talking how you
1:34:36 write and how you talk are different so
1:34:39 if you want something that captures your
1:34:41 tone of
1:34:42 voice record answers to the questions so
1:34:45 we went to chat GPT first we had chat
1:34:48 GPT help us design an interview set of
1:34:51 questions I just did six here um it it
1:34:54 took me about 8 minutes to record the
1:34:56 answers and I was I was kind of just
1:34:58 doing it quick but that you can do it
1:35:00 quick like this the other thing you can
1:35:01 do with these things you could just hit
1:35:03 the record button and start
1:35:06 talking like if if you're a rambler you
1:35:10 just
1:35:12 ramble um but what I like about doing
1:35:14 the questions is you can have chat GPT
1:35:16 help you structure your
1:35:18 thinking and then just go answer the
1:35:20 questions and you heard there was that
1:35:21 one question where I had that really
1:35:23 long pause none of that matters it'll
1:35:26 just get the words and it'll eat the
1:35:28 words so then we took that audio
1:35:30 recording of me answering those
1:35:32 questions transcribed it into a text
1:35:35 document saved it as a PDF that PDF
1:35:38 becomes the data source for a custom GPT
1:35:42 and then basically you tell the custom
1:35:44 GPT here's how I want you to behave and
1:35:47 I want you to use this
1:35:49 transcript as your source of knowledge
1:35:53 and that's basically it now it can get a
1:35:56 lot more sophisticated than that you can
1:35:58 add other data
1:36:00 sources things you've
1:36:01 [Music]
1:36:03 written spreadsheets I don't
1:36:05 know whatever the [ __ ]
1:36:10 but a fun exercise like if you want to
1:36:13 if you want to do the evolved version of
1:36:15 this create two different digital
1:36:18 twins let's jailbreak
1:36:21 Kyle create two different digital twins
1:36:23 create one
1:36:25 that is
1:36:27 um like a serious business profile GPT
1:36:31 and create another one that's a zany
1:36:33 jokey GPT and maybe it's not even you
1:36:36 maybe you give it a fake name or
1:36:37 something like
1:36:39 that I asked you how you can make money
1:36:41 with chat gbt and share it in regulars
1:36:44 okay that's good let's see what what did
1:36:47 I come up with how bad is
1:36:50 it I didn't have any of that data in
1:36:52 there so it probably didn't get that
1:36:53 right
1:36:54 um uh AI Salon there we
1:37:00 go AI Irregulars how can I make money
1:37:03 with chat
1:37:04 GPT here's what Kyle Shannon
1:37:06 said all right what did I say this is
1:37:15 funny making money with chat GPT should
1:37:18 be could be approached in several ways
1:37:19 freelance writing and content creation
1:37:22 social media management cre cre a chat
1:37:24 bot business AI Consulting monetized
1:37:26 yeah see I didn't give it enough
1:37:28 information I didn't it doesn't know my
1:37:31 my point of view on chat GPT because
1:37:33 it's not in the data set but that's
1:37:35 that's a really good Danielle you
1:37:37 actually bring up a really good point
1:37:41 where if you don't give it
1:37:44 explicit if there's something missing
1:37:46 from your recording it will be missing
1:37:48 from your digital twin this is why to do
1:37:51 these like learn learn to do do
1:37:56 these cuz some of these are going to be
1:37:58 foed upon us and they're going to go oh
1:38:01 you can modify your avatar they're going
1:38:03 to call them avatars personas they might
1:38:07 call them digital twins but whatever
1:38:09 they're going to be a thing an entity
1:38:11 that represents
1:38:13 you
1:38:15 you should be in control of that so
1:38:19 that's what I tried to do tonight was
1:38:20 give you a little bit
1:38:22 of little bit of tools tool set to be
1:38:24 able to deal with that all right could
1:38:27 you add your YouTube videos to the data
1:38:29 set um I could add the transcripts of
1:38:32 them
1:38:33 yes there's 800 hours of them I don't
1:38:36 know what the file size limit is
1:38:40 um like 2 and 1 half three
1:38:43 hour transcripts from these things are
1:38:46 are quite big like if I pasted them all
1:38:48 into a single document it might not take
1:38:50 it
