aiLL 10/31/23 - OpenAI Drops Massive AI Announcements: GPTs, Democratization of Specialization, More

Description

Chapters

Transcript

0:07 [Applause]
0:21 boom bill number one in I like it
0:26 overachiever Silver Fox was shaking
0:28 Sherry d
0:31 all right let's
0:33 see not much going on today anyone else
0:37 getting bored with this this whole
0:39 sector there just like it's like nothing
0:42 it's like it's totally
0:44 stagnated nothing
0:47 happening I mean in the last 32
0:53 seconds holy crap what a day
0:56 dag dag that was a day that was a day I
1:02 also had a lot of stuff at work so I
1:04 didn't get to play so what happened was
1:08 open aai had their first ever developers
1:11 conference which is surprising because
1:13 they've been around for a while and
1:14 they've primarily been targeting
1:16 developers until this past
1:19 year and they made some major
1:21 announcements which I will go over and
1:24 and uh Ted Ted made us this Time
1:26 Magazine cover that uh Time Magazine is
1:29 covering the aiar learning
1:31 lab and uh they they put out a new issue
1:35 today advertising that we're going to be
1:37 talking about all the new chat J Asos
1:40 they talked about this you can make
1:42 money with ch no seriously now you're
1:45 goingon to be able to make money with ch
1:47 gbt it's just that easy it's actually
1:50 pretty easy I I don't know what details
1:54 look like Clint Copperhead calling
1:55 everyone to order the uh the Irregulars
1:59 and the mods and things like that if
2:00 you're new here welcome welcome welcome
2:02 if you're not new here welcome welcome
2:04 welcome if you're an irregular you
2:10 weirdo the Irregulars here they claim
2:12 not to be weird they're just
2:15 irregular have you met
2:21 them so the Irregulars Thank You Be not
2:25 Afraid the Irregulars are people like
2:27 they reintroduced browse with Bing
2:32 um and then
2:35 today changed the game like it this is a
2:39 direct assault on
2:40 Claude the Claude
2:44 model it's cuckoo it's just
2:47 cuckoo Cortana started talking to me on
2:50 my phone it updated it was a bit
2:52 surprising what's
2:54 Cortana Dolly wish it was more like mid
2:57 Journey what that you have to do it in
3:00 do in
3:03 Discord do do you mean you wish the
3:05 quality was a bit more like mid Journey
3:07 cuz mid Journey can't do text and mid
3:08 Journey can't do generation Claud
3:12 doesn't sound as robotic as gp4 uh it's
3:15 going to be interesting to see um GPT 4
3:18 Turbo that they announced today is they
3:20 confirmed is GPT
3:22 4.5 so this is this is a this is a Next
3:26 Step Up that he's claiming that it's
3:28 better faster and and 2.75 five times
3:32 cheaper wait I just joined L what
3:35 happened today say what that's an
3:38 appropriate name hey say
3:42 what um what was that sorry there's
3:45 nothing to
3:46 repeat
3:52 Siri missed the news about the gpts
3:55 today okay um missed all the news about
3:58 the gpts today okay this is the real
4:02 deal wait there's no reason to subscribe
4:04 to anything else this is the real deal I
4:06 I agree with that I came for the therapy
4:08 tonight I spent half the day just
4:10 stunned need community so so lab five
4:13 Futures uh who's in here that's that's
4:16 uh um Cindy [ __ ] is going to be speaking
4:18 at the AI salon tomorrow night about the
4:23 future and she had prepared something
4:26 she had prepared something when on like
4:28 Sunday right she prepared something on
4:31 Sunday and then all this [ __ ] dropped
4:33 today and it just it
4:35 just holy crap the future's coming too
4:39 fast should I not tell you I made my
4:41 first G GPT oh jrc you should tell me
4:45 that but I'm so jealous you got a GPT I
4:48 didn't get access yet a although I made
4:52 an assistant in the playground and he's
4:55 a he's a 1970s muscle car
4:58 geek so no matter what you ask him he
5:01 turns he turns the conversation into a
5:04 conversation about a specific 70s muscle
5:07 car so I'll show you how to make
5:09 that spill the tea all right all right
5:12 I'll do it okay so so okay Where Do We
5:15 Begin if you're brand new here we're
5:18 going to go geeking out a little bit um
5:21 and I'm I'm just going to talk mostly
5:23 because I can't show a whole ton um but
5:26 but let's let's let's flip over and
5:28 start show and telling some stuff so
5:30 I'll get back to that well actually here
5:32 let me just zoom in and get all that
5:35 [ __ ] down there okay if you're brand
5:37 brand new
5:39 to um AI just go go there just go there
5:44 right now chat.
