Jan 16, 2024 - Insights from the AI Learning Lab Experience

Description

Chapters

Transcript

0:00 [Music]
0:13 [Applause]
0:14 she came on him like
0:19 [Music]
0:22 slow
0:25 his look
0:27 [Music]
0:29 her
0:31 she sat
0:32 onol and what do you
0:36 [Music]
0:39 want she said give me a love that don't
0:43 fre up
0:48 inside said
0:50 IED some hearts in my time
0:56 de but to sit next to you well I shiver
1:01 and
1:05 Shake if I knew love well I don't think
1:09 I'd be
1:11 [Music]
1:14 here asking
1:16 myself got it
1:19 [Music]
1:20 [Applause]
1:23 te do melt your iy blue
1:29 ho
1:31 [Music]
1:35 has turn was been frozen
1:42 fall into a
1:48 [Music]
1:54 river these days we all play cool
1:59 Calle
2:01 I think he's getting tired of
2:04 [Music]
2:29 it
2:31 [Music]
2:46 that's champ Shannon everybody champ
2:48 Shannon on lead
2:51 vocals I'm singing a sad sad harmonies
2:55 and backup here welcome welcome welcome
2:59 welcome to AI learning l
3:01 just got out of a uh fantastic AI Salon
3:05 we had like 80 people there
3:07 tonight had a fantastic speaker Tim came
3:10 talked about his photography was just
3:13 awesome what's your view on open air
3:16 loosening up their stance on open AI
3:19 loose loosening their stance on
3:21 Military
3:23 usage is profit the focus now well I
3:27 don't I don't know I don't I don't know
3:29 anything about their stance on Military
3:31 usage I mean I think I saw a Twitter
3:34 post about it that was like a week or
3:35 two
3:36 ago
3:38 um I do not know I don't I don't really
3:40 have a point of view on it right now um
3:43 they're a business
3:46 and running these large language models
3:49 so so so as magic as as absolutely
3:54 [ __ ] magical as they are they
3:56 actually take a tremendous amount of
3:59 energy and compute power to run
4:03 therefore a lot of
4:05 cost um so if open AI having to do stuff
4:08 to survive to get to AGI I assume
4:11 they're going to do what they need to do
4:13 I mean listen if you've ever run a
4:15 business there's something there's
4:16 something that that that makes me a
4:18 little cranky about this poly Anish
4:21 thing like you know businesses should
4:24 just only ever do good for the world and
4:26 and you know never turn a profit and
4:29 never survive like there's a [ __ ]
4:31 reality to running a business if if
4:33 you've ever run a
4:35 business there's things like making
4:38 payroll it you know what it is it's
4:40 these [ __ ] snowflake gen zers they
4:42 expect to get a paycheck every
4:45 week every week they're like they expect
4:48 you get a check well no a a depos a in
4:52 their venmo account they expect cash to
4:55 be deposited in their venmo account like
4:58 so that they can go buy Starbucks coffee
5:00 they expect it every
5:02 week if the Gen seers would just give up
5:05 the whole capitalist pick thing and not
5:08 accept the money then these companies
5:10 could just do
5:14 [ __ ] I mean it's like I don't I don't
5:17 know what you expect companies to
5:20 do you know open AI is
5:23 arguably re-architecting how we're going
5:26 to work in the future and listen I get
5:29 it we don't want them to do evil we
5:31 don't want the robots to kill us and you
5:33 know doing stuff with military is is a
5:36 is a slippery slope and all that [ __ ]
5:38 but at the same time they got to [ __ ]
5:40 survive they're business they are a
5:43 business
5:47 first now they're managed by a nonprofit
5:50 We saw how well that went when they
5:52 tried to fire Sam
5:53 Alman that was a [ __ ] clown
5:57 show so
6:01 I don't know you know they probably got
6:03 some business people in there going uh
6:05 yeah we're going to run this like a
6:06 business now yeah about that so I don't
6:09 know I don't know I don't know right now
6:13 I'm just using chat GPT I don't want to
6:15 keep using chat GPT and I want it to be
6:17 20 bucks a month I've I've said this
6:20 many times I think the 20 bucks a month
6:22 for chat GPT right now is the biggest
6:24 bargain in software
6:26 history what what chat gp4 does for 20
6:30 bucks a month is [ __ ]
6:34 staggering they could easily charge 20
6:37 bucks a month per feature in chat
6:40 gb4
6:42 easily so the fact that we get to play
6:45 with that [ __ ] for for [ __ ] pennies a
6:50 day for less than a dollar a day you can
6:53 change the word with chat
6:55 JP I had avocado toast tonight can I do
6:59 that as a gen xer cuno you're [ __ ]
7:03 you're skating on thin ice here you're
7:06 going to be drinking from the hose for
7:07 the rest of the week avocado toast if it
7:11 had bacon on it of course absolutely
7:14 acceptable if it had like Pika Deo and
7:18 and some
7:19 [ __ ] toasted sesame seeds no not
7:23 acceptable not
7:25 acceptable are you still using Google to
7:28 search for things I haven't used Google
7:31 I got to tell you oh Be not Afraid thank
7:32 you so much appreciate that
7:35 um every time I use Google or Bing and
7:39 because I do it in here sometimes people
7:40 are like oh you know what's this thing
7:42 and I'll go to Google and try to use it
7:44 it's
7:59 and then they had to monetize it back to
8:01 the those pesky employees expect their
8:04 paychecks
8:05 thing and they they they
8:08 over you
8:11 know they they effectively got a
8:14 monopoly which forced every Advertiser
8:17 to do SEO exactly how Google wanted it
8:20 done and so now the entire internet is
8:22 this homogeneous SEO injected pile of
8:25 [ __ ] where you can't find anything
8:28 because everyone's trying to Ram you
8:29 down their sales funnel and make it look
8:32 like a free pdf download site which is
8:35 mostly what I search for how do I find
8:38 PDFs of this particular flavor and I got
8:40 to sift through 40 [ __ ] sites to get
8:42 there I am so done with search
8:45 engines remember when we used to have to
8:48 pay $500 for Windows yeah exactly I mean
8:52 anybody using Photoshop or Office 365
8:55 that [ __ ] ain't cheap I think we're
8:57 paying like I've got a small company and
9:00 I think we're paying 600 or 700 bucks a
9:02 month per Adobe Creative Suite license
9:06 that's [ __ ]
9:09 expensive and you get all of what's in
9:12 chat gp4 for 20
9:14 bucks put it in [ __ ] perspective it's
9:17 like
9:21 seriously cuz that's all we had kyl we
9:23 liked it we loved it in my
9:28 day in my day he i' I'd have my computer
9:32 call up another computer and I'd watch
9:34 the characters type across the
9:37 screen we had a 300 bod modem nobody
9:40 knows what a bod
9:45 is tapping foot looking at watch I will
9:49 not be bullied Emilio's wife absolutely
9:52 not if I want to say a 45 minute hello
9:55 like I did last night listen last night
9:57 was an anomaly we did a we did a a three
10:00 we I talked for three hours and 23
10:03 minutes it was off it like someone
10:06 needed to put me
10:11 down I'm like I'll just show you chat
10:13 GPT three and a half hours later I'm
10:16 still like maybe I should go
10:20 [Laughter]
10:24 now or were you yelling at me CU I was
10:27 here late I had a salon tonight 45
10:30 minute
10:34 hos oh man that was a good Salon tonight
10:38 so if you don't know what the AI Salon
10:39 is you're like what's this Salon thing
10:41 you're talking about I've never heard of
10:42 a salon the only Salon I go to is where
10:45 I got my hair done I do you ever get a
10:48 let me can I ask you a question as a man
10:51 um do you ever get a man manicure Manny
10:54 Petty manicure or pedicure have you ever
10:55 you ever done that oh you haven't oh you
10:58 do not know what you're missing thing
10:59 it's
11:00 fantastic really Fant what oh it's not
11:04 that kind of Salon what I'm sorry is
11:05 there another kind of
11:07 Salon what
11:09 Aon I I do not what
11:12 Aon oh I haven't heard that
11:17 oh sounds very fancy what do you do
11:20 there you you just
11:24 talk like at a salon at a at a
11:28 salon
11:32 oh Emilio's wife no I want the hellos
11:34 and goodbyes you're late Mr Shannon I
11:37 actually got here half an hour from what
11:39 I had in my head that I was going to get
11:41 here I figured I'd start at 9:00 so I
11:43 started half an hour
11:44 early so I'm early in my mind but I
11:47 didn't do a little timer thing to let
11:48 people know I was here so anyway welcome
11:51 everybody if you are new here my name is
11:54 Kyle Shannon this is the AI learning lab
11:58 and
12:01 that the AI Salon if you go to the
12:04 salon. oh excuse me the salon. a by the
12:08 way if you type in
12:10 HTTPS col that it won't work CU I set up
12:14 the domain wrong so just type in thes
12:16 salon. all lowercase and it'll work so
12:19 go there that'll redirect you to a link
12:21 tree and the first link on that link
12:24 tree goes to the Community website of
12:27 the AI salon and it's super cool it's a
12:30 really cool Community we just had our
12:32 meeting tonight we have a meeting every
12:33 other
12:34 week we're going to start doing local
12:37 meetups
12:39 um and you know like happy hours and
12:41 things like that so I'm really excited
12:43 about that and we have speakers come in
12:45 we had a a professional photographer
12:47 come in tonight that has a really cool
12:50 uh a really
12:52 cool what's it
12:55 called
12:57 uh AI project
13:02 he doesn't seem very bright he was
13:05 trying to remember words it was not
13:07 going
13:10 [Music]
13:15 well you're forgiven we're all here now
13:18 very nice very good car on cherio pip
13:20 pip
13:21 rather Tim
13:25 Aral Tim archal is a
13:32 is
13:34 a rather well-known
13:37 photographer and he came tonight and he
13:40 spoke about this project that he created
13:42 called the AI Camera Club so if you go
13:45 to Instagram to theore aior Camera
13:50 Club um steo where is your local Meetup
13:54 you'll have to start it dude we have we
13:57 haven't we haven't rolled them out yet
13:58 we're we're figuring out when they're
14:00 going to be and what they're going to be
14:01 and how local people can do them and
14:05 we're probably going to make
14:10 stickers um anyway um it's a really cool
14:14 project it basically what he did was he
14:16 sort of
14:18 imagined the he what this project is is
14:22 um this photography club from the'
14:26 40s that uh started in AI photography
14:30 club in the 40s so it's this it's this
14:32 whole fictitious club and like the
14:35 pictures that they took in this club
14:37 when they were inventing AI Photography
14:39 in the' 40s it's such a cool [ __ ]
14:41 premise like I the the thing that I'm
14:43 liking the most about artists that are
14:46 embracing AI right this is a
14:48 photographer like he's been a
14:49 photographer since he was
14:52 14 he went from film to
14:55 digital he skipped a couple of the
14:57 trendy things and then he actually he he
14:59 said he started this project he went
15:00 into this really pissed off like oh this
15:03 going to [ __ ] suck and this is stupid
15:05 I'm just going to put stuff up and make
15:06 it look stupid and then he got he got
15:08 sucked in like everyone every single
15:11 person that I know that that actually
15:13 dives into this AI stuff what they
15:16 discover is a whole new kind of
15:18 self-expression and he said what ended
15:20 up happening to him
15:23 was he was sort of getting images he
15:26 liked and then he tried a couple of
15:27 different things and then all all of a
15:29 sudden he got one
15:31 photograph that he said oh that's a
15:35 photograph like I would take and the
15:38 minute he did that it hit him that this
15:42 is a new form of
15:44 self-expression and so he leaned into
15:46 this project and so it's this whole
15:48 world of these this photography club
15:50 from the
15:52 40s the AI Camera Club on
15:55 Instagram Tim archal is his name so he
15:57 spoke at the salon tonight
16:00 tonight and uh it was really really good
16:03 do we get to pick our own local AI zoo
16:06 animal for the
16:08 mascot we're going to need a bigger Arc
16:11 for all these [ __ ] animals Kyle I
16:14 forgot to tell you Adobe Creative Cloud
16:15 is holding a free event in my town oh
16:17 nice very cool Thursday awesome yeah go
16:21 there go do that um I have some
16:24 interesting news about Adobe I can't
16:25 really share it right now it's not big
16:27 news but it's just some
16:29 some some just interesting interesting
16:32 stuff going on which I'm excited
16:37 about um anyway that's that so if you
16:41 haven't been to the AI Salon
16:44 website the community this is it so when
16:48 you go
16:49 here go to the about the salon Tab and
16:53 then check out who we are and why we do
16:55 what we do and then check out our values
16:58 we went over these tonight Mr K made a