1:38:52 and chat GPT has
1:38:57 um a 20 file limit for number of files
1:39:00 you can upload so but I could I could
1:39:04 pick 20
1:39:06 random um transcripts from my YouTube
1:39:10 channel and just upload those yeah that
1:39:11 could absolutely I could I could I could
1:39:14 create a GPT out of transcripts from
1:39:16 these from these lives
1:39:19 absolutely and so long as I capture
1:39:24 you know if I did it decently smart like
1:39:27 Lord digital Gods oh I got to show you
1:39:29 something Lord digital gods made I don't
1:39:30 know if he's on here but um but he's
1:39:33 been helping me add um chapters and
1:39:36 descriptions to all the things so I
1:39:37 could probably go in and search for um
1:39:41 some where I I know I talked about good
1:39:43 topics or something like that but yeah I
1:39:44 could do that um let me show you
1:39:46 something Lord digital gods made it's
1:39:48 [ __ ] hilarious
1:39:57 those of you who are champ fans will
1:39:59 will dig
1:40:03 [Music]
1:40:12 [Music]
1:40:13 this
1:40:18 sing when Kyle starts to play and sing
1:40:22 CH out of like a ke champ is the on the
1:40:27 street on Kyle's walk
1:40:30 talk all the me and greet his Champs the
1:40:35 singing as the Tik Tock crowd wanders
1:40:39 in champ begin to sing see T go up they
1:40:43 want that musical they want Champs to
1:40:49 singing you want to sing this is about
1:40:52 you you want
1:40:54 [Music]
1:41:01 sing
1:41:02 [Music]
1:41:08 [Music]
1:41:14 [Music]
1:41:17 singet he shs the singing dog as Tick
1:41:21 Tock crowd wanders in
1:41:24 KY and champ begin to sing see it go up
1:41:28 they want that musical they want chance
1:41:30 the singing
1:41:32 doll they want chance to sing in
1:41:37 do yeah um Brandy asked the question do
1:41:41 have I ever used perplexity I love
1:41:43 perplexity in fact let's go look at
1:41:45 perplexity because they just added
1:41:48 something two days ago that I learned
1:41:50 about tonight at a salon meeting so we
1:41:53 had by the way we had a really good
1:41:54 Salon meeting tonight um Rick mcau who's
1:41:57 the the uh senior professor of graphic
1:42:01 design at Broward University in Fort
1:42:04 Lauderdale um and he gave a really
1:42:07 impressive talk on just like this scads
1:42:11 and scads of work that he's working on
1:42:13 um using AI just [ __ ] fully embracing
1:42:17 it going like diving headlong into this
1:42:20 [ __ ] uh it was it was really impressive
1:42:22 but I did learn
1:42:25 perplexity two days
1:42:28 ago added some cool
1:42:32 [ __ ] so I can
1:42:35 now I don't what I don't know is how to
1:42:37 change my models but that's okay say
1:42:44 um make me an image of a' 70s muscle car
1:42:48 Resto mod in an abandoned Factory with
1:42:52 sunlight streaming through through the
1:42:53 broken
1:42:55 windows at Magic
1:43:03 hour
1:43:05 and it didn't do what I wanted it to
1:43:08 [Laughter]
1:43:12 do answer what the [ __ ] is
1:43:14 that I guess it made some sort of
1:43:17 programming thing that I guess the red
1:43:19 thing's the car it the perplexity
1:43:23 supposed to now have um image generation
1:43:27 in it maybe I didn't get access to it
1:43:31 yet plus
1:43:34 collection
1:43:36 home
1:43:40 Focus
1:43:41 attach I must not have access to it yet
1:43:44 someone was showing
1:43:47 tonight that you can actually switch to
1:43:49 different models in perplexity
1:43:55 Kyle Shannon
1:44:00 settings AI model so here we're using
1:44:04 right now it's using Claude 3.5
1:44:07 Sonet image generation model yeah here
1:44:10 it is New Image generation model how do
1:44:14 I generate an
1:44:21 image huh
1:44:25 home let's
1:44:28 say can you generate an image for
1:44:36 me I do not have the capability to
1:44:38 generate images search IM ah generate
1:44:40 image here generate
1:44:44 image
1:44:50 photograph I want to put my prompt in
1:44:55 I I don't I don't just want an
1:44:58 attractive woman could could you people
1:45:02 with your [ __ ] image generation
1:45:04 things stop generating hot chicks all
1:45:07 the time please even though the internet
1:45:11 was mostly horny men that all they