5:45 open.com nothing else really even
5:48 matters right now
5:51 um you likely won't experience the brand
5:54 brand new stuff they announced today cuz
5:56 I don't have it a lot of people don't
5:58 have it yet it's going to be rolling out
6:00 out probably over the probably over this
6:01 week today's
6:04 Monday
6:05 um but just go there and start playing
6:08 now if you've been here so so couple of
6:12 things happened today
6:15 um chat open AI had their first ever
6:20 developer conference and they had a big
6:23 keynote
6:24 presentation uh that talked about all
6:26 the stuff that they were releasing and
6:28 it's a lot it's a lot someone prior to
6:32 this over in the Discord the salon
6:34 Discord um asked about could I do a
6:38 um a vocabulary primer to get started
6:43 for for people that aren't Geeks so so I
6:46 thought I would I thought I would
6:48 actually split up what happened today
6:51 into two columns um chat GPT so so this
6:55 is one of the most confusing things
6:56 about this whole world right now there's
6:59 's a thing called GPT playground or open
7:02 AI playground that's where developers go
7:05 to make applications like chat GPT chat
7:10 GPT is an application built on top of
7:13 the developer playground built on top of
7:15 the apis it just happens to be made by
7:18 open AI okay so they made major
7:21 announcements in both Arenas okay they
7:25 made major announcements in chat GPT and
7:28 they made major announcements in the
7:29 developer side so I'll go over them both
7:32 because anything that was announced in
7:33 the developer side because they made it
7:37 so easy and there's so much there we're
7:39 going to start very very quickly seeing
7:41 applications that start delivering a lot
7:43 of the stuff on the developer side and
7:46 all of that stuff is also going to
7:48 likely end up in the chat GPT side the
7:50 consumer side so just think of it as
7:52 chat GPT is for the rest of us the
7:55 developer stuff's for developers all
7:56 right the developer stuff's been there
7:58 for I don't know four 5 years it's been
8:00 there for a while um but they just made
8:02 the apis really slick really really
8:05 slick all right are you sure you don't
8:09 have it Kyle it's in the upper left in
8:12 the four
8:14 dots in what in in the app in the IOS
8:25 app or
8:27 in uh uh uh uh or in the web app let me
8:32 go to the web
8:36 app I just looked at it right before I
8:38 came on here
8:42 reload there's no dots in the upper left
8:44 I don't know what you're talking
8:46 about chat J see I've still got see how
8:50 I've still got five modes that means I I
8:53 do not have so the when you have the
8:57 single mode stuff that icon turns from a
8:59 little picture to like this little open
9:01 AI paperclip upload looking thing so I
9:05 ain't got it man I ain't got it man
9:08 listen listen listen listen you can make
9:12 money
9:14 with okay all right let's go through
9:17 this crap I mean this fantastic
9:21 news okay
9:24 so if you want to watch the video
9:31 it's go to open AI just search for open
9:34 AI on
9:38 YouTube you got to actually finish
9:40 typing
9:41 it and it should be the first
9:46 thing open AI
9:54 keyote here we
9:58 go so yeah just go to YouTube and it's
10:02 called opening open AI Dev day opening
10:07 keynote all right so you you you you can
10:11 and should go watch it I'll I'll play
10:14 some pieces of
10:15 it but it's so here's the deal it it was
10:20 so I live streamed it today I did one of
10:22 these Ticky takies and we watch Sam do
10:24 his thing and it was just like one thing
10:28 after another where everyone on the chat
10:30 was like oh my God oh my God people were
10:33 crying Sherry D was like I should why am
10:35 I crying why am I crying it it was oddly
10:38 emotional there was something really
10:40 amazing about watching that live stream
10:43 with a group who's been talking about
10:46 this stuff for so long so a lot of the
10:48 people that come here night after night
10:50 they've been figuring this [ __ ] out and
10:52 they've been figuring out where the
10:53 boundaries of of chat GPT is and and we
10:56 all kind of know like what we would like
10:57 it to do and a lot of what was announced
11:01 today just like knock down barriers
11:04 knock down barriers and it's just the
11:06 the amount of power that's being put in
11:09 our hands in fact I think one of the
11:11 things Alman says at the beginning here
11:14 let me see if I if he says it really
11:16 soon thank you for joining us today
11:18 please welcome to the stage Sam
11:21 [Music]
11:28 Alman
11:31 good morning Welcome to our first ever
11:33 open AI Dev day we're thrilled that
11:35 you're here and this energy is
11:42 awesome and welcome to San Francisco San
11:45 Francisco has been our home since day
11:46 one the city is important to us and the
11:49 tech industry in general we're looking
11:51 forward to continuing to grow here so
11:54 we've got some great stuff to announce
11:56 today but first I'd like to take a
11:58 minute to talk about some of the stuff
12:00 that we've done over the past year about
12:03 a year I don't know what I I don't know
12:04 when he says it but he he says this
12:06 thing about like he he wants to put like
12:08 magic in people's hands he wants to give
12:11 them power I went to see what a 300 page
12:15 prompt looks
12:17 like I want to see well so you know