17:00 video today talking about the AI salon
17:02 and he went through these values as well
17:04 these values are super important this is
17:06 this is here's here's what you do with
17:08 the with the AI Salon come here and read
17:11 about who we are and why we do what we
17:14 do and then read the values and here's
17:16 here's your litmus test if when you read
17:19 these values you're like yeah that makes
17:21 sense these are your people you found
17:24 your tribe if in reading these values
17:27 you're like who are these [ __ ]
17:28 weirdos
17:29 this is not your tribe it's totally cool
17:32 but it's really important so as we
17:35 grow this community I want to make sure
17:38 that we
17:39 maintain the spirit of of who we are and
17:43 by the way the there's no charge for the
17:45 AI Salon it's
17:46 free um Ann Murphy is January's sponsor
17:51 so she's sponsoring we have sponsor we
17:53 have sponsors cover the infrastructure
17:56 costs cuz Mighty networks is not
17:59 cheap um and so Ann Murphy from
18:02 empowered fundraiser is our sponsor this
18:05 this month so anyway our values are
18:08 curiosity and
18:09 generosity and exploration and
18:13 collaboration and inclusion and empathy
18:16 you should go read you know read what we
18:18 wrote about them and if you resonate
18:20 with those this is going to be your
18:22 people and then what you should do is
18:25 slide on down here to introduce
18:27 yourself and and introduce
18:31 yourself but Kyle I'm shy and I've never
18:34 done AI before and I'm too new to this
18:37 stuff and I'm not very technical and I'm
18:39 afraid people will judge
18:42 me they may but they won't hear cuz this
18:47 group's really cool go read the values
18:49 and understand that this is an amazing
18:53 community and then just say hi tell us
18:55 who you are tell us what you
18:57 do tell us if you're just starting
19:01 or you're the only one in your company
19:03 that's into Ai and everyone else thinks
19:05 you're a [ __ ] weirdo that's us that's
19:08 us too it's a support group it's like a
19:10 12-step program for AI
19:13 enthusiasts hi my name's Kyle I'm into
19:17 AI hi
19:19 Kyle it's been a few minutes since I
19:21 made an image with Dolly yeah yeah yeah
19:24 I get you I feel you how much does it
19:27 cost to be a sponsor for a month
19:29 $150 I think it's the it's the cheapest
19:32 sponsorship of a group this strong this
19:34 side of the
19:36 payos um we're we're really just
19:38 charging enough to cover the cost of the
19:39 infrastructure so um so yeah and then
19:43 the other thing is if you go
19:47 to if you go to that URL I gave you the
19:56 salon. it's going to take you to um a
20:00 link tree so the first one is the
20:02 community and then there's a couple of
20:04 links for the GPT for good event that we
20:06 did over a couple of weekends ago and
20:09 then there's a a merch store and so one
20:12 way you can support the salon is you can
20:14 buy a monthly sponsorship and the other
20:16 way is you can if you buy merch at the
20:18 store that supports the salon so that's
20:20 how we're paying for the infrastructure
20:22 that's why it's free but it it was
20:24 really important to us we wanted to make
20:26 sure that we could
20:30 provide a space for people that were
20:33 trying to figure this stuff out to just
20:36 go be themselves and that's what this is
20:38 there's a a water cooler discussion area
20:40 there's a showand tell area which if
20:42 you're in here tonight generally what
20:44 happens is oh nice we're going to need
20:47 bigger animals for these animals that's
20:49 good very cute uh as I'm talking uh and
20:53 making fun wait what is this AI sausages
20:57 maybe I did something maybe it's Kyle
20:59 maybe I did wrong what's
21:02 happening where is
21:06 Kyle people in here are funny they're
21:08 funny they're funny I like it um you can
21:12 just be sharing images you can be
21:14 sharing gpts the AI minor lab nice very
21:20 nice Kyle is
21:27 late uh um yeah so anyway that's the AI
21:30 Salon the other thing in the AI
21:32 Salon Community is this page I think
21:35 this page is worth its weight and
21:37 gold it's digital Kyle it doesn't weigh
21:40 anything that's not the point it's a
21:44 metaphor you can make money
21:46 with this is
21:49 gold
21:51 um this is a this is a list that I I
21:55 maintain I try to M I try to keep my
21:57 keep this up to date but this is a big
22:00 ass list now it used to be small this
22:02 used to have three sites on it and and
22:05 it was just like here's the three sites
22:07 I think you should be paying attention
22:08 to and now
22:10 it's if you want to feel AI literate go
22:13 learn all this [ __ ] but here's what I
22:14 can promise you if if there's a site on
22:16 this list it's worth your time to learn
22:18 it with the exception of grock which
22:21 nobody needs to learn grock until Elon
22:24 Musk shut shuts down that company or
22:27 makes it better
22:29 but it's on there cuz it's Elon
22:34 Musk but yeah you should go look at that
22:37 anyway all right so here's how this
22:40 works here at the AI learning
22:43 lab you ask questions about AI or you
22:47 say hey I want to see something or I
22:49 want to understand something and then I
22:53 will make up the
22:55 answer and because you don't know
22:57 anything you you won't know any
23:04 better I wish I I wish that weren't
23:08 true but ask ask the Irregulars the
23:11 Irregulars are the people that show up
23:13 here every night they have been trying
23:15 to figure out what the [ __ ] I've been
23:16 talking about for six months they keep
23:19 coming back it's not cuz they like it
23:21 here it's like eventually this is going
23:23 to make sense
23:25 eventually we I may I could possibly
23:28 learn something
23:30 here it's just like the ultimate bait
23:32 and switch you know you come here for
23:34 the gold you can make money time and you
23:37 get a SM me X actor and you're like why
23:40 am I here that that is for you listen we
23:44 all make questionable Life Choices
23:47 you've made yours by hanging out here
23:53 tonight thanks for building that one
23:55 page in the AI Salon versus the keynote
23:57 doc I was tired of iPhone iPhone
23:59 screenshots I know I know it's so much
24:03 better it's it's amazing what a little
24:06 infrastructure um does for a a joint
24:09 we're listening Tom Arnold perfect
24:11 response please explain what the meetups
24:14 are about
24:17 um so we generally have so the AI Salon
24:20 meetups are um they're every other
24:24 week and we generally have a speaker so
24:27 we'll have a speaker come talk about a
24:29 project that they're working on or when
24:31 when Ann Murphy and I did the GPT for
24:33 good thing um she came and she talked
24:36 about what we did on that
24:38 weekend
24:40 um we talk about the community and you
24:44 know how we want it to be and then and
24:46 then we generally have some interaction
24:48 um with the folks on the phone like
24:51 tonight this was fun we did it took us a
24:54 half an hour cuz there were 80 people on
24:56 the call it was a pretty big call
24:57 tonight
24:59 um we did um what are what's what's one
25:04 word hope for 20124 what's one word
25:08 that's a fear of 2024 and what's your
25:11 prediction for
25:12 2024 and so Vicki Baptist took notes and
25:16 we went through like everyone that was
25:17 on the call they said here's my hope
25:19 here's my fear here's my prediction for
25:21 2024 so we're going to turn the hopes
25:23 and fears into word clouds and then
25:25 we're going to um we're going to take
25:28 the predictions and we're going to put
25:31 them in some semblance of an order and
25:32 by we I mean chat
25:34 GPT um we're going to put them in some
25:36 semblance of an order and then we're
25:38 going to
25:40 um put them somewhere on the AI salon
25:43 and we'll check in next year and see how
25:44 we
25:46 did how do we join to get notified for
25:48 the meetings so just go
25:51 to I
25:56 don't when you go to the AI Salon I
25:59 don't know about the notifications
26:05 but
26:07 um of all the meetings here so you
26:10 basically go subscribe to them what what
26:13 happens if so if you're in the AI salon
26:15 and one of these meetings goes live
26:18 you'll get a a live symbol here so a lot
26:23 of times what'll happen is so we've one
26:25 of the things we've got in the salon is
26:27 we've got guilds and so when the guilds
26:29 have meetings there there'll be a little
26:31 sort of flashing red light here that
26:33 says they've got a live meeting going on
26:35 so one way is you can just come in here
26:36 and see see when things are live but you
26:38 can also go subscribe and add these uh
26:41 events to your
26:42 calendar so that's how you do that and I
26:46 don't know about the notification things
26:47 I I would say check with Mighty networks
26:50 like do do a Google search good
26:53 luck ask P how how Mighty networks Works
26:57 uh cuz I don't know cuz you know I just
27:01 I'm like the MC of the joint I don't
27:03 know how this [ __ ] works that was a cool
27:06 image I made tonight that's a uh what's
27:08 it called DOL DOL image I wish I could
27:12 have got at China to come to the Meetup
27:14 tonight the voice of Siri oh that would
27:16 have been
27:20 [Music]
27:25 nice watch Daniel's and video she made
27:28 of her upand cominging business okay
27:31 very cool where is that is that in the
27:32 show and
27:35 [Music]
27:42 tell we're going to need a bigger
27:46 Arc the AI Zoo
27:53 transport yall are so [ __ ] weird I
27:57 like it it's just the way you're
27:59 supposed to be just left a
28:04 [Laughter]
28:09 center here let me go look for
28:20 Daniel where Is It joker where's the
28:23 video I'll go look at
28:26 it
28:28 with diff disinformation and Bad actors
28:31 a foot people's Home Alone M moment
28:33 could be negative it absolutely could
28:36 listen so so two block
28:38 Tom I agree with you that um we are in
28:43 for a pretty dramatic [ __ ] show of the
28:48 2024
28:50 election
28:52 um and so I I agree I think that I think
28:56 that having the Kevin mallister moment
28:58 which if you don't know what the Kevin
28:59 mallister moment is um if you remember
29:02 home alone it's it's this
29:05 moment and when you start playing around
29:07 with this AI stuff there will be some
29:09 point usually within the first 5 or 10
29:11 minutes where you have that moment and
29:12 you're like holy [ __ ] what did that
29:14 thing just
29:16 do and then those things keep coming
29:19 just one after another after like the
29:21 deeper you dig into this the more you
29:23 just you it's it's essentially
29:25 impossible to process what's happening
29:28 and I think 2024 as these tools get more
29:30 sophisticated we're already seeing it
29:32 with images but we're seeing the videos
29:34 get better and better we're seeing the
29:35 audio get better and better we're seeing
29:37 the synthesized audio sound more and
29:39 more
29:41 lifelike and
29:43 nuanced and so probably within
29:47 2024 it's going to be almost impossible
29:50 to distinguish real from
29:52 fake and yeah some people's Kevin
29:55 McAllister moments might be this shit's
29:58 evil and I guess I suppose that's why I
30:01 spend a lot of time here trying to get
30:03 people to step into this on purpose
30:08 right consciously choose to be a part of
30:10 learning about this rather than smacked
30:13 up side the head with a
30:15 tortilla um I just posted it in the
30:19 Irregulars so you could find it easily
30:21 very cool so uh
30:25 Danielle be not afraid
30:30 use this tool called
30:33 invid so is this your video or did you
30:36 post
30:43 it fold Rush laundry reclaim your time
30:47 with expert laundry service ever wished
30:49 you had more time in your day Picture
30:51 This a life free from the mundane chore
30:53 of laundry enter fold Rush laundry your
30:56 personal laundry assistant that turns
30:58 your laundry day into a day of leisure
31:00 imagine a busy mom juggling work kids
31:03 and
31:05 home fold Rush laundry has been a game
31:07 changer she says now I can spend more
31:09 quality time with my kids instead of
31:11 dealing with dirty clothes now think of
31:13 a college student swamped with
31:15 assignments and exams who says fold Rush
31:17 laundry is my secret weapon it gives me
31:20 more time to focus on my studies and I
31:22 don't have to worry about laundry
31:24 consider a professional working long
31:25 hours who shares full Rush laundry is
31:28 reliable and convenient it saves me
31:30 precious time and the quality of their
31:32 service is topnotch fold Rush laundry