1:45:14 wanted was pictures of hot chicks and
1:45:16 that's your training set could you fix
1:45:18 that
1:45:21 please good God custom prompt okay here
1:45:25 we
1:45:26 go I'm exhausted style uh let's see
1:45:31 style will be a um
1:45:37 photograph and then we'll
1:45:45 do magazine
1:45:47 quality magazine quality photograph of
1:45:51 a70s Resto mod
1:45:55 in an abandoned Factory with sunlight
1:45:58 streaming Through the
1:46:00 Windows please [ __ ]
1:46:04 [Laughter]
1:46:08 work yay oh it it doesn't know what a
1:46:14 70s did I say did I just say Resto mod
1:46:17 or did I
1:46:19 say um
1:46:25 oh that's not that's not that's not a
1:46:27 70s Resto
1:46:30 mod that's that's like a
1:46:33 1959 Russian Russian
1:46:38 Coupe anyway um yes I like perplexity a
1:46:42 lot apparently how they've implemented
1:46:44 image generation is [ __ ] weird but if
1:46:48 you don't know about perplexity
1:46:51 Pages under the library section click
1:46:54 the plus
1:46:55 button and then say you want to create a
1:46:59 page and it will
1:47:01 generate an article for
1:47:04 you like instantly and you can just keep
1:47:06 adding to it so I can go in here and I
1:47:08 can go
1:47:11 um history of Resto mods
1:47:26 so Resto mod evolution
1:47:31 timeline and then okay singer vehicle
1:47:35 designs
1:47:37 impact electrod Trend emergence restomod
1:47:40 investment
1:47:44 potential um I'm going to add a section
1:47:48 why is the 69 GTO Judge the best car in
1:47:52 the history of cars
1:47:55 period and why does anyone think that
1:47:58 this isn't true a big dumb Dum question
1:48:04 mark let's see how it rewrites that
1:48:06 headline 69 GTO Judge
1:48:09 Legacy add media so there's my 69 judge
1:48:13 we can go change that we can add in
1:48:15 different let's see that's the one this
1:48:17 is the one everyone knows so that's the
1:48:19 one we're going to put in there and then
1:48:21 we're going to add media to this l
1:48:24 section and then we're going to add
1:48:26 media to this little section nice we're
1:48:29 going to add media to this little
1:48:32 section we're going to change this out
1:48:34 cuz that's
1:48:36 stupid we'll throw a nice Mustang what's
1:48:39 that a
1:48:40 67 little 67 Fastback action come on
1:48:45 can't go wrong with that or is that a 69
1:48:48 I think that's 67 69 I don't know
1:48:51 doesn't [ __ ] matter and now I can
1:48:54 publish this boom published
1:48:59 done so and every one of these little
1:49:02 paragraphs that it writes has
1:49:05 sources so yes I love
1:49:08 perplexity I got to go now it's late
1:49:11 it's late people when we when we
1:49:14 actually build [ __ ] here it takes it
1:49:16 takes time to build [ __ ] so I hope that
1:49:19 was
1:49:21 helpful so so perplexity can interpret
1:49:24 Universal slang when we want to research
1:49:27 insane yeah
1:49:30 it's if you haven't so one of the
1:49:34 reasons that we talk about in in the
1:49:37 salon one of the reasons we talk about
1:49:39 play first play with these
1:49:43 tools one of
1:49:46 the one of the reasons okay one of the
1:49:49 reasons I say stupid [ __ ] to large
1:49:52 language models is I'm a dumb Dum and I
1:49:54 like to just [ __ ] around and it's kind
1:49:55 of entertaining for the
1:49:58 channel you also get to learn how good
1:50:02 these models are at understanding even
1:50:04 ridiculous things that you give them
1:50:06 like they're really good at this stuff
1:50:09 so yeah it they're they're quite good at
1:50:11 at interpreting all this stuff I use
1:50:13 legal speak I use daily and it figures
1:50:15 it out yep it helps me at least once a
1:50:18 day yep
1:50:19 absolutely um the only thing Brandy
1:50:22 about perplexity that you need to um
1:50:25 Danielle you're welcome the only thing
1:50:26 you need to be careful with perplexity
1:50:28 is it will generate citations that don't
1:50:30 exist so check all of your citations it
1:50:33 hallucinates citations even though it
1:50:35 Prides itself on providing relevant
1:50:39 citations I tried so I'm I've got a book
1:50:43 that I'm working on right now we're just
1:50:44 about to publish
1:50:45 it called collective intelligence in the