12:19 brother 52 here's what a 300 page prompt
12:22 looks like you take a novel that you
12:25 wrote and you want to have it edited or
12:28 you want to write an epilogue for it or
12:29 a prologue for it you paste in your 300
12:32 Page novel and and then you start
12:35 interacting with it um there's so many
12:39 things there's so many things I was
12:40 trying to get Dolly 3 to set a
12:42 Thanksgiving table for 12 it kept
12:44 miscounting oh Becky yeah asking these
12:47 things to do any counting is is a
12:50 disaster right now that'll get better
12:51 but it ain't right now it ain't it ain't
12:55 it ain't all right all right pap what do
12:59 we got here I mean on the live actually
13:01 here is on the
13:03 live on Tic Tac anyway okay all
13:07 right all right let's get
13:10 going
13:12 okay so so
13:15 the they they split these up a little
13:17 bit today what I tried to do was was I
13:20 tried to put stuff that's definitely
13:22 going to be in chat chpt so this is so
13:25 this column is the stuff you pay 20
13:27 bucks a month for this column is if
13:30 you're paying for API access basically
13:33 which you can set your own limit for
13:34 that I set my limit and it's a different
13:37 it's a different bucket of money
13:38 basically but they only charge you for
13:41 the tokens you use so I don't think I've
13:43 had a bill on this side more than $9 and
13:45 that was when I was pretty busy and they
13:48 just made things cheaper so so that's
13:50 kind of how I split things up here okay
13:53 so for Chachi BT um he confirmed today
13:58 it's going to it's going to go true
14:00 multimodal uh
14:01 meaning you're not going to have what I
14:04 have right
14:06 now which are these five modes where
14:10 you've got you can upload images here
14:13 this is connected to the internet here
14:15 this does Advanced Data analysis here I
14:17 think plugins they didn't mention
14:19 plugins today plugins have essentially
14:20 been replaced with gpts which is a
14:23 brilliant [ __ ] name I'll talk about
14:25 that and then Dolly 3 all of these are
14:27 going to get rolled into the single the
14:29 single line so you'll just ask it to do
14:31 [ __ ] and it's going to do
14:34 [ __ ]
14:36 um they confirmed that there's an
14:38 updated April 2023 cutoffs and he you
14:42 know Altman said he was embarrassed that
14:46 the the most recent um data in in GPT 4
14:50 was September 2021 and he said it's now
14:53 April 2023 and they're going to try to
14:55 improve that constantly it's a pretty
14:57 big deal to update those models
15:00 um so there's that um I am not clear it
15:05 it was it it might have been clear but I
15:07 I missed it or forgot
15:09 it GPT 4
15:12 Turbo has
15:15 128,000 token context window which is
15:18 bigger than claws what was not clear to
15:20 me is is if that's coming to chat GPT
15:23 right away or not so so that I that I
15:25 missed I don't know I think I think it
15:27 is I think it has but I'm not sure the
15:31 the big announcement on the consumer
15:34 side on on the chat GPT side are these
15:36 things called custom gpts or just
15:40 gpts so gpts are just like Poots we've
15:44 we've talked about po.com here before
15:46 and and a and a bunch of people a bunch
15:48 of the Irregulars have been off building
15:50 Poots turns out that was that was a
15:52 really good use of their
15:54 time because they're not calling them
15:57 Bots they're calling them GP pts I think
16:00 that is like a stroke of Genius from a
16:03 branding standpoint here's
16:05 why chat Bots people the tech industry
16:09 has been talking about chat Bots for 15
16:11 or 20 years they have just sucked for 15
16:14 or 20 years so if you call it a bot it's
16:16 immediately in a category of tech that
16:18 everyone already hates plus Hollywood's
16:21 told us that the robots are going to
16:23 kill us so if they called it Bots that
16:25 People's Natural fear is you know
16:29 gonna bias them toward hating these
16:32 things by calling them
16:35 gpts they are instantly defining chat
16:39 GPT
16:41 gpts as a
16:43 verb let me GPT that let me Google that
16:47 make me a Xerox hand pass me a Kleenex
16:51 right all these brand names that became
16:53 verbs in in you know popular vernacular
16:57 that's what they're going to do that's
16:58 what G PT is going to be it's and it's
17:00 going to be instant cuz there's going to
17:02 be thousands of these things they made
17:04 these so easy to make so so I can what I
17:08 can show you is I might even be able to
17:11 have someone well I don't know if
17:13 anyone's set up to show can you talk
17:15 through the process of how you do the
17:17 API what you
17:20 use
17:22 versal wait I can't see what that
17:26 says oh versal um
17:30 I don't when I'm doing the API stuff
17:31 I've just been using zapier I'm not
17:33 doing I'm not doing coding my my
17:35 Engineers are are accessing some of the
17:38 API stuff but I'll show you where it is
17:40 we'll go we'll go over there we'll go
17:42 over there and
17:43 play um so so when you go to
17:48 PO let me go to PO here for a
17:52 second po.com
17:56 log
17:57 in I going to log in it's going to be so
18:00 good
18:03 mhm so in po you can create a bot okay
18:07 so gpts are essentially
18:10 that and when you get in here you you
18:13 have to name it so that's the same and
18:15 then you pick a model and I don't think