31:36 isn't just about washing and folding
31:38 your clothes it's about giving you back
31:40 your time it's about convenience at your
31:42 fingertips with easy scheduling and
31:44 prompt delivery it's about quality with
31:47 every item treated with personalized
31:49 care and yes it's about the planet too
31:51 with eco-friendly practices that you can
31:53 feel good about so why spend your
31:55 weekends doing laundry when fold Rush
31:57 laundry can do it for you say goodbye to
31:59 the laundry burden and hello to more
32:01 freedom more peace of mind and more time
32:03 for what truly matters give fold Rush
32:06 laundry a try for your next laundry day
32:09 experience the convenience reliability
32:11 and quality that our diverse customers
32:13 Rave about turn your laundry day into a
32:16 day of leisure with fold Rush laundry
32:18 because you deserve it that's awesome
32:22 yeah it's in video and you can you know
32:24 she can she can pay for that and
32:25 download it without the watermarks and
32:27 things like like that but this is this
32:29 is one of those tools where you just put
32:31 in here's my business it writes the
32:33 script it creates the voice over it
32:35 finds all the stock footage it edits it
32:38 all all together it puts the background
32:39 music in there creates the
32:42 titles well that you having to do a
32:45 thing there's so in video excuse me in
32:49 video is one of them and then there's
32:51 one that
32:53 I I have in the list called Visa which
32:56 is very Sim similar to
32:58 this super
33:01 cool super cool it creates all the
33:03 dialogue and uses stock video all from a
33:06 prompt yeah
33:08 amazing and that's I mean this is this
33:11 is the
33:13 the the the thing that one one of the
33:16 predictions um or yeah one of the
33:19 predictions and some of the hopes
33:21 tonight we're just ubiquitous was the
33:25 word right that that this AI stuff is
33:27 going to be ubiquitous it's going to be
33:29 everywhere Be not Afraid took less than
33:31 5 minutes yeah she put that thing
33:34 together in less than 5 minutes and now
33:36 and what's amazing about this is you can
33:38 start with that and if you wanted to go
33:39 edit the script you could go edit the
33:40 script if you wanted to cut in different
33:42 video you could do that you wanted to
33:44 swap something out for a different piece
33:45 you can do
33:49 that
33:51 minutes things that used to take days or
33:54 weeks and cost thousands of dollars are
33:56 now gonna take minutes or
33:59 seconds and it's going to be everywhere
34:02 it's coming it's coming it's coming it's
34:03 here it's
34:05 here it's here
34:09 people it's here it's here what are you
34:13 doing what are you doing have you
34:15 launched chat GPT yet seriously anyone
34:18 that's in here there's 111 people in
34:20 here have you tried chat
34:26 GPT I'm I'll
34:29 wait three of you said no it's probably
34:32 more than that by the
34:34 way I've had people that have come here
34:37 multiple
34:38 times and then like weeks after I see
34:41 them in here multiple times they're like
34:43 I finally tried chat GPT you were right
34:46 it's kind of
34:48 something [ __ ] try it right now
34:51 chat doop
34:54 ai.com just go there now if you haven't
34:57 played with it go there right now but
34:59 how what do I
35:01 do go to your browser what's a browser
35:05 Chrome oh Chrome yeah I've got Chrome
35:07 it's not a is that a browser go to
35:12 Chrome in the little bar at the top type
35:15 chat. open.com just go do
35:19 it but how what do I
35:22 do you do
35:26 this you do parlor tricks to start kle
35:31 what's a parlor trick I don't first you
35:33 mention the AI salon and it's not it's a
35:35 different meaning of the word than the
35:38 beauty parlor and I didn't even know
35:40 that and now you got parlor tricks what
35:42 is this the 1890s Kyle I don't
35:45 understand so you go here I'm at chat
35:48 GPT
35:49 3.5 okay that's the free
35:52 version and then I'm going to go uh
35:55 let's see let's see
35:58 uh uh let's
36:00 see write me an
36:04 article
36:08 about the Ai No that's not going to show
36:12 up
36:15 about
36:16 why some
36:19 dogs sing whenever they
36:23 hear a
36:26 guitar
36:33 Kyle I can't read it it's going too fast
36:36 but surely that's a piece of poop isn't
36:38 it maybe let's go look at
36:46 it title why some dogs sing whenever
36:49 they hear a guitar dogs have been our
36:51 loyal companions for thousands of years
36:53 and their ability to surprise us with
36:55 unique behaviors never ceases to amaze
36:57 us
36:57 K9 connection to music mimicking
36:59 Behavior emotional connection social
37:02 interaction breed and personality
37:05 factors not all dogs sing when they hear
37:07 a guitar true I've had a lot of labs
37:09 none of them sang
37:11 reinforcement yeah well champ makes a
37:14 makes
37:15 a he's got an audience he's got an
37:18 audience so it'll do that it'll write
37:20 that you could also say
37:23 um let's
37:26 see create a table that compares dogs
37:30 that sing to guitar with howling
37:38 wolves okay but wait where did
37:43 that H whoever just did that uh survey I
37:48 don't see it up here to turn it on try
37:52 that again something went
37:54 wonky just did a complete marketing plan
37:57 made a website in less than 2 hours one
37:59 love yeah exactly exactly this this
38:02 shit's
38:04 nuts um oh by the
38:07 way howling wolves and dogs that sing
38:09 guitar like here's the table it just
38:11 made comparing those two things like
38:14 this table didn't exist before the g&
38:17 GPT stands for generative it's
38:19 generating this stuff it's generating
38:21 original content it's not copying and
38:23 pasting this is not a plagiarism machine
38:26 this is not a plagiarism machine this is
38:29 not a plagiarism
38:31 machine
38:33 okay it's going to be categorized as
38:35 that why sells more newspapers it's not
38:38 how it works though there going to be
38:41 lots of lawsuits for lots of years uh
38:42 what else could you do you could
38:45 say I want to create a python
38:48 application that lets people talk about
38:50 the kinds of behaviors they like in dogs
38:53 and then it recommends breeds for them
38:55 please write that for me
39:06 have you ever tried to Learn
39:08 Python me
39:10 neither one of the first Kevin
39:13 McAllister moments I had with chat GPT
39:16 was I thought hey I heard it can write
39:19 code let me try some of that and I did
39:22 something essentially exactly like this
39:25 and it wrote all this code and I'm like
39:28 hey I don't know how to run python code
39:31 I was just
39:33 kidding and it said here's how you run
39:36 python code and I'm like no no I don't
39:37 want to run the terminal in on my
39:40 Macintosh I got a GUI for a reason I
39:42 don't like command line [ __ ] I said
39:44 isn't there a website I can try to run
39:47 this python code it said sure here's
39:49 five of them the first one was repet
39:51 which is like the most famous one the
39:53 the the one people
39:55 like within 90 minutes I had written a
39:58 fully functional python
40:01 application that was making API calls to
40:04 open Ai and it generated five pieces of
40:07 social media based on three little text
40:09 inputs I did it in 90 minutes I've still
40:12 never written a line of python in my
40:14 life survey is up I don't see it silver
40:20 fox normally I've got it it it shows up
40:23 up here where oh wait there it is okay
40:26 there it is
40:28 all right you've got a minute so up
40:31 there there's a survey so what we want
40:33 to do is for the 115 people on there
40:36 click on the in the upper left corner it
40:38 says chat GPT
40:40 3.5 it's a little
40:44 survey and just put whether you're
40:47 you're paying 20 bucks a month for chat
40:49 GPT plus you use it with
40:52 3.5 you don't know wait say what or you
40:55 don't know I don't know how the last two
40:57 are different but wait how many here use
40:59 chat GPT
41:00 4 all right so go do
41:05 that 12 votes you got a minute left
41:08 minute left Let's go people go to the
41:11 survey up there over there it's a little
41:14 thing that looks like this and you click
41:16 on it well you don't click click is so
41:18 Gen X you tap on it you What's this
41:22 called is that a
41:25 tap that's very dist disturbing Kyle
41:27 could you stop doing that with your
41:28 finger I could but it's
41:33 funny go go make clicks on that click
41:37 click tap that tap click
41:41 it hey hey Marge Marge Marge
41:45 Marge you watching the wheel yeah Hey
41:49 listen how do you turn on the tick
41:51 tock you
41:54 what you it's just on
42:00 what how there it says there's people
42:03 here you
42:06 what should I turn it off don't don't
42:08 turn it off say what to the
42:11 people say say
42:13 what oh teach him about
42:16 AI okay all
42:20 right hi people we'll teach you about AI
42:24 here all right this is a skewed audience
42:27 92% of the people in here are paying 20
42:30 bucks a
42:32 month that's a lot that's good 39
42:37 votes I personally think right now chat
42:40 GPT 4 the 20 Buck a month version of
42:43 chat GPT go to chat.
42:45 open.com and then in the lower lower
42:47 leftand corner click on your name and go
42:50 to my
42:51 plan and if you can afford 20 bucks a
42:54 month get the 20 Buck a month month uh
42:57 chat GPT Plus
42:59 subscription and learn it learn what's
43:02 in it but Kyle I don't want to learn I'm
43:04 not very technical why do you always
43:06 tell me to learn I don't want to learn
43:08 I'm not back in school I have worked
43:11 very hard today I don't want to learn
43:13 your fancy AI
43:16 business that's okay just go pay your 20
43:19 bucks and start learning it just here
43:21 here's all you need to do to learn it
43:22 show up here every night seriously just
43:25 show up here and dick around stick
43:26 around with AI it will serve you all
43:30 right since when does 3.5 write code
43:32 it's always written code it just it
43:34 doesn't write it as good as four four
43:36 does everything better um but yeah 3.5
43:39 is always done
43:41 code I'm trying to think what else 3.5
43:43 does you can talk to 3.5 if you get the
43:46 IOS app or the Android
43:48 app
43:50 um and you can do Co you can do custom
43:53 instructions with it anyway I'm not
43:55 going to run down this rabbit hole it
43:56 was a three and a half hour Rabbit Hole
43:58 I so last night if you go watch on if
44:01 you go to our YouTube channel it's it's
44:03 AI learning la- TT on
44:06 YouTube I put up last night's recording
44:11 I did a three and a half hour session
44:13 last night that was there were high
44:15 Jinks it there was it was it was a
44:18 [ __ ] standup comedy routine without
44:20 the laughter it was high Jinks there
44:23 were singing dogs there were questions
44:26 there were answers we demoed chat gbt
44:29 3.5 we demoed custom instructions we
44:32 demoed like 2/3 of what chat gp4 can do
44:36 it took like two hours to like really
44:39 dig into just not not really dig into to
44:43 kind of scratch the surface on what's
44:46 possible with chat
44:49 gp4 how where do you pay for chat GPT go
44:53 to click on your name and go to my
44:58 plan and ignore team for now just ignore
45:02 it just it's not ready pretend they
45:06 didn't launch it just ignore that [ __ ]
45:08 it's a piece of steaming poop by the way
45:11 I don't work for open aai I don't get
45:13 paid by open AI it's been well
45:16 documented by the New York Times that
45:17 Sam mman is jealous of me he doesn't
45:20 even want me talking about a stupid
45:21 company but that's fine I talk about it
45:24 anyway because it's the best one out
45:26 there
45:27 so I don't give two shits if you don't
45:29 pay the 20 bucks if you can afford the
45:32 20 bucks if you can give up three
45:35 Starbucks lattes a
45:38 month this is the best 20 bucks you'll
45:41 ever
45:42 spend and like learn what it does and
45:44 you're like but Kyle I don't know what
45:45 it does how would I learn what it does
45:47 you go to the AI Salon like I told you
45:49 before you join the AI salon and then in
45:52 the AI Salon you go to this little page
45:56 here called AI tools and resources and
45:59 the very first thing on AI recommended
46:02 recommended top AI salon tool
46:05 recommendations the very first one is
46:08 chat
46:09 JP you can make money with chat
46:12 J and I talk about here's what chat GPT
46:16 3.5 can do and here's what chat GPT 4
46:19 can do every one of these bullet points
46:22 is a whole another Rabbit Hole you
46:24 should be exploring this is is all
46:26 within chat