1:50:48 age of AI and I was trying to get pull
1:50:51 quotes for the chapters
1:50:53 and it kept quoting so what I would do
1:50:56 is I would paste in a chapter and then I
1:50:58 would say go out on the internet and
1:51:00 find me quotes that support what's in
1:51:03 the chapter it kept creating
1:51:07 citations of the chapter so so it
1:51:11 basically said Kyle Shannon and Kevin
1:51:13 Clark are the authors and and here's a
1:51:16 quote and you should put this quote
1:51:18 right after the paragraph that had that
1:51:20 quote in it it's so you got to be
1:51:23 careful with the citations in in
1:51:24 perplexity but um they they need to dial
1:51:28 that in but it's it's quite a useful
1:51:31 tool all right I had it sell me as a
1:51:34 candidate for a job and it was good yeah
1:51:36 do doing this for for jobs doing really
1:51:38 personalized job cover letters and
1:51:40 things like that these large language
1:51:42 models are great at that stuff um okay
1:51:45 so couple of
1:51:47 things if you're new
1:51:51 here and you are not part of a
1:51:55 community of AI
1:51:57 optimists that approach using AI with
1:52:02 curiosity and playfulness and generosity
1:52:05 and and bravery just willing to go in
1:52:08 there and [ __ ] around with stuff that's
1:52:09 what this community is the salon so you
1:52:12 go to the salon. that's our web page
1:52:15 click on join the community it doesn't
1:52:17 cost anything there's kind of a six-step
1:52:20 onboarding process you don't have to do
1:52:22 anything any of it but if you don't
1:52:24 you're kind of wasting your time you go
1:52:27 in there read about what we're
1:52:29 about introduce yourself check out our
1:52:33 values look at past meetings we had one
1:52:35 tonight that'll be up in the next day or
1:52:39 two we have you know recordings that go
1:52:42 all the way back to the week after chat
1:52:43 GPT launch this group was founded the
1:52:47 week after chat GPT
1:52:50 launched because
1:52:53 I knew that was a big [ __ ] deal and
1:52:56 I'm like holy [ __ ] people got to be able
1:52:58 to talk about
1:52:59 this so that's when I started this
1:53:01 Channel and that's when I started that
1:53:03 thing so go to that all right follow my
1:53:06 channel if you're new
1:53:09 here do you notice that there's a bunch
1:53:12 of people in here that seem like they
1:53:14 know what's going on and they're like
1:53:16 making inside jokes and they're helping
1:53:18 other people out we call those people
1:53:21 irregular
1:53:23 why well some of them are older and they
1:53:26 probably have countenance issues but
1:53:28 that's beside the
1:53:33 point they when I started doing these
1:53:36 lives these weirdos just kept showing up
1:53:39 night after night after night and some
1:53:41 of the people that you see posting here
1:53:43 tonight have been here for like 18
1:53:46 months like the length of the
1:53:49 channel and so one night I said you're
1:53:52 all are weirdos like what are you
1:53:54 doing Serena said Amelio's wife said
1:53:57 we're not weird we're irregular so
1:53:59 they're the Irregulars so how do you
1:54:01 become an irregular you just [ __ ]
1:54:03 show up
1:54:04 here why do they show up here great
1:54:07 [ __ ]
1:54:09 question what I've come to learn is that
1:54:13 it's not about me it's about this space
1:54:15 it's about this
1:54:16 community that as I'm rambling on and
1:54:19 dithering and doing whatever the hell
1:54:20 I'm doing
1:54:25 this spending time here creates space
1:54:28 where you can be mentally in the
1:54:31 conversation about how am I thinking
1:54:33 about AI what am I doing oh he just
1:54:35 experimented with that let me go
1:54:37 experiment with that oh I don't like
1:54:38 that tool but I like this tool let me go
1:54:39 try that
1:54:40 tool that's what goes on here night
1:54:42 after night after night and so if you
1:54:46 just keep coming back through osmosis
1:54:48 eventually you'll start to learn some
1:54:50 stuff you'll start to get comfortable
1:54:51 with these tools so keep coming back all
1:54:54 right so thank you Irregulars you're