18:17 there'll be a lot of models at at GPT I
18:20 think it'll just be gp4
18:22 turbo oh I just saw breaking news we
18:25 work valued at 47 billion files for
18:28 bankrupt oops see there's no big
18:31 surprise okay then you type in a prompt
18:34 and then you type in greeting messages
18:36 and things like that and that that's
18:37 pretty much it so but but you really
18:40 have to write your own
18:41 prompt
18:43 the the GPT maker at chat
18:48 GPT
18:49 um does it for you so in fact we can go
18:53 watch I can go find that part of the the
18:57 video
19:04 uh oh that's tokens wait
19:10 where I haven't quite memorized where
19:13 all this [ __ ]
19:14 happens oh I think it's at the
19:19 beginning here this is where he says
19:21 what he wants to do for
19:23 [Applause]
19:25 people okay so we have shared a lot of
19:29 great updates for developers already and
19:31 we got a lot more to come but even
19:33 though this is a developer conference we
19:35 can't resist making some improvements to
19:37 chat GPT
19:38 yay so a small one chat GPT now uses GPT
19:42 4 Turbo with all the latest improv
19:45 including the latest knowledge cuto
19:46 which will continue update that's all
19:48 live today it can now browse the web
19:51 when it needs to write and run code
19:53 analyze data take and generate images
19:55 and much more and we heard your feedback
19:57 that model picker extremely annoying
19:59 that is gone starting today you will not
20:01 have to click around the drop down menu
20:03 all of this will just work together chat
20:05 gbt yeah y okay so so that was that so
20:08 now he goes and he demos okay so let's
20:11 here he is talking
20:13 about the future of Agents important to
20:16 move carefully towards this future of
20:18 Agents it's going to require a lot of
20:20 technical work and a lot of thoughtful
20:22 consideration by
20:24 Society So today we're taking our first
20:27 small step that moves us towards this
20:29 future we're thrilled to uh we're
20:32 thrilled to
20:33 introduce gpts gpts gpts are tailored
20:37 versions of chat GPT for a specific
20:40 purpose you can build a GPT a customized
20:44 version of chat GPT for almost anything
20:46 with instructions expanded knowledge and
20:49 actions okay with
20:53 instructions expanded knowledge and
20:57 actions expanded knowledge and actions
21:01 expanded knowledge and
21:03 actions those are he sort of says them
21:06 in passing they're a huge [ __ ] deal
21:08 they're a huge deal so right now with um
21:13 with a poot you basically just have your
21:15 prompt and whatever you put in your
21:17 prompt um but but you'll see what you
21:20 can do with these
21:22 gpts is you can use code interpreter you
21:26 can do a thing called retrieval which is
21:29 um uh retrieval augmented generation
21:33 where you can upload documents to this
21:36 to this
21:37 GPT um so so let's say you've got a
21:40 particular kind of uh sales Consulting
21:43 that you do you can take your documents
21:45 from your sales Consulting business and
21:48 build that into the bot and so not only
21:50 will it um do its normal chat GPT kind
21:54 of you know consulting stuff it'll pull
21:56 from your documents pull
21:59 relevant data from your
22:03 documents um and then taking actions you
22:06 can actually connect this to other
22:08 things in the world you publish it for
22:10 others to use and then you can publish
22:11 it for others to use now po does that
22:13 right
22:14 now but you're going to now be able to
22:18 go from chat GPT to making a custom chat
22:20 GPT to publishing that so think about
22:23 that for publishing that inside your
22:25 business think about publishing that for
22:27 your friends and family you you make the
22:29 family reunion bot and anyone in your
22:33 family can go use that and get all the
22:34 information about the family reunion or
22:38 you could publish a bot to start to
22:40 monetize it you could make money with
22:46 Jing it's
22:50 coming and because they combine
22:52 instructions expanded knowledge and
22:54 actions they can be more helpful to you
22:57 they can work better in any context and
22:59 they can give you better
23:00 control they'll make it easier for you
23:02 to accomplish all sorts of tasks or just
23:04 have more fun and you'll be able to use
23:06 them right within um madcap does that
23:10 mean buy zappier No in fact one of the
23:12 things that he shows here is is a
23:14 zappier bot that that can can now take
23:17 actions over to zappier I think I think
23:20 this amplifies zapier um in in a in a
23:24 pretty big way be because of all the
23:25 stuff they just announced with on the
23:27 API side this um the zappier calls are
23:31 going to get much fancier you're going
23:33 to be able to build some really
23:34 sophisticated apps either through zapier
23:37 but a lot of the apps that I would have
23:39 built in zapier I I'll now just be able
23:42 to do directly in we we will all be able
23:44 to do just directly in in chat GPT in
23:46 chat
23:47 GPT you can in effect program a GPT with
23:51 language just by talking to it it's easy
23:54 to customize the behavior so that it
23:55 fits what you want this makes building
23:57 them very accessible
23:59 and it gives agency to
24:01 everyone so it gives agency to
24:04 everyone this is this thing this is the
24:07 thing alman's really about is he wants
24:10 to
24:12 democratize
24:15 specialized
24:17 knowledge this gives agency to everyone
24:19 what's that mean it means non-developers
24:22 are going to be able to make incredibly
24:24 sophisticated
24:26 applications without having to know any
24:31 programming we're going to show you what
24:33 gpts are how to use them how to build
24:36 them and then we're going to talk about
24:38 how they'll be distributed and
24:39 discovered and then after that for
24:41 developers we're going to show you how
24:43 to build these agent like experiences
24:45 into your own apps so first let's look
24:48 at a few
24:49 examples our partners at code.org are
24:52 working hard to expand computer science
24:54 in in schools they've got a curriculum
24:57 that is used by tens of millions of
24:58 students worldwide code.org crafted
25:02 lesson planner GPT to help teachers
25:04 provide a more engaging experience for
25:06 middle
25:07 schoolers if a teacher asks it to
25:09 explain for Loops in a creative way it
25:11 does just that in this case it'll do it
25:14 in terms of a video game character
25:16 repeatedly picking up coins super easy
25:18 to understand for an eighth
25:19 grader as you can see this GPT brings
25:22 together code.org extensive curriculum
25:25 and expertise and lets teachers adapt it
25:27 to their needs quickly and
25:29 easily next wait wait who just said that
25:32 who just said that um Vic Van blonde so
25:37 so terrorists can can also use it well
25:39 sure I mean that's this is one of this
25:42 is one of the reasons that people are
25:44 wondering about you know safety guard
25:46 rails and things like that um yes they
25:49 can um although terrorists won't be able
25:53 to um publish their naughty gpts in the
25:57 GPT store that he's about to
26:00 announce check this one
26:02 out canva has built a GPT canva it lets
26:05 you start designing by describing what
26:07 you want in natural language if you say
26:10 make a poster for de a Dev dat reception
26:12 this afternoon this evening and you give
26:14 it some
26:15 details it'll generate a few options to
26:17 start with by hitting canvas apis now
26:20 this concept may be familiar to some of
26:21 you we've evolved our plugins to be
26:24 custom actions for gpts wait wait we've
26:27 evolved our plugins to be custom actions
26:30 for gpts so you're now going to be able
26:32 to create a b who just be Not Afraid
26:34 create a GPT oh coming soon Oh I thought
26:37 you had access to it dang
26:40 it
26:41 um we've evolved our plugins to become
26:44 actions within gpts so you're going to
26:46 be able to make these
26:47 Bots that that call plugins right that
26:51 call actions so so an action is
26:53 basically calling through your Bot to
26:57 some other service
26:58 right so out to zapier out to canva out
27:01 to you know other things out to your
27:04 company's data
27:06 set you can keep chatting with this to
27:08 see different iterations and when you
27:10 see one you like you can click through
27:12 to canva for the full design
27:15 experience so now we'd like to show you
27:18 a GPT live zapier has built a GPT that
27:22 that lets you perform actions across
27:24 6,000 applications to unlock all kinds
27:26 of integration possibilities
27:28 I'd like to introduce Jessica one of our
27:30 Solutions
27:32 Architects Kyle didn't he say he was
27:34 adding security who
27:37 Sam today he may have I don't remember
27:40 what he said about it who's going to
27:41 drive this demo welcome
27:43 Jessica thank you s it's cool
27:46 demo thank you
27:50 all thank you all for being here my name
27:53 is Jessica Shay I work with partners and
27:55 customers to bring their product alive
27:57 hey I know
27:58 this going to be annoying but since we
28:00 have a bunch of people in there hey
28:01 everyone if you're new here my name is
28:03 Kyle Shannon this is the AI learning lab
28:05 we are we are on a deep dive about open
28:07 ai's um Dev day announcement today their
28:10 keynote announcement today um so I'm
28:12 going to just keep digging through what
28:14 they what they talked about and
28:15 explaining what I what I think are the
28:17 the important things and explaining ask
28:18 answering any questions do me a favor if
28:21 you would share this live since since
28:23 we're going through this as many people
28:25 can see this uh I would appreciate that
28:27 so if you'd share the live and then also
28:29 this is this is totally selfish and
28:31 arbitrary if you like what's going on
28:33 here follow the channel follow this the
28:35 AI learning lab Channel um cuz I'm like
28:38 350 people away from 30,000 and and
28:42 November 30th is the one-year
28:43 anniversary of chat gbt so the OCD and
28:46 me just wants to wants to do that so oh
28:50 somebody gave Deborah A Tik Tok hat that
28:52 was nice
28:54 um wait was it name Deborah no it wasn't
28:57 was it what's their
29:00 [Music]
29:04 name thank you all for being here my
29:06 name is Jess I work with Partners oh
29:09 thank you very much be Not Afraid thank
29:10 you very much appreciate the gifts I
29:12 can't wait to show you how hard we've
29:13 been working on this hey Joseph Baker's
29:15 happen so to start where your gbt will
29:18 live is on this upper left corner I'm
29:21 going to start with clicking on the zaap
29:23 your AI actions and on the right hand
29:26 side you so so I guess these will be
29:31 your open AI Dev day oh yeah that's
29:35 that's that's a
29:37 GPT so I so I guess I guess you'll
29:41 you'll have a specific sort of GPT
29:45 bucket that's SE separate from your prom
29:47 prompt history you can see that's my
29:48 calendar for today so it's quite a day
29:52 I've already used this before so it's
29:54 actually already connected to my
29:55 calendar to start I can ask
29:58 what's on my schedule for today we built
30:01 gpts with security in mind so before it
30:04 performs any action or share data it
30:07 will ask for your permission so right
30:09 here I'm going to say allowed so Drew
30:12 learning AI can I summarize this yeah
30:18 so I I'll let her play it but but
30:21 basically she's got this
30:24 zappier chatbot this this GPT
30:28 it's connected to things like her
30:30 calendar and to some some other actions
30:32 so she can just in her bot go what do I
30:35 have going on today and it will go look
30:37 at her calendar so that's the taking
30:39 actions part so because it's zapier and
30:42 zapier has ACT has access to 6,000
30:45 different services from Salesforce to
30:48 Discord to slack to Google Calendar
30:51 things like that the zappier GPT in
30:55 particular is going to be really really
30:58 powerful because you can just ask it
31:00 [ __ ] and if you've got it connected to a
31:02 bunch of their different Services you
31:04 know you would be able to do something
31:06 like if you had Z zappier hooked up to
31:09 RSS feeds of the latest news in AI you
31:14 could just say you know in this bot
31:16 what's the latest news and it would just
31:18 go grab it suck it in there post it to
31:21 your slack Channel things like that so
31:23 so she's just sort of demo demoing what
31:25 a single GPT can be if it's connected to
31:28 other services and because zapier can
31:30 connect to 6,000 it's a good one to to
31:33 demo so gbt is designed to take in your
31:37 instructions make the decision on which
31:40 capability to call to perform that
31:42 action and then execute that for you so
31:44 you can see right here it's already
31:46 connected to my calendar it pulls into
31:49 my my information and then I've also
31:52 prompted it to identify conflicts on my
31:54 calendar so you can see right here it
31:57 actually was able able to identify that
31:59 so it looks like I have something coming
32:01 up so what if I want to let Sam know
32:04 that I have to leave early so right here
32:05 I say Let Sam no I got to go um chasing
32:12 gpus so with that I'm GNA swap to my
32:17 conversation with Sam and then I'm going
32:20 to say yes please run
32:24 that Sam did you get that
32:28 I
32:29 did
32:32 awesome
32:34 so this is only a glimpse of what is
32:37 possible and I cannot wait to see what
32:39 you all will build thank you and back to
32:41 you Sam yep
32:45 so that's that now he's going to build
32:48 one thank thank you Jessica so those are
32:51 three great examples in addition to
32:54 these there are many more kinds of gpts
32:56 that people are creating and many many
32:58 more that will be created
33:00 soon we know that many people who want
33:02 to build the GPT don't know how to
33:04 code we've made it so that you can
33:07 program the GPT just by having a
33:10 conversation we believe that natural
33:11 language is going to be a big part of
33:13 how people use computers in the future
33:15 and we think this is an interesting
33:16 early example so I'd like to show you
33:18 how to build one now this is very
33:23 cool all right so I want to create a GPT
33:27 uh that helps give Founders and
33:28 developers advice when starting new
33:30 projects
33:32 um I'm going to go to create a GPT here
33:35 and this drops me into the GPT Builder
33:38 uh I worked with Founders for years at
33:40 YC and still whenever I meet developers
33:42 the questions I get are always about how
33:44 do I you know think about a business
33:45 idea can you give me some advice I'm
33:47 gonna let this run I'll be right back
33:48 see if I can build a GPT to help with
33:49 that so to start GPT Builder asks me
33:53 what I want to make and I'm going to say
33:55 I want to help startup Founders think
33:59 through their business
34:02 ideas and get advice after the founder
34:08 has gotten some
34:10 advice uh Grill
34:13 them on why they are not growing
34:19 faster all right so to start off I just
34:21 tell the GPT a little bit about about
34:23 what I want here and it's going to go
34:24 off and start thinking about that and
34:27 it's going to write some detailed
34:28 instructions for the GPT um it's also
34:30 going to let's see ask me about a name
34:33 how do I feel about startup Mentor
34:35 that's fine uh that's good so if I
34:38 didn't like the name of course I could
34:39 call it something else but it's you know
34:41 going to try to have this conversation
34:42 with me and start there and you can see
34:45 here on uh on on the right in the
34:47 preview mode that it's already starting
34:49 to fill out the GPT um where it says
34:52 what it does it has some like ideas of
34:55 additional questions that I could ask um
34:59 and you know what I so I just generated
35:01 a candidate of course I could regenerate
35:03 that or change it but I sort of like
35:05 that so I will say that's
35:10 great and you see now that the GPT is
35:13 being built out a little bit more as we
35:15 go now what I want this to do um how it
35:18 can interact with users I could talk
35:20 about style here but what I'm going to
35:22 say uh is I am going to upload
35:26 transcript
35:28 of some lectures about startups I have
35:31 given please give advice based off of
35:36 those all right so now uh it's going to
35:41 go figure out how to do that and I would
35:42 like to show you the configure tab so
35:44 you can see some of the things that were
35:46 built out here as we were going um by by
35:48 the Builder itself and you can see that
35:50 there's capabilities here that I can
35:51 enable um I could add custom actions
35:54 these are all fine to leave um I'm going
35:56 to upload a file uh so here is a lecture
36:00 that I picked that I used to that I gave
36:03 with some startup advice um and I'm
36:04 going to add that here in terms of these
36:06 questions uh this is a dumb
36:09 one the rest of those are reasonable uh
36:13 and like very much things Founders often
36:14 ask um I'm going to add one more thing
36:16 to the instructions here which is be
36:19 concise and constructive with
36:23 feedback all right so again if we had
36:26 more time I'd show you a bunch of other
36:27 things but this is uh this is like a
36:30 decent start and now uh we can try it
36:33 out over on this preview tab so I will
36:35 say um what's a common question what are
36:39 three things to look oops what are three
36:43 things to look for when hiring employees
36:46 at an early stage
36:51 startup now it's going to look at that
36:53 document I uploaded um it'll also have
36:55 of course all of the background
36:56 knowledge of GPT
37:01 4 that's pretty good those are three
37:03 things that I definitely have said many
37:04 times um now we could go on and it would
37:07 start following the other instructions
37:09 and you know Grill me on why I'm not
37:10 growing faster but in the interest of
37:12 time I'm going to skip that uh I'm going
37:14 to publish this only to me for now uh I
37:18 can work on it later I can add more
37:19 content I can add a few actions that I
37:21 think would be useful um and then I can
37:23 share it publicly so that's what it
37:25 looks like to create a GPT
37:28 with thank
37:34 you by the way I always I always wanted
37:37 to do that after like all of the YC
37:39 office hours I always thought man
37:40 someday I'll be able to make a bot that
37:41 will do this and that'll be
37:44 awesome so with gpts we're letting
37:47 people easily share and discover all the
37:49 fun ways that they use chat GPT with the
37:53 world you can make private gpts like I
37:55 just did or you can share your Creations
37:59 publicly with a link for anyone to
38:01 use or if you're on chat GPT Enterprise
38:05 you can make gpts just for your
38:08 company and later this month we're going
38:11 to launch the GPT
38:13 store you can list a GP thank you I
38:17 appreciate
38:23 that you can list a GPT there and we be
38:26 able to feature the best and the most
38:27 popular gpts of course we'll make sure
38:30 that gpts in the store follow our
38:32 policies before they're
38:35 accessible Revenue sharing is important
38:37 to us we're going to pay people who
38:40 build the most useful and the most used
38:42 gpts a portion of our Revenue we're
38:45 excited to foster a vibrant ecosystem
38:47 with the GPT store just from what we've
38:49 been building ourselves over the weekend
38:51 we're confident there's going to be a
38:52 lot of great stuff we're excited to
38:53 share more information
38:55 soon so those are gpts and we can't wait
38:58 to see what you'll
39:00 build but this is a developer conference
39:02 and the coolest thing about this is that
39:04 we're bringing the same concept to the
39:12 API many of you have already been
39:15 building agent-like experiences on the
39:18 API for example Shopify sidekick which
39:22 lets you take actions on the platform
39:23 discords Clyde lets Discord moderators
39:26 create Custom Custom personalities for
39:29 and snaps my AI a customized chatbot
39:32 that can be added to group chats and
39:33 make recommendations these experiences
39:36 are great but they have been hard to
39:38 build sometimes taking months teams of
39:41 dozens of Engineers there's a lot to
39:43 handle to make this custom assistant
39:46 experience So today we're making that a
39:48 lot easier with our new assistance
39:55 API the assistance API includes
39:58 persistent threads so they don't have to
40:00 figure out how to deal with long
40:01 conversation history built-in
40:04 retrieval code interpreter a working
40:07 python interpreter in a sandbox
40:09 environment and of course the improved
40:11 function calling that we talked about
40:14 earlier so we'd like to show you a demo
40:16 of how this works and here is Raman our
40:19 head of developer experience
40:21 [Music]
40:23 welcome thank you son good morning
40:28 wow it's fantastic to see you all
40:31 here it's been so inspiring to see so
40:34 many of you infusing AI into your apps
40:37 today we're launching new modalities in
40:39 the API but we also very excited to
40:42 improve the developer experience for you
40:44 all to build assistive agents so let's
40:47 Dive Right
40:48 In imagine I'm building one delas the
40:51 travel app for Global explorers and this
40:53 is the landing page I've actually used
40:55 gp4 to come up with with these
40:57 destination ideas and for those of you
40:59 with the Keen ey these illustrations are
41:01 generated programmatically using the new
41:03 darly 3 API available to all of you
41:05 today so it's pretty
41:08 remarkable but let's enhance this app by
41:11 adding a very simple assistant to it
41:14 this is the screen we're going to come
41:15 back to it in a second first I'm going
41:17 to switch over to the new assistant
41:19 playground creating an assistant is easy
41:21 you just give it a name some initial
41:24 instructions a model in this case I'll
41:26 Pi GPT for Turbo and here I'll also go
41:29 ahead and select some tools I'll turn on
41:31 code interpreter and retrieval and save
41:34 and that's it our assistant is ready to
41:37 go next I can integrate with two New
41:40 Primitives of this assistant API threads
41:43 and messages let's take a quick look at
41:45 the
41:46 code the process here is very simple for
41:49 each new user I will create a new thread
41:52 and as these users engage with their
41:54 assistant I will add their messages to
41:56 this threads very simple and then I can
41:59 simply run the assistant at any time to
42:02 stream the responses back to the app so
42:05 we can return to the app and try that in
42:07 action if I say hey let's go to
42:13 Paris all right that's it with just a
42:16 few lines of code users can now have a
42:18 very specialized assistant right inside
42:21 the
42:21 app and I'd like to highlight one of my
42:24 favorite features here function cooling
42:26 if if you have not used it yet function
42:28 calling is really powerful and as Sam
42:30 mentioned we're taking it a step further
42:32 today it now guarantees the Json output
42:35 with no Ed latency and you can invoke
42:38 multiple functions at once for the first
42:40 time so here if I carry on and say hey
42:43 what are the top 10 things to
42:47 do when I have the assistant respond to
42:50 that
42:51 again and here what's interesting is
42:53 that the assistant knows about functions
42:55 including those to annotate the the map
42:57 that you see on the right and so now all
42:59 of these pins are dropping in real time
43:02 here yeah it's pretty
43:07 cool and that integration allows our
43:10 natural language interface to interact
43:12 fluidly with components and features of
43:14 our app and it truly showcases now the
43:17 harmony you can build between Ai and UI
43:20 where the assistant is actually taking
43:23 action but next next let's talk about
43:25 retrieval and retrieval is about giving
43:27 our assistant more knowledge Beyond
43:30 these immediate user messages in fact I
43:32 got inspired and I already booked my
43:34 tickets to uh to Paris so I'm just GNA
43:36 drag and drop here this PDF okaying I
43:39 can sneak PE minor emergency with the
43:42 doggy everything's good I'm back okay
43:46 it's I'm unclear what happened okay so
43:50 um pop any questions you have down below
43:53 let me let me go back to where I was um
43:56 let's see the building the GPT thing was
43:59 pretty
44:00 amazing
44:07 um all right am I the only one that's
44:09 totally lost probably because I let it
44:11 sort of bleed into developer stuff so
44:13 let me go back let me go back and talk
44:19 about the
44:21 gpts all right so this was
44:24 him building the thing
44:31 oh you were flying to
44:33 Paris um okay so the couple of things
44:37 that I wanted to talk about here Grill
44:40 them on why they are not growing
44:45 faster all right so to start off I just
44:48 tell the GPT a little bit about about
44:49 what I want here and it's going to go
44:51 off and start thinking about that and
44:53 it's going to write some detailed
44:54 instructions for the GPT um it's also
44:57 going to let's see all right so so so
45:01 the fact that you can basically just
45:02 have chat GPT build the bot is a big
45:04 deal right um someone just posted what
45:08 what's elon's bot supposed to do what
45:11 elon's AI is going to do is the primary
45:15 thing differentiator right now is it's
45:17 going to be connected to Twitter it's
45:20 going to be connected to Twitter
45:22 data so wait you said you paid $9 not
45:26 $20 how okay no I paid there's two
45:29 different things so
45:32 on this
45:36 chart these are two
45:39 different areas of open AI chat GPT is
45:43 like the consumer front end the
45:46 developer playground where you actually
45:48 build applications is a separate fee
45:51 that's the one I paid like no more than
45:52 $9 on so it's 20 here and between like
45:57 five and 10
45:58 there all right who will this help who
46:01 will what help the creating these bots
46:05 so basically anyone that wants to create
46:08 a
46:09 customized tool that they can share with
46:12 other people it's going to help so maybe
46:15 you're really good at gardening maybe
46:16 you're really good at uh we've got Corey
46:19 in here who's who's into Pottery um she
46:23 might be able to make a a thing that
46:24 will train people how to do their own
46:26 pottery right and it might even be able
46:28 to make images and see images and and
46:32 talk about Pottery things like that like
46:33 literally anyone is going to be able to
46:35 make an
46:37 application that is connected to this
46:40 really powerful large language model and
46:42 is connected out to the rest of the
46:46 world all right
46:49 um let's see let's go back here Were You
46:53 underwhelmed