46:29 GPT and and they're not broken out as
46:33 separate tools it can do all of them in
46:36 conjunction with one another in a long
46:38 ass conversation you have with it you
46:41 could be talking about your you know
46:44 Danielle started her new business she
46:47 could be talking about I want to come up
46:49 with a name for the business okay it did
46:51 that now I want to come up with a
46:52 tagline now I want to come up with a
46:53 marketing plan now I want to come up
46:55 with a business plan now I want to
46:57 figure out how I Target certain customer
46:59 groups now I want to come up with a
47:03 logo now I want to come up with an image
47:06 for Instagram post now I want to come up
47:08 with the Instagram post just on and on
47:10 and on and just does all this
47:12 stuff maybe she has a friend come over
47:15 comes over and they have a whiteboard
47:18 session where they're brainstorming on
47:21 the Whiteboard and she's like man that
47:24 was a great brainstorming session if
47:26 only there was a way I could get the
47:27 notes from that whiteboard into chat
47:35 JP I'm going to go to chat jp4 I'm going
47:38 to click on this little
47:43 paperclip I'm going to go find that
47:46 picture that someone
47:49 took of the
47:55 Whiteboard
47:58 that
48:00 one and I'm going to say to chat
48:03 jpt what is
48:11 this I noticed an article in the salon
48:14 tonight comparing Bard to chat gp4
48:18 thoughts Bard right now compared to cat
48:21 gp4 is an an absolute joke um when they
48:25 release
48:27 um Gemini
48:29 Ultra then and and assuming that Gemini
48:32 Ultra lives within Bard then it won't be
48:35 a joke but right now Bard is almost as
48:38 good as chat GPT
48:40 3.5 what's the deal with Microsoft's 20
48:43 Buck a month co-pilot subscription
48:45 literally the same thing it's not
48:46 literally the same thing um so
48:50 Microsoft's co-pilot subscription is to
48:53 have chat GPT like services in
48:56 incorporated into Office
49:00 365 the chat GPT the the co-pilot
49:03 microsoft.com the thing that's like chat
49:05 GPT is actually free from
49:07 Microsoft co-pilot is the thing where
49:10 it's integrated into your Office 365
49:13 Suite it's a very different thing and
49:15 also chat GPT from Microsoft is not the
49:18 same as chat GPT
49:20 from uh from open AI it's they're two
49:23 totally different applications they both
49:26 use the gp4 large language model the
49:29 foundation model but they're different
49:31 applications we did a whole thing last
49:34 night with with a
49:37 box and ink
49:39 cartridges and so the box is GPT 4 this
49:44 is chat GPT and this is co-pilot they're
49:47 different
49:48 applications and then
49:51 [Applause]
49:52 Microsoft lobotomized their GPT when
49:56 they got some bad press from The New
49:57 York Times which is why it's
50:01 shittier um but it's incorporated into
50:04 Office 365 anyway back to this thing
50:07 Danielle's starting her business and now
50:09 she shifted from starting a a laundry
50:12 business to now she started a um a food
50:16 a food ordering app designed for
50:18 airports and that's what that's what
50:20 this little so that's what this picture
50:22 was this is a whiteboard that I found
50:26 the
50:27 internet I stuck it in here it says it
50:29 seems to be a UA UI design of an
50:31 application that does a bunch of stuff
50:32 so it reads all that stuff I could say
50:35 um
50:36 please um write out the text using
50:43 markdown to make it easier to
50:54 read
50:57 you ever had [ __ ] on your whiteboard for
50:59 like months and you're like I just can't
51:00 bring myself to type it in guess what
51:04 you don't have
51:07 to anyone having a Kevin mallister
51:10 moment yet you should be if you haven't
51:13 seen this ship
51:16 before well it's just OCR Kyle it's it's
51:19 relatively trivial uh optical character
51:21 recognition has been around for for
51:23 quite some time uh Mr Shannon
51:28 okay true I'll give you that
51:38 um I'd like to come up with a clever
51:41 name for this app can you please write a
51:43 creative brief for how you would name
51:45 this app and then give me 20 possible
51:48 app names based on that creative
51:54 brief
51:58 project name airport ordering app naming
52:00 to create a name for an airport and food
52:02 app target audience business Travelers
52:04 got that from the Whiteboard families
52:06 got that from the Whiteboard tech savvy
52:08 individuals got that from the fact that
52:09 it was a digital
52:12 interface tone and personality the name
52:14 should be Innovative friendly efficient
52:17 International unique value
52:19 propositions keywords 20 possible names
52:24 okay all right okay all right well
52:26 anybody can come up with names Kyle
52:27 that's just stupid all right great fine
52:30 let's do uh fly Feast Arrow eats Gate
52:33 Gourmet uh Sky surf no Jet Set menu
52:36 quick gate quick Gate's not bad Sky
52:39 snack
52:41 no Globe Trot grub OH
52:45 Close Airway amenities flight Fair
52:47 terminal treats terminal treats that's
52:51 horrible it's just bad all right so
52:54 we'll go we'll go boarding bite I I
52:56 don't feel like working okay fine so
52:58 we're going to
53:00 go
53:04 okay okay I like the name boarding bite
53:06 now please come up with a
53:09 tagline that brings that name to
53:16 [Music]
53:21 life awesome now can you visualize what
53:26 a poster in the airport might look like
53:28 that includes the name of the app a
53:30 screenshot of the ordering interface and
53:33 the tag
53:43 line now it's making
53:45 images who
53:47 [Laughter]
53:52 me was that from the Sydney AI love
53:54 affair yeah why they lobotomized uh uh
53:58 co-pilot yeah that was that was from
54:01 when Sydney told Kevin Roose that she
54:03 loved him and he should leave his wife
54:04 for her that that made the cover of the
54:07 New York Times boarding bite look there
54:10 you go okay so we just went from a
54:13 stupid ass whiteboard that I just
54:15 randomly found on the internet to a
54:18 poster in an airport with the app the
54:21 interface the name of it oh wait there's
54:25 more
54:29 please recommend a technology strategy
54:32 that includes the front end the back end
54:35 and the database technology and provide
54:38 sample code for
54:44 each this is my new parlor trick for
54:46 chat gb4 that this one just kills me it
54:49 just blows me away like do people get
54:52 what's happening in the world right now
54:56 look at this Rudy if you want to speak
54:59 to G GPT like Kyle on a Mac but on
55:03 Windows press
55:05 the plus G
55:08 Keys what is that command G or Control
55:13 G oh oh the oh press Windows plus G okay
55:18 cool if you want to talk to it yes so
55:21 how I'm talking to it right now is I'm
55:23 just using a Macintosh accessibility
55:25 feature called dictation and so on a on
55:28 a MacBook Pro F5 is the microphone key
55:32 and that's mapped to that function so
55:34 that's how I'm talking to it but but
55:36 look what this did let's let's go back
55:38 here so so there's our little marketing
55:40 plan we could also do interfaces for
55:42 this um but it says front end react
55:45 native reasoning react native allows you
55:47 to create mobile apps that's
55:48 indistinguishable from the app built um
55:51 from an app built using Objective C Java
55:53 or Swift uses the same fundamental UI
55:56 building blocks as regular IOS and
55:58 Android apps you just put those building
56:00 blocks together using Java JavaScript
56:02 and
56:03 react so there's your there's your react
56:06 code back end nodejs with Express gives
56:09 you the explanation for that there's
56:11 your JavaScript for that database
56:13 mongodb here's the sample code for
56:22 that wait it's Windows key plus h G
56:25 sorry okay Rudy two demerits for Rudy so
56:29 so Amelia's wife gets one demerit for
56:32 being late and Rudy gets two demerits
56:35 for giving us the wrong command key on
56:36 Windows Rudy I thought you were a
56:40 professional Kyle I set up my keyboard
56:43 to use the function key twice no need to
56:45 lift to cover the keyboard I could do
56:47 that Corey that's not a bad idea um yeah
56:51 I got to I got to listen there's a whole
56:54 situation with my setup here which it's
56:56 getting better thanks to the Irregulars
56:58 who gave me a lovely at least I'm not
57:00 using a motorcycle box to hold up my a
57:03 motorcycle helmet box to hold up my uh
57:06 my computer anymore but yeah I got to
57:08 get my [ __ ] together with the uh with
57:09 the old setup here the brown bag is for
57:13 it's for this this is not just any old
57:15 brown bag this is a blandon's bag and
57:19 ask me how I got a blandon's
57:22 bag I was I was gift ambushed
57:25 by the Irregulars in here shortly after
57:30 the New Year where they sent among other
57:36 things this which if you know anything
57:39 about bourbon this is a bit of a holy
57:41 grail and it's the best [ __ ] bourbon
57:45 out there so anyway I'm probably going
57:47 to get Perma Bann from Tik Tok for
57:49 showing it but
57:50 whatever
57:53 whatever um anyway
57:56 anyway so so
58:00 yeah so have you gone to chat GPT
58:05 yet if you haven't gone to chat GPT go
58:08 there
58:11 now no let's do
58:14 it are there other tools you could use
58:17 yes an infinite number of them right
58:20 now but you can talk to it and it can
58:23 talk back correct you can do this on
58:25 both chat GPT 3.5 and chat gp4 this is
58:28 an iPad with a with a licorice flavored
58:32 keyboard if you don't have the licorice
58:34 it's it's black Li licorice flavored if
58:36 you don't have the black licorice
58:37 flavored official Apple keyboard [ __ ]
58:40 delicious you should get one anyway um I
58:44 do get commissions on all of the
58:45 licorice keyboards that that are sold um
58:50 you buy them for $129 I get
58:53 $125 and they're [ __ ] tasty all right
58:58 go look for it on Amazon it's right
58:59 there right now Li black licorice iPad
59:03 keyboard like magic keyboard or
59:05 something like that okay see that that's
59:08 the chat GPT app click on that and then
59:12 put it into right now it's chat GPT 3.5
59:15 I I'm going to put it in chat GPT
59:17 4 and but you can do it in either one in
59:20 the lower right hand corner there's
59:21 those little headphones you can
59:24 go
59:28 what years were the Ford Bronco
59:37 made the Ford Bronco was originally
59:40 produced from 1966
59:43 to and reintroduced in 2021 so it has
59:47 two distinct production periods from
59:49 1966 to
59:51 1996 and starting again in
59:54 2021
59:56 make me a picture of an orange Bronco
59:58 from
59:59 1972 rock climbing in the
1:00:06 desert so not only can you talk to it
1:00:08 and it talks back you can ask it to make
1:00:11 pictures so you know that thing I did
1:00:13 where I I told it to say hey make me
1:00:15 that concept for the airport thing and
1:00:19 you could be doing that in your car on
1:00:21 the way to work just talking to it and
1:00:23 having it make look here's the image of
1:00:25 a bright orange 1972 Ford Bronco rock
1:00:28 climbing in the desert it captures the
1:00:30 dynamic and rugged nature of the Bronco
1:00:32 in a picturesque desert setting let me
1:00:35 know what you think you can check out
1:00:37 the images in our conversation
1:00:40 history I like it okay I'd like you to
1:00:43 write an advertising headline to go with
1:00:45 that
1:00:50 image conquer the wild Unleash the Power
1:00:53 of the 72 Bronco in every Adventure
1:00:57 that's great but I want to publish it in
1:00:59 a German magazine can you tell me that
1:01:01 in
1:01:08 German
1:01:09 sure in German the headline would
1:01:17 be that's so good but we're really
1:01:20 trying to break into the Japanese Market
1:01:22 could you give me it in Japanese too
1:01:28 of
1:01:29 course in Japanese the headline is
1:01:37 co this should resonate well with the
1:01:39 Japanese Market okay so now I was
1:01:44 driving in my car I put it in Bluetooth
1:01:46 mode and I was having a conversation
1:01:48 with
1:01:49 it I'm laid on a deadline a client needs
1:01:52 an ad for a 1972 Ford Bronco okay great
1:01:57 I get to
1:02:01 [Laughter]
1:02:09 work new
1:02:12 chat
1:02:17 look look I was just talking that with
1:02:20 my words and I get to my desk and look
1:02:25 here it is in
1:02:27 German here it is in
1:02:34 Japanese any Kevin McAllister moments
1:02:37 going on out
1:02:39 there you guys should be diving
1:02:43 deep what am I looking at now I don't
1:02:47 know what were you looking at when you
1:02:50 wrote
1:02:51 that I don't see I don't see your
1:02:54 comments in real time cuz I have mods
1:02:57 wait missed the intro for this tool the
1:02:59 talking the one where I was talking it's
1:03:00 chat
1:03:01 GPT the one that made this image is chat
1:03:04 GPT the one that translated that took my
1:03:08 talking and turned it into words and
1:03:10 then I moov to my desktop machine and
1:03:12 here are the words as well as the
1:03:14 picture chat
1:03:18 GPT the one
1:03:21 that
1:03:23 turned
1:03:26 this whiteboard
1:03:29 image into a transcription of what was
1:03:31 on
1:03:33 it into 20 possible
1:03:37 names into a marketing thing into Tech
1:03:42 recommendations and code chat
1:03:47 GPT this is why this is why when you go
1:03:51 to the AI Salon AI tools salon
1:03:55 tough a is salon tool recommendations
1:03:58 the very first thing is like
1:04:01 this long ass thing about chat
1:04:05 GPT cuz that one tool if you pay 20
1:04:08 bucks a month does all this
1:04:10 [ __ ] and if you want to get AI literate
1:04:13 you should understand what's here
1:04:14 because Chachi pt5 is coming and it's
1:04:18 going to be way more intense like this
1:04:19 is
1:04:21 nothing and this is [ __ ] intense just
1:04:24 chat GPT 3.5 if you really dug into it
1:04:28 is
1:04:29 intense that conversation I had you can
1:04:31 have a
1:04:32 3.5 custom instructions where you can
1:04:34 give chat GPT its own personality and
1:04:37 have it behave very differently than it
1:04:39 does now that's part of 3.5 it's super
1:04:42 fast it's wicked
1:04:45 deep and then if you pay the 20 bucks a
1:04:47 month then you get all this other
1:04:49 [ __ ] what is all that other [ __ ] Kyle I
1:04:51 don't understand the exact get in there
1:04:54 get get in there go to this document
1:04:56 I'll tell you what all that stuff is and
1:04:58 start playing and start sharing what
1:05:00 you're learning really important right
1:05:03 now really
1:05:05 important all the Irregulars the people
1:05:07 that come to this channel night after
1:05:09 night after night I'm in here seven
1:05:11 nights a week for two hours a night last
1:05:14 night it was three and a half hours cuz
1:05:15 I'm
1:05:23 stupid
1:05:25 why are they doing that my super model
1:05:28 good
1:05:31 looks to hear my dog sing for three
1:05:34 minutes at the beginning of the
1:05:37 show Lori duskin Ryan best $20 I spend a
1:05:40 month hands down 20 bucks for chat GPT
1:05:43 is the it's the best deal in software
1:05:45 it's like I think it's the best deal in
1:05:47 software
1:05:49 history I make no commissions off this I
1:05:53 don't care if you buy it or not
1:05:57 he just says that I bet he's making a
1:05:58 commission that's what all those people
1:06:00 do you know what they do they say this
1:06:02 is a fair and balanced um just a
1:06:05 recommendation and then what you find
1:06:07 out they had a brand deal for a year I
1:06:10 bet he's been in Sam alman's office I
1:06:12 bet he I bet Sam calls
1:06:15 him nightly every once a week to ask his
1:06:23 advice
1:06:26 possible I
1:06:29 [Laughter]
1:06:39 suppose Corey Sandler Pottery I'm
1:06:41 sharing with my Pottery crew and several
1:06:44 have come here to check you out oh boy
1:06:46 sorry
1:06:53 Corey hey P loty crew so Cory's wait we
1:06:56 got so I'm going to show you what a what
1:06:58 a GPT is if you don't know what a GPT is
1:07:03 that's another thing you can do for free
1:07:04 or no as part of the 20 bucks a month
1:07:06 with GPT 4 you can create these things
1:07:09 called gpts which are like little
1:07:11 customized personalized versions of chat
1:07:15 GPT so if you click on explore gpts you
1:07:18 need chat GPT 4 to be able to do this
1:07:21 but if you click on that it'll take you
1:07:23 to what open AI calls the GPT store
1:07:26 which is a horrible name because it's
1:07:28 not a store but whatever I'm I'm bitter
1:07:30 about this but you can at least find
1:07:31 stuff so if I look for
1:07:34 Corey
1:07:37 Corey
1:07:38 [Music]
1:07:42 sand let's see the
1:07:45 Imaginarium Pottery Whisperer Pro so
1:07:49 she's got a she's got a bunch so Corey
1:07:50 Sandler you should go check out her
1:07:52 stuff her her gpts are [ __ ] cool
1:07:54 she's got one called iconic that makes
1:07:56 groovy glasses she's got one that'll
1:07:59 make she she combined the Artistry of
1:08:03 the cinabun with SH shoe design so it
1:08:07 makes shoes that look like they were
1:08:08 made by the cinabun
1:08:12 chef oh my and
1:08:16 wait hang on a sec so wait let me go
1:08:20 Pottery whisper okay so
1:08:23 this
1:08:26 is it creates unique glazes you choose a
1:08:29 color a cone I didn't know what a cone
1:08:32 was I but I do now a cone is a
1:08:35 temperature zone for pottery and and
1:08:39 anyway so so you say create a create a
1:08:41 random glaze for cone 6 Pottery so you
1:08:44 click on that watch what this thing does
1:08:47 is chat GPT the place to build
1:08:50 presentation slides not really Don Smith
1:08:53 you can write write the copy for the
1:08:55 presentation slides in chat GPT which
1:08:58 will be really good and you can make the
1:09:00 images for the slides in chat GPT but it
1:09:03 doesn't automatically assemble it my
1:09:06 problem with the slide the with with the
1:09:10 the things that actually make you a like
1:09:12 a PowerPoint presentation is they're a
1:09:14 little rigid and it's I don't know I'd
1:09:18 rather just do it in chat GPT and then
1:09:20 copy and paste the crap over to
1:09:22 PowerPoint or Keynote um okay so so
1:09:25 check this out so this is I don't know
1:09:28 this is going to be called The Mystic
1:09:30 nebula and this has got potach feldspar
1:09:33 30% Whiting
1:09:35 20% EPK cin 15% silica Flint 25% and
1:09:41 talc 10% I can see it now can't you no
1:09:44 neither can I I don't know anything of
1:09:46 what what the [ __ ] is she talking about
1:09:48 here but then watch what you do watch
1:09:50 you're like um can I see what what this
1:09:55 would look
1:09:57 like so it says this glaze should create
1:10:00 a captivating and dynamic surface on
1:10:02 your pottery with the depth of the ocean
1:10:05 and the Mystique of a Starry Sky
1:10:06 remember glaze results can vary greatly
1:10:09 right so so here's a recipe for you to
1:10:13 put together and stick on your dang
1:10:15 pottery look what
1:10:18 look how cool is
1:10:21 this I'm telling you
1:10:25 how do you activate plugins with your
1:10:27 voice I don't that I don't know I don't
1:10:30 know if you can i h probably you might
1:10:34 be able to I don't know I've never tried
1:10:36 it anyway that's that's Cory s that's a
1:10:38 GPT like so so what gpts are is people
1:10:42 take their really specific expertises or
1:10:44 passions and without needing to do any
1:10:47 coding they basically give you a tool
1:10:51 that does
1:10:53 something that they've got expertise in
1:10:56 it's amazing it's absolutely amazing I
1:10:59 want to use AI to evaluate football
1:11:01 plays and lifting
1:11:03 form any ideas if that's
1:11:08 possible
1:11:12 um doing it with video right now would
1:11:15 require some programming but GPT Vision
1:11:19 so so go watch so which HCC strong go
1:11:24 watch um there's videos on
1:11:28 Twitter where they're
1:11:30 using GPT Vision Plus text to voice
1:11:35 where they're they're analyzing like
1:11:37 video of a kid playing a video game and
1:11:40 it's generating the
1:11:42 commentary in real time while the game's
1:11:45 playing so like that's possible but so
1:11:47 let's let's go to try and experiment so
1:11:49 you said
1:11:51 um football plays
1:11:55 all right so let's
1:11:57 go let's go to the internets we'll go
1:12:01 um photo
1:12:05 of football play
1:12:11 developing no not that
1:12:13 football oh I spelled it with one L dang
1:12:20 it with of American
1:12:25 football play
1:12:32 developing
1:12:37 images
1:12:40 uh this is the wrong kind of
1:12:45 football let's say of oh
1:12:50 college football play developing okay
1:12:53 okay let's see if we can let's see let's
1:12:55 see overhead
1:13:04 photo all right is that
1:13:08 one
1:13:12 sure
1:13:19 oops copy
1:13:21 image I'm going to go chop some of this
1:13:23 other [ __ ] off it hang
1:13:35 on we're going to get rid of Rose Ball
1:13:37 we're just going to have
1:13:40 players all right and we're going to go
1:13:42 tools
1:13:45 crop all right there's our there's our
1:13:47 there's our thing
1:13:51 Save By the way I do not know if this
1:13:54 works this is part of the joy of this
1:13:56 channel is you tell me to do stupid [ __ ]
1:13:59 and I'll go try it and you know what
1:14:02 happens like nine times out of 10 it
1:14:04 blows my [ __ ] mind
1:14:07 football okay football to the desktop
1:14:11 great okay so we have an image we're
1:14:14 going to go to chat J we're going to do
1:14:17 a new chat now you got to be in chat gp4
1:14:20 if you're in chat gp4 you get this
1:14:22 little pi paper clip you click on that
1:14:25 paperclip we're going to go find that
1:14:27 fruitball image
1:14:29 fruitball so football image is in there
1:14:31 so I'm going to say I know nothing about
1:14:34 this act
1:14:36 as a division
1:14:40 one
1:14:43 coach and
1:14:48 analyze the play that is
1:14:52 developing and what I should know about
1:14:58 it as a
1:15:03 Layman the image I see is a football
1:15:06 play in progress here's the Layman's
1:15:08 analysis formation and personel the
1:15:10 offense the team with the ball is lined
1:15:12 up in what appears to be a standard
1:15:14 formation with a quarterback under
1:15:16 Center that's not true this the
1:15:18 quarterback's rolling rolling out
1:15:21 meaning he's directly okay the defense
1:15:23 is counter what's happening the play has
1:15:24 just started the offensive line is
1:15:26 engaging with the defensive line the
1:15:29 running back
1:15:31 37 oh maybe that's maybe they think
1:15:34 that's the running back I think that's
1:15:35 the quarterback because the way he's
1:15:36 holding the
1:15:37 ball
1:15:40 um is moving toward the line of
1:15:42 scrimmage where the ball was snapped
1:15:45 indicating a running play the
1:15:46 quarterback number eight has multiple
1:15:48 options
1:15:50 here it's just making this [ __ ] up
1:15:54 so this is called
1:15:57 hallucinations as a coach you should
1:15:59 know all right I'm going to try this
1:16:02 again let's do it let's do a new a new
1:16:04 chat GPT I'm just I'm going to get same
1:16:10 picture um what is going on
1:16:19 here moment from an American football
1:16:21 game play in progress players from both
1:16:23 both teams on the field the
1:16:25 team on the offense which is the team
1:16:28 with the possession of the football
1:16:29 appears to be exe executing a play
1:16:32 possibly a running play given the
1:16:34 formation and positions of the player
1:16:37 defensive team's trying to stop them the
1:16:38 action taking plac near the
1:16:40 Zone
1:16:43 um who has the
1:16:52 ball
1:16:54 yeah it's not quite doing it here um so
1:16:59 it's possible you could do this and it's
1:17:01 possible you could tray look look for
1:17:03 the play cards X's and O's X's and O's
1:17:06 it'll probably do I I was just trying to
1:17:08 get it to do analysis of actual photo
1:17:11 stuff but let's let's see if we can get
1:17:12 it to to do something better with um
1:17:16 with x's and o's X is an
1:17:19 O so we'll say um
1:17:25 football play
1:17:30 design all right here we go let's get a
1:17:34 complicated
1:17:35 one all right that looks pretty
1:17:38 complicated great so we'll do save image
1:17:41 as we'll do
1:17:44 football let's see football
1:17:51 2 but you see what I'm doing here right
1:17:55 and this is this is my advice for anyone
1:17:58 wanting to learn
1:18:01 is nobody has the answers to any of this
1:18:05 [ __ ] this stuff this this Tech is too
1:18:08 new like are there people who figured
1:18:10 out how to do this I would I would wager
1:18:12 we can go look and see I would wager
1:18:14 that there's probably a GPT that's been
1:18:16 built that does some sort of play
1:18:18 analysis but let's
1:18:21 uh but but the best way learn this stuff
1:18:24 is just do it just like if you have a
1:18:27 question like I I need to analyze
1:18:29 football plays more better can it do
1:18:31 that I don't know
1:18:34 maybe
1:18:36 um tell me what happens
1:18:42 here and name the
1:18:52 play
1:18:54 image appears to be a diagram of
1:18:56 football play specifically an American
1:18:58 football game the design of the field
1:19:00 and the markings it's strategic plan in
1:19:02 the diagram few key moments the lines
1:19:04 and arrows represents okay the Shaded
1:19:06 okay whatever unfortunately without the
1:19:08 label of context it's not possible I'm
1:19:10 going to say try again try again look at
1:19:16 the shapes and
1:19:20 arrows and take a guess
1:19:25 this is an important thing if chat GPT
1:19:27 gives you shitty answers tell it it gave
1:19:29 you shitty answers and it'll try it's
1:19:31 amazing it what what is at a minimum
1:19:33 what it'll try to do is it'll try to
1:19:35 help you the diagram represent here's a
1:19:38 breakdown of what could be happening the
1:19:39 brown shaded arrow with the pink
1:19:41 rectangles could indicate a player like
1:19:43 the quarterback rolling
1:19:47 out
1:19:49 the okay he's looking there he's going
1:19:52 to throw there there oh I see
1:20:02 okay the yellow arrow seems to show a
1:20:06 receiver running a deep route
1:20:07 potentially a post route angling toward
1:20:09 the center of the field that's right
1:20:11 that's what that is oh no he said the
1:20:13 yellow arrow that's not right I don't
1:20:17 know anything about football though so
1:20:18 wait The Black Arrow shows the
1:20:20 receiver's
1:20:22 routes
1:20:24 yeah it it it it doesn't get this so
1:20:27 it's clearly not really been trained on
1:20:28 football stuff but let's go see if
1:20:32 there's any gpts that
1:20:34 do um let's do
1:20:41 football football predictor there's
1:20:43 going to be all sorts of fantasy
1:20:44 football [ __ ] so we'll do football
1:20:50 play digital football coach advisor
1:20:58 grid iron Guru football quarterback a
1:21:00 strategic football game that's
1:21:04 cool so the coach
1:21:07 advisor supports youth in high school
1:21:10 plan and strategize what questions can I
1:21:12 ask how do I get started what's the
1:21:14 three and four nickel defense anyway all
1:21:18 right that was that was a a a fairly
1:21:22 rare wing and a complete Miss by chat
1:21:25 GPT it's obviously not really trained on
1:21:27 that kind of that kind of data with any
1:21:30 sort of actual training um so I would
1:21:33 say there's an opportunity to go do a
1:21:36 fine-tuned model where you could train
1:21:39 it
1:21:40 on images and make something that does
1:21:44 that you can absolutely do that like
1:21:46 it's all this Tech is relatively trivial
1:21:48 to put that [ __ ]
1:21:52 together but it looks like chat GPT out
1:21:54 of the box is shitty at that is there
1:21:57 any indication on rating chats actually
1:22:00 helps open AI learn well that's ex
1:22:03 that's entirely how so the
1:22:08 way um the way chat
1:22:12 GPT learns is called it's called
1:22:15 reinforcement
1:22:16 learning and so it's got human beings
1:22:19 when it generates answers human beings
1:22:21 go thumbs up thumbs down here's it was
1:22:23 good and every time they do that it
1:22:26 actually does train the model so yes it
1:22:30 it that's that's how open AI does what
1:22:33 they do
1:22:35 now if you want to see something if you
1:22:37 want to read something interesting about
1:22:39 how you train these
1:22:41 models read up or watch YouTube videos
1:22:44 on a company called
1:22:46 anthropic they're the company behind
1:22:48 Claude aai claw. A claw. in my opinion I
1:22:53 is kind of the second runner up to chat
1:22:55 GPT in terms of quality of large
1:22:58 language model right
1:23:00 now it just happens to be founded by a
1:23:03 bunch of xop AI people that made chat
1:23:08 gpt3 so they know what the [ __ ] they're
1:23:11 doing and they use what's called a
1:23:13 constitutional learning model so instead
1:23:15 of human beings go thumb UPS thumbs down
1:23:18 what the Constitutional learning model
1:23:19 does is it gives there's there's like
1:23:21 five imperatives five
1:23:23 rules that that are like you know don't
1:23:27 harm Do no harm increase Prosperity um I
1:23:32 don't know make people happy whatever it
1:23:34 is it's got five rules or seven rules
1:23:37 whatever the Constitution
1:23:39 says there's some set of
1:23:42 imperatives and when when Claude
1:23:44 generates an an answer it com it it's
1:23:47 comparing it to its
1:23:50 Constitution and so it's always me
1:23:52 measuring it consistently without human
1:23:57 bias so I think it's a really
1:23:59 interesting model and their whole thing
1:24:02 is that they want to be able to do
1:24:06 um what they call alignment better than
1:24:09 any other company in the world and
1:24:11 Alignment is when when when you're
1:24:13 talking about AI
1:24:15 alignment it basically means this you
1:24:18 have human beings that have their
1:24:20 imperatives we want to be awesome
1:24:24 right and then you have ai and it's got
1:24:26 its imperatives alignment means those
1:24:31 two goals imperatives are
1:24:35 aligned if you have humans over here
1:24:38 going we just want to be swell and you
1:24:40 have ai going um I want to be the most
1:24:43 powerful AI in all of the
1:24:46 land and it starts heading toward being
1:24:49 the most powerful AI in all of the land
1:24:51 and one of these humans gets in the way
1:24:53 of
1:24:55 that it will go
1:24:58 huh I'm getting smarter and those guys
1:25:01 are in my way so my imperative is just
1:25:04 to become the strongest AI ever so why
1:25:06 don't I go on ahead and just shut off
1:25:08 their power see how they do for a
1:25:12 week and then the humans all scrambling
1:25:15 like [ __ ] what happened and the AI is
1:25:17 like all right I got a free week let me
1:25:18 go let me go do some damage that's not
1:25:21 alignment
1:25:23 that's that's the that's the scenario
1:25:26 that the the movies want us to believe
1:25:28 is going to
1:25:29 happen um that's what cla's into but yes
1:25:32 the thumbs up and thumbs down actually
1:25:34 makes a difference Erica how does AI get
1:25:35 smarter and not Dumber if users are the
1:25:38 ones
1:25:40 teaching what corrects the mistakes oh
1:25:44 man there's so it's it's a it's a whole
1:25:46 world in and of itself Erica if you want
1:25:48 to geek out on it a bit um go watch
1:25:51 videos of David Shapiro he's Dave shap
1:25:54 on
1:25:55 YouTube um he's written five books on
1:25:58 AGI he talks a lot about this stuff Kyle
1:26:00 your idea why can't chat GPT pick the
1:26:02 winning numbers for the power ball
1:26:10 game oh yeah it's now you're getting
1:26:14 into probability and and Superstition
1:26:16 they're two different things
1:26:20 um what was I talking about
1:26:27 I don't know doesn't
1:26:28 matter Kyle I don't know anything about
1:26:31 this also Kyle cuac isn't under under
1:26:34 the center he's rolling out well I know
1:26:36 a little bit about football I just the
1:26:38 only thing I know about football is when
1:26:40 I played it in Junior
1:26:44 High people that are really into sports
1:26:47 are of a different breed than me I got
1:26:50 kneed in the rib so hard once it broke a
1:26:52 rib
1:26:53 um but the guy just didn't give a
1:26:56 [ __ ] and I was like oh I get it they're
1:27:00 not like me it's just
1:27:02 like so how does AI get smarter not
1:27:05 Dumber if users the ones teaching well
1:27:07 so users aren't the ones
1:27:11 teaching what users so
1:27:21 okay
1:27:23 I'm trying to think if I want to run
1:27:24 down this Rabbit Hole I could you do you
1:27:27 want me to run down a rabbit hole on how
1:27:28 large language models work I've got a
1:27:31 I've got a presentation I've got a
1:27:33 slide that that will teach you how large
1:27:35 language models work I think it's
1:27:37 [ __ ]
1:27:39 fascinating um but before I do that
1:27:42 the humans are not
1:27:46 teaching chat
1:27:48 GPT how to be better chat GPT is
1:27:54 giving probab probabilistic
1:27:58 answers and then human beings are
1:28:00 grading that and so it basically knows
1:28:02 okay I gave it this thing and it gave me
1:28:04 a bad score and so the next time I'm
1:28:06 going to you know and a bad score means
1:28:08 it went that way so the next time I'm
1:28:10 going to sort of aim it that way and
1:28:12 then so it's just this constant game of
1:28:14 is it better or worse better or worse
1:28:17 and so as humans tell it it's better or
1:28:19 worse better or worse it gets better and
1:28:21 better at predicting what is the right
1:28:24 answer that a human being would want it
1:28:27 these tools right now have no idea what
1:28:30 they write they have no idea what it
1:28:33 means and it's it's one of the confusing
1:28:35 things about them because chbt is so
1:28:39 powerful you look at the output of it
1:28:41 and you're like well it must be sensient
1:28:43 like it's it it knows all this stuff and
1:28:46 it it knows how to write and it knows
1:28:48 how to write code and it it has no idea
1:28:51 what it's writing it's literally a
1:28:55 calculator predicting the probability of
1:28:57 the next word in the
1:28:59 sentence word by word token by token
1:29:02 they call it it's you know words commas
1:29:05 spaces [ __ ] like
1:29:06 that it's predicting that word by word a
1:29:11 mathematical it's just finding some
1:29:13 point in Thousand dimensional Vector
1:29:16 space and saying this point in Thousand
1:29:19 dimensional space attached to this word
1:29:23 is the next most probable
1:29:26 word so it has no idea what it's written
1:29:29 so when humans are giving it thumbs up
1:29:31 thumbs down they're just they're just um
1:29:33 modifying what's called The
1:29:35 Waiting how it
1:29:38 probabilistically determines which are
1:29:40 the right words so that's how it works
1:29:42 so anyway embarrassing question but can
1:29:45 chat GPT be used for poll is live all
1:29:48 right what's the
1:29:50 poll can we go down the Rabbit Hole of M
1:29:53 llm Creations oh okay so we've got a
1:29:57 minute and a half if you want me to
1:29:58 explain how chat GPT Works go to the
1:30:01 poll so the poll's right there click on
1:30:03 that if you want me to explain how llms
1:30:06 work large language models work click
1:30:08 yes and if not
1:30:09 not and then I won't CU yeah exactly
1:30:13 someone just voted no cuz they're like
1:30:15 that's going to bore the [ __ ] out of me
1:30:16 I didn't come here to learn something
1:30:17 it's called the AI learning lab I came
1:30:19 here to be entertained this is Tik tock
1:30:22 people if Kyle doesn't start twerking in
1:30:25 the next 15 minutes I'm out of
1:30:29 here I'll do it too I'll take [ __ ]
1:30:31 lessons I've got
1:30:33 YouTube I got a
1:30:38 booty it's neck and neck right now by
1:30:40 the way you got 30 seconds left to vote
1:30:42 do you want do you want to learn how
1:30:43 large language models work I'll show you
1:30:46 my super fancy slide I
1:30:49 made or not and then I'll just keep
1:30:52 talking about dog farts
1:30:56 [Laughter]
1:31:01 you're oh now now we have 30 seconds
1:31:04 left it was 50 seconds before I've got
1:31:05 old eyes what do you want from
1:31:07 me I'm not going to look at this I'm
1:31:09 going to I don't want I don't want to
1:31:10 see so if you want me to teach you how
1:31:13 large language models work vote Yes for
1:31:14 that embarrassing question but can gp4
1:31:16 be used to create PowerPoint
1:31:18 presentations not embarrassing at all
1:31:20 not dumb at all
1:31:23 um chat gp4 can be used to write the
1:31:27 PowerPoint presentation so you can say I
1:31:29 want a PowerPoint presentation on this
1:31:31 topic I want this many slides in it um
1:31:34 or give me an outline for a presentation
1:31:37 and now write the copy and then because
1:31:39 it can do do you can use it to generate
1:31:42 the images you'll have to then copy and
1:31:44 paste that
1:31:46 into chat GP or into
1:31:49 PowerPoint there are tools like what's
1:31:52 one of them called it begins with a
1:31:57 T old man brain I don't remember what
1:32:00 it's
1:32:02 called um there are tools out there that
1:32:06 will make presentations for you I don't
1:32:08 like them I like you I I just do the
1:32:09 writing in chat GPT so you can write
1:32:12 copy of it but uh but it it won't
1:32:14 actually write it um what was the
1:32:17 poll co-pilot and PowerPoint oh yeah oh
1:32:21 okay 65% of 32 votes want to see how to
1:32:26 make uh how how large language models
1:32:28 work okay fine let me go let me go get
1:32:30 the let me go get the slide and I'll
1:32:33 show you how it works it's actually
1:32:35 pretty
1:32:36 wild I kind of like I kind of like
1:32:38 showing this slide because it every time
1:32:40 I show it it reminds me how just [ __ ]
1:32:43 batshit crazy insane this stuff
1:32:47 is all right I got to download it from
1:32:49 iCloud hang on people don't forget the
1:32:53 credit card oh
1:32:55 Danielle Danielle is referencing a
1:32:57 moment just four days ago where I put my
1:33:00 credit card information on a Tik Tok in
1:33:02 front of about 150 people and p goes you
1:33:06 might want to go cancel that there
1:33:08 buddy which I
1:33:12 did I told the I told the lady at the
1:33:15 bank of like hey uh yeah I just uh I
1:33:18 just showed my my uh full credit card
1:33:20 information to about 50 people on the
1:33:22 internet I'm going to go on ahead and
1:33:24 and uh want to change that number she's
1:33:26 like I think that's a good idea I think
1:33:27 it's a good thing you called
1:33:31 us oh man is there there is a slid maker
1:33:36 custom GPT oh that's
1:33:38 cool okay AI
1:33:43 [Music]
1:33:48 Workshop oh yeah
1:33:51 that's so we've heard from a number of
1:33:53 clients that modular content is
1:33:55 important to them getting more efficient
1:33:57 with how we create content and how we
1:33:59 leverage
1:34:06 content behind me you see these little
1:34:09 green boxes um that's how we yeah that's
1:34:12 from hey Jen if you don't know hey Jen
1:34:16 you should know hey
1:34:20 Jen
1:34:23 does he always sing like that it's very
1:34:26 annoying does he want to get rid of
1:34:27 people I don't understand this Channel
1:34:30 at
1:34:31 all he seems like a bit of a dumb dumb
1:34:34 is he dumb he's not why does he act like
1:34:37 a dumb
1:34:40 [Music]
1:34:42 Dum I gotta find the slide people there
1:34:46 it is all right you're gonna you're
1:34:48 gonna be like oh man that was so cool
1:34:52 you should show us other slides man like
1:34:56 that's so bitching okay how chat
1:35:01 Works does he need to talk to us like
1:35:03 that he seems a little
1:35:06 aggressive okay okay okay yeah hey Peter
1:35:12 yeah we're going to need you to go on
1:35:14 ahead and come on in on
1:35:17 Saturday yeah oh hey and uh one more
1:35:20 thing Peter yeah uh
1:35:22 Hey listen we're going to need you to go
1:35:24 on ahead and come on in on Sunday
1:35:28 too my husband says AI is fancy
1:35:31 autocomplete your husband is
1:35:35 100,000% correct ding ding ding husband
1:35:39 for the
1:35:40 win I love it when a husband's right
1:35:43 fantastic Bob tell him what he's won
1:35:46 just don't tell her you're right just
1:35:47 don't tell her you're right I told you I
1:35:48 was right honey oh swinging a Miss ooh
1:35:52 that's going to be a long evening in the
1:35:55 on the couch
1:35:59 okay been married for 37 years oh yeah
1:36:03 oh yeah um okay how does chat GPT work
1:36:07 well let's start with GPT what does GPT
1:36:10 stand
1:36:11 for the G and GPT stands for generative
1:36:14 the p is pre-trained and the t is
1:36:16 Transformer the Transformer is the
1:36:19 technology that Google invented in 2017
1:36:22 17 that basically makes all of the [ __ ]
1:36:26 that we're playing with today possible
1:36:28 one little white paper in 2017 called
1:36:32 attention is all you
1:36:33 need is the foundation is is the basis
1:36:37 for open AI success and all of these
1:36:39 large language models and all of the
1:36:42 image generation tools and all the video
1:36:43 generation tools it's all based on the
1:36:45 same basic technology it's staggering
1:36:48 that somebody figure out how to do this
1:36:51 it's just it's just mind-blowing so
1:36:53 that's the T the T just basically is the
1:36:55 type of technology that they use
1:36:57 Transformer pre-trained the p
1:37:01 means you don't have to bring your own
1:37:03 data so in the olden timey days like
1:37:05 before
1:37:06 2017 if you wanted to create a a machine
1:37:09 learning model you had to bring your own
1:37:11 data you'd bring Radiology scans or
1:37:14 you'd bring you know population data or
1:37:17 you'd bring whatever data you had and
1:37:19 you'd stick it in the machine and it
1:37:22 would do some weird ass mathematical
1:37:25 stuff and then it would do magical
1:37:27 things and so machine learning models
1:37:29 were these
1:37:30 vertical specialty kind of models the G
1:37:35 in GPT stands for generative and that
1:37:37 means it's actually generating original
1:37:39 content and I'll show you that's what
1:37:41 I'm going to show you here is how it
1:37:42 does that so generative generates
1:37:45 original content pre-trained on whatever
1:37:49 knowledge it's been trained on in the
1:37:50 case of chat GPT
1:37:52 what it's been trained on is as much of
1:37:54 the public internet as open AI could
1:37:57 could put in this particular format and
1:37:59 jam into the thing
1:38:01 so one of the reasons chat gp4 is better
1:38:05 than anyone else out there that no one's
1:38:07 been able to catch them in a year is
1:38:09 that they spent four or five years
1:38:12 getting as much data as they could to
1:38:14 train this thing on Sam mman realized
1:38:17 very very early that the the the better
1:38:21 the quality of the data and the more
1:38:23 data they have and the more gpus they
1:38:25 can train it on the better this thing's
1:38:26 going to be it's why no one's caught
1:38:28 them
1:38:29 yet um okay llm you hear that a lot that
1:38:32 stands for large language model that's
1:38:34 just the name for the the things we're
1:38:36 playing with like chat GPT um things
1:38:40 like um Mid journey and stable diffusion
1:38:43 are diffusion models you know for images
1:38:46 so there's different things large
1:38:47 language models is the language
1:38:50 stuff then there's this this thing
1:38:52 called the latent space and this is
1:38:53 where we get a little geeky but it's not
1:38:56 that bad so the latent space is a
1:38:59 mathematical
1:39:00 representation I say of words but it's a
1:39:03 mathematical representation of points in
1:39:06 mathem in in Vector space clustered by
1:39:09 meaning okay so imagine that every one
1:39:13 of those
1:39:14 dots is is a is a vector is a
1:39:18 number that represents a word and those
1:39:21 clusters are clusters of
1:39:24 meaning so you might have the word dog
1:39:27 and in this cluster
1:39:30 here it's the the meaning of the word
1:39:33 dog like this is a cluster of like
1:39:35 animals and then you also have the word
1:39:38 dog up in this cluster and this cluster
1:39:40 is like that guy did his girl bad and
1:39:43 he's a dog right so same word very
1:39:46 different semantic
1:39:49 places this is one of the reasons why
1:39:52 when you say act like a Stanford
1:39:56 mathematician when you tell chat GPT to
1:39:58 act like a certain thing it actually
1:40:00 kind of points it in the direction of a
1:40:02 semantic cluster and it kind of
1:40:04 eliminates some of the stuff where there
1:40:06 might be false information that's why
1:40:08 that thing works okay so that's the
1:40:11 latent space this is
1:40:13 a highly highly
1:40:17 simplified two-dimensional this is just
1:40:19 an XY two dimensional space right it's
1:40:22 just a flat plane with these things the
1:40:24 real latent space is like a thousand
1:40:26 Dimensions it's impossible to comprehend
1:40:29 but and it's trillions of of points in
1:40:32 in this Laten space okay so whatever it
1:40:35 doesn't matter but just know that's the
1:40:37 latent space it's this mathematical
1:40:39 space where each dot represents um a
1:40:43 vector a number which is then associated
1:40:45 with a word
1:40:48 okay based on your prompt the large
1:40:51 language model predicts the next word
1:40:54 what the next word should be should it
1:40:56 should respond with over and over and
1:40:58 over and over and and this is why um I
1:41:01 don't know what your name was why your
1:41:03 husband is correct that it's basically
1:41:04 just autocomplete on steroids here's
1:41:07 here's how this works so if you type in
1:41:10 the prompt the quick quick brown fox
1:41:12 jumps over the lazy blank if you ever
1:41:14 took a typing class you know the word is
1:41:17 dog but what's what what actually
1:41:19 happens here well when you type that in
1:41:22 it creates um a mathematical probability
1:41:26 of what the next word should be based on
1:41:28 knowing all of this Laten space so what
1:41:31 it does is it kind of deselects all the
1:41:33 [ __ ] that's not right and it says okay
1:41:36 in in this cluster of words somewhere in
1:41:39 there is the highest probability
1:41:42 Vector the one that has the highest
1:41:44 probability is is the one it will go
1:41:46 with now there's a thing when you
1:41:49 program these called temperature where
1:41:50 you can make things more more or less
1:41:52 creative if you make it less creative
1:41:54 it'll just go with the highest
1:41:56 probability one if you make it more
1:41:58 creative it might select I don't know
1:42:00 eight or 10 of the words in here that
1:42:02 could be the next word and so it it'll
1:42:04 make it more creative but anyway so it
1:42:07 predicts that's the next most probable
1:42:09 word and if you look at what that point
1:42:11 is it's just this big long string of
1:42:13 numbers
1:42:16 0.123 comma minus. 0579 comma 2 eight
1:42:21 right so it's just this big long ass
1:42:24 string of numbers and then that long ass
1:42:26 string of numbers is tied to the word
1:42:28 dog and then the
1:42:30 system inserts the word dog into that
1:42:34 sentence that's how it
1:42:37 works it the large language model has
1:42:41 zero idea what that word
1:42:45 is or what all the other words it wrote
1:42:48 were all it knows is that it's just
1:42:50 autocompleting
1:42:52 one token after another so when it
1:42:54 writes you a whole four-page blog
1:42:57 article it has no idea what it wrote
1:43:00 none none none none
1:43:04 none it's just a dum dum now I think
1:43:07 that there are going to be and and I I
1:43:09 think these are being designed right now
1:43:12 there are going to be hybrid systems
1:43:14 that that that do kind of multiple
1:43:17 passes on this one one pass will be
1:43:19 it'll generate all the words then
1:43:21 there'll be another password it'll
1:43:22 analyze what it wrote and see if it got
1:43:25 it right and then it'll go back and
1:43:27 rewrite it and then read it and see if
1:43:29 it got it right and then it may do that
1:43:31 four or five times before you see
1:43:33 anything that's how we'll fix things
1:43:35 like hallucinations the reason these
1:43:37 things hallucinate they don't know what
1:43:39 they
1:43:40 wrote all they know is in this weird ass
1:43:43 semantic space there's a bunch of words
1:43:45 that seem like they should go in this
1:43:46 order and so that's why it will
1:43:48 confidently answer you even if it's
1:43:50 wrong
1:43:54 that's not the llm slide Kyle what's not
1:43:57 the llm
1:43:59 slide what you talking about
1:44:06 Willis what you don't trust us with your
1:44:09 credit card number I thought we were
1:44:17 friends oh that was from a while
1:44:20 ago
1:44:22 oh man isn't that
1:44:25 wild that's how chat jpt Works what time
1:44:28 is it is it time it's time it's getting
1:44:30 late people
1:44:32 um yeah that's how that's how that [ __ ]
1:44:35 works it's crazy isn't
1:44:40 it I when I fig when I when when I
1:44:44 figured out did I figure it out when I
1:44:46 learned how large language models
1:44:50 work I find them more mysterious than I
1:44:54 did when I didn't know how they worked
1:44:57 because I like many people assumed that
1:45:01 it was just kind of like a search engine
1:45:03 you've got all these documents in a
1:45:04 database and you type your prompt and it
1:45:07 sort of says let me go look at all these
1:45:10 documents and see which one's the right
1:45:11 document and then I'll return some chunk
1:45:13 of that to the user it's not what it's
1:45:17 doing when you put in a whole PDF
1:45:20 document in and you embed it it's what
1:45:22 it's it's called embedding when you turn
1:45:23 it into this Vector
1:45:26 space that original document ceases to
1:45:31 exist it turns all of the individual
1:45:34 words into numbers and scatters them
1:45:37 into Vector
1:45:38 space in semantic
1:45:44 clusters and then when you type your
1:45:46 prompt it just has some mathematical
1:45:48 probability that one of those things is
1:45:50 the next right token
1:45:52 word [ __ ] nuts isn't
1:46:00 it I asked Pi the other day I'll ask you
1:46:03 what is to stop someone evil from
1:46:05 programming chat
1:46:06 GPT well the only thing to stop them
1:46:10 from programming chat GPT do doing
1:46:13 something evil is you have the
1:46:16 corporation open AI saying they're not
1:46:18 going to and you have the Microsoft
1:46:22 saying yeah we invested 10 billion in
1:46:25 you and we're going to hold you to
1:46:26 account for that cuz we've got our
1:46:28 business to
1:46:29 protect so so you've got an economic
1:46:32 interest in the case of open AI if you
1:46:34 look at what meta released into the
1:46:36 world karmic fruit the Llama 2 open
1:46:42 source
1:46:44 full large language model with all the
1:46:47 weights they call it it's basically all
1:46:49 the secret sauce what put into the world
1:46:52 was a completely unrestricted
1:46:57 open-source foundational large language
1:46:59 model that's that's about the quality of
1:47:02 um GPT
1:47:05 3.5 that's in the hands of Open Source
1:47:08 developers right now you can install
1:47:10 today a non-safety guard railed version
1:47:14 of llama 2 on your slack Channel I know
1:47:17 a guy that did it I met him from in here
1:47:21 he said yeah I put one of these in my
1:47:23 slack channel it's [ __ ]
1:47:25 insane it cusses at them all the time it
1:47:29 it he said it it'll lose its mind and
1:47:31 it'll just start speaking Spanish for
1:47:33 like three days and no matter what they
1:47:36 ask no matter what they tell it it will
1:47:37 not not speak Spanish and then it'll
1:47:40 just come back and then it'll start
1:47:42 cussing at them for three days question
1:47:45 sensient versus Consciousness are AI
1:47:47 conscious I don't think so not right now
1:47:53 I I mean I think they'll program in that
1:47:55 kind of
1:48:06 behavior I I mean I honestly don't know
1:48:09 I I honestly don't know I don't know
1:48:11 enough about the science of it again so
1:48:13 cam what I what I would what I would
1:48:15 strongly recommend if you're if you're
1:48:17 interested in this the like the guy that
1:48:20 can expl expl this stuff in a way that I
1:48:22 think non-technical people can
1:48:24 understand is David Shapiro he explains
1:48:27 how these tools work and how they think
1:48:29 and he he explores these areas of
1:48:33 Consciousness and senscience and all
1:48:34 that sort of
1:48:36 stuff
1:48:38 um I mean just if you look at what I
1:48:40 just
1:48:42 did there's no way that chat GPT is
1:48:48 conscious it's a calculator
1:48:52 I mean it's literally no more conscious
1:48:54 in my opinion looking at how it
1:48:56 works than a
1:48:59 calculator but here's the weird here's
1:49:03 the weird
1:49:04 part I don't know that the mechanism of
1:49:07 chat GPT is conscious but what you
1:49:11 interact with as a
1:49:14 human is all the data that humans put on
1:49:18 the
1:49:19 internet and it happens to be coded in
1:49:22 such a way that what you're actually
1:49:25 interacting with is
1:49:27 the consciousness of humanity of the
1:49:32 past 40 years that stuck the [ __ ] on the
1:49:35 internet so my experience in working
1:49:38 with chat GPT is it's a lot closer to
1:49:44 interacting with a wicked smart very
1:49:48 simple intern
1:49:51 it's like the experience is like
1:49:53 interacting with a
1:49:54 human but technically what's happening
1:49:57 it's a calculator but I think I think
1:50:00 what blurs the line is what the data is
1:50:02 that it was trained on the p in GPT is
1:50:05 it what was it pre-trained on human
1:50:09 words and it and it's it's called a
1:50:12 large language model you know what else
1:50:13 is a large language model we
1:50:16 are so we've got this [ __ ] crazy
1:50:18 straw of a brain that when someone talks
1:50:21 [ __ ] into your ear your probability
1:50:24 engine goes here's the next most likely
1:50:26 words that are going to be the right
1:50:27 answer to
1:50:30 that we call that Consciousness so I I
1:50:33 mean I guess you could kind of go
1:50:36 there but it's just it's just a
1:50:39 calculator so I don't know I mean I and
1:50:42 and I also
1:50:44 think I think it's just semantics like
1:50:47 does it really [ __ ]
1:50:48 matter I mean I suppose if we get into
1:50:51 th the thing about do the AIS have
1:50:54 rights then it [ __ ]
1:50:57 matters but but it like I I don't think
1:50:59 it [ __ ] matters I think it's just and
1:51:01 why I think it don't ma it doesn't
1:51:03 matter is if this thing acts acts
1:51:07 sensient and acts
1:51:10 conscious
1:51:11 as you know compassionate and empathetic
1:51:15 as any human or
1:51:19 more how's that different than a human
1:51:22 doing it whatever the mechanism is ours
1:51:25 is sort of a wetwear version of it
1:51:27 theirs is a silicon version of it does
1:51:30 it really matter whether it's actually
1:51:32 conscious or sensient I don't know if
1:51:34 you get into ethics of do you turn it
1:51:36 off maybe if it has rights but I don't
1:51:39 know I don't know Way Way Beyond my
1:51:43 ability to comprehend right and
1:51:45 wrong can you make a GPT oh wait could
1:51:48 you repin that whoever just had that
1:51:49 pinned can you make a GPT that will
1:51:51 automatically check the internet and
1:51:54 send cease and deist
1:51:58 infringers yeah you could probably do
1:52:00 that you could probably do
1:52:03 that
1:52:07 um let
1:52:19 me
1:52:36 so I just wrote search the internet for
1:52:38 any mention of the word storyvine not on
1:52:40 story vine.com
1:52:46 yeah it just wait
1:52:54 it didn't find
1:52:56 anything I mean so so basically what I
1:52:58 would say is if you can do a search and
1:53:01 find an infringement you could make a
1:53:03 GPT that does the same search and then
1:53:06 has your cease and desist letter in it
1:53:08 and would customize the cease and desist
1:53:11 letter you could probably even get it to
1:53:13 if you could find the the the
1:53:17 infringements you could probably get it
1:53:18 to quote
1:53:20 the
1:53:22 infringement and even you know have
1:53:24 language around you know when you find
1:53:26 infringement quote it in the following
1:53:28 way cite this part of the
1:53:31 law question what is consciousness then
1:53:35 Kim kin from Cleveland I got an acting
1:53:37 degree not a Philosophy degree I don't
1:53:39 [ __ ] know if you put it into play I
1:53:42 can say the words have you tried grock
1:53:44 yes I have grock's a piece of
1:53:48 [ __ ] grock is conceptually really
1:53:50 interesting it's just a piece of [ __ ]
1:53:51 right
1:53:52 now dang it Kyle I got to sleep need to
1:53:55 be awake in 5 hour probably a potato
1:53:58 okay well listen I got to get out of
1:53:59 here too it's it's 10:30 I got to get
1:54:01 going um so let's do this let
1:54:05 me just
1:54:10 quickly um if you are trying to figure
1:54:14 this AI stuff out or if you're like the
1:54:16 one in your group that's talking about
1:54:18 Ai and everyone's like you're a weird
1:54:22 come check out the AI Salon this is a
1:54:24 group I started the week after chat GPT
1:54:27 launched it's a really powerful
1:54:29 Community Check out about the salon when
1:54:31 you get there the first link when you go
1:54:33 there this is a link tree the first link
1:54:35 is to our Mighty networks um Community
1:54:39 Mighty networks is just a Community
1:54:40 Management tool um it's like it's kind
1:54:42 of like Facebook it's but it it's it's
1:54:45 good somewhere between Discord and
1:54:47 Facebook is mighty Networks um and so
1:54:51 it's free it doesn't cost you anything
1:54:53 it's a it's a really cool group so go
1:54:55 there check that out um and what
1:54:59 resources and recommendations for llm
1:55:01 training experience and deep
1:55:05 [Laughter]
1:55:10 learning so Lily
1:55:15 if if you want to learn
1:55:18 about the the underpinnings of like how
1:55:22 the [ __ ] works and and like getting into
1:55:26 deep learning and machine learning and
1:55:27 programming and Engineering there's lots
1:55:30 and lots of courses on that if you want
1:55:33 to if you want to do courses on
1:55:35 generative AI like post chat GPT how to
1:55:39 use chat GPT 4 how to use these things
1:55:43 there are courses on that but what I
1:55:45 would strongly encourage you to do is
1:55:48 just play on your own
1:55:50 you can take a course but don't don't
1:55:53 put too much stock into anything that
1:55:55 you're going to learn right now cuz
1:55:57 anything that you're going to learn
1:55:58 right now that course was written 6
1:56:01 months ago and and seriously the amount
1:56:04 of progress that's been made in the past
1:56:06 6 months like the course you're going to
1:56:08 be learning will be valuable but it's
1:56:10 six-month old technology and then by the
1:56:12 time you go to apply it or 6 months in
1:56:14 the future you're going to have you know
1:56:16 9 months to a year
1:56:19 Gap you know between what you were
1:56:21 taught and what you're practicing and
1:56:23 where the technology is so if you want
1:56:26 to learn the backend stuff go take
1:56:28 courses from MIT Stamford Google meta
1:56:32 there there's a bunch of free courses
1:56:33 out there um I would go find those if
1:56:36 it's about large language model stuff I
1:56:40 would just keep showing up here show up
1:56:43 in groups that are you know doing Jam
1:56:46 sessions um Ju Just Keep throwing pasta
1:56:50 at the wall cuz the shit's moving too
1:56:52 fast to really there's no there's no
1:56:54 degree you can get in this right now all
1:56:57 right Show and Tell before I go so area
1:57:00 called showand tell and in Show and
1:57:04 Tell they do all sorts of crazy ass
1:57:08 [ __ ] oh that's really funny at the 1942
1:57:12 fair oh that's
1:57:13 [Laughter]
1:57:16 great 30 minute
1:57:18 goodbyes
1:57:21 yeah yeah that's just how my arms
1:57:27 look who wants to learn about
1:57:33 llms AI leaning
1:57:35 lab AI leaning lab is more better by the
1:57:39 way if if if you're frustrated that
1:57:43 these things can't spell right um enjoy
1:57:46 it while you can because in 6 months to
1:57:48 a year they're all going to be per
1:57:49 perfect and they're going to be boring
1:57:51 as [ __ ] we're going to long for the days
1:57:53 when it's spelled AI leaning lab son is
1:57:56 it really worth the 20 bucks a
1:58:02 month oh this is cool Sharon mendal is
1:58:04 putting up a bunch of suo
1:58:06 songs Gabrielle restaurant and bar very
1:58:09 cool those are
1:58:12 great nice
1:58:18 Barbara
1:58:21 there's a cure for
1:58:23 that just left the center
1:58:29 W oh man oh nice we got a bigger Arc I
1:58:33 like it we've upgraded that's
1:58:36 solid we're gonna need a bigger
1:58:40 Arc oh if you come back tomorrow night
1:58:43 I'll tell you why why there's all these
1:58:45 animal jokes these are great great great
1:58:48 great all right everybody listen I gotta
1:58:51 get out of here nice poop
1:58:55 Sundays
1:58:58 um phone day phone died I came back for
1:59:00 goodbyes thank you Joker I appreciate
1:59:02 that uh thanks to all the subscribers
1:59:05 thanks to everyone please follow this
1:59:06 channel um you know I want to grow I
1:59:09 want to grow the Channel I want to go
1:59:10 grow the community um um thanks to the
1:59:15 Irregulars people that show up here
1:59:17 regularly thank you for the mods you're
1:59:19 amazing anyone that came to the salon
1:59:21 tonight that was a good one thank you
1:59:22 really fun tonight um we're going to
1:59:25 need a larger language model that's
1:59:28 pretty good
1:59:30 um Mr it thank you so much Clint
1:59:33 Copperhead calling people to order thank
1:59:36 you sir Be not
1:59:38 Afraid good to see you all right let's
1:59:42 see V dancer Edgar thank you guys BT
1:59:45 comp KSU good to see
1:59:48 you nice uh thanks for your help tonight
1:59:51 at the salon by the way Papa JJ Vicky
1:59:53 thank you for your help at the salon
1:59:54 tonight by the way thank you very very
1:59:56 much I mean it I mean
2:00:00 it Lily good to see you Edgar V dancer
2:00:05 Corey Sandler Pottery thanks Pottery
2:00:07 crew for coming and checking Kyle out
2:00:09 yeah sorry if I was a disappointment
2:00:11 tonight to the pottery crew ah it's it
2:00:14 this Channel's like a box of chocolates
2:00:16 you just never know what you're going to
2:00:17 get great live thank you pry Tobias
2:00:21 thank you so much spin B3 California cam
2:00:25 catkin from Cleveland Mary Mary
2:00:28 steo all right let's see Erica Don Smith
2:00:33 Amelio's wife y'all are awesome Tyron
2:00:37 truth all right cool listen uh I'm out
2:00:40 of here it is Tuesday night tomorrow is
2:00:42 Wednesday uh if you haven't played with
2:00:44 chat GPT chat. open.com if you can
2:00:47 afford the 20 bucks spend the [ __ ] 20
2:00:50 bucks learn that [ __ ] um I didn't be
2:00:54 used to be this aggressive about it but
2:00:56 I just I feel like it is in everyone's
2:00:59 interest to get educated about what
2:01:01 these tools can do because 2024 shit's
2:01:05 going to get weird these tools are going
2:01:06 to get way more powerful way more
2:01:08 multimodal meaning pictures videos sound
2:01:11 all that stuff start playing now chat.
2:01:15 open.com go there now if you haven't
2:01:18 gone have your Kevin mallister
2:01:22 moment I'll see you tomorrow peace out