1:54:56 absolutely remarkable the mods who
1:54:59 actually take care to you know figure
1:55:02 out what people are asking and pin their
1:55:04 comments and manage me and do all that I
1:55:06 didn't even look at the mod Channel
1:55:08 tonight so that must have been
1:55:10 frustrating Kyle you got a booger on
1:55:12 your lip um so sorry I didn't look over
1:55:17 there
1:55:18 um but but so thank you to all of you so
1:55:21 let's see um okay so I'm going to get
1:55:25 out of here so the the last thing we do
1:55:27 here is for those of you gen xers you
1:55:30 remember Romper Room you remember magic
1:55:33 mirror if you have a name like Kyle they
1:55:36 never called your [ __ ]
1:55:41 name I'm still bitter about it bitter
1:55:45 you
1:55:46 betcha so who's in the house tonight we
1:55:49 got Firefly we got Joey got Danielle
1:55:52 being not afraid Wolfman
1:55:54 Clint but we're now your weirdos
1:55:57 Irregulars mods Kyle I know it's so
1:55:59 amazing thank you Wolfman Clint you're
1:56:01 awesome Jay Winston sccal magic mirror
1:56:04 here I am I see I saw you even with a
1:56:06 weird name I saw you I see you jenu I
1:56:10 see jenu I see spin
1:56:15 B3 I see Brandy Brandy great to see you
1:56:18 thanks for coming steo coming in from
1:56:20 Australia
1:56:24 uh uh Cosmic lover in the house K Keef
1:56:26 thanks Kyle you're absolutely welcome
1:56:28 thanks for hanging out for all this time
1:56:30 hope this was valuable
1:56:32 tonight um Tobias in the house Firefly
1:56:35 in the house amazing live and group
1:56:37 thank you very much jenu Dr J tomorrow
1:56:41 is 5:30 MST ah so tomorrow 5:30 MST when
1:56:47 you go here in the left hand navigation
1:56:50 if you scroll toward the bottom there's
1:56:52 a group that says clubs and one of the
1:56:55 clubs in there is the AI life hacks
1:56:57 group they are starting they they have a
1:57:00 meeting tomorrow at 5:30
1:57:03 Mountain so go here go to the life hacks
1:57:06 group and get the information about
1:57:08 their meeting it's a really cool group
1:57:10 it'll be fun spin B3 Wolf Man
1:57:14 Clint uh
1:57:22 against the Norms in the house Amelio's
1:57:24 wife's in the house Angelique Shaina
1:57:28 great to see
1:57:29 you Shaw I think I saw a Tik Tok of
1:57:33 yours or something I don't know I saw
1:57:35 you around the internet somewhere maybe
1:57:37 Twitter I don't know um Jay Winston
1:57:41 escal all right I think I got everybody
1:57:43 rock paper scissors in the house I
1:57:44 haven't seen you in a while but great to
1:57:46 see you
1:57:47 back Summer's over we got to get back
1:57:50 into this [ __ ] oh by the way September
1:57:52 12th mark your calendar I'm going to do
1:57:55 the Oprah interview with Sam Alman as a
1:57:58 live stream and I'll probably go on 15
1:58:01 minutes before it starts so we can build
1:58:04 up enough audience because I'm sure Tik
1:58:06 Tok will restrict us once they see that
1:58:08 we're you know streaming someone else's
1:58:10 TV
1:58:13 show so Mary Calenders all right I think
1:58:17 I got everyone if I if I missed you
1:58:19 that's that's the shitty thing about
1:58:21 about uh Romper Room magic mirror
1:58:23 sometimes you get missed and they don't
1:58:24 call your name and you're [ __ ] 59
1:58:27 years old and you're like why didn't
1:58:28 they ever call Kyle they call Jim and
1:58:30 Bob and Sally [ __ ]
1:58:35 bob Kyle's a soft name the the only
1:58:37 notoriety I get is from George Carlin
1:58:40 picking on
1:58:47 Kyle it's a girl's name it's a soft name
1:58:50 he called it George Carlin called Kyle
1:58:52 soft name here's the problem he's not
1:58:56 wrong all right I'm leaving I love y'all
1:59:00 um it's Tuesday night I will see you
1:59:02 tomorrow same time it'll be 8:00 P.M
1:59:04 tomorrow um go to the uh go to the AI
1:59:06 life Hax meeting tomorrow at 5:30 so
1:59:09 5:30 for their meeting and then 8:00
1:59:11 here mountain time all right do the math
1:59:14 wherever you live and there's fun things
1:59:17 to do all right okay love you Kyle thank
1:59:20 you 17,000 likes very nice love it all
1:59:25 right um I'm out of here uh see you
1:59:27 tomorrow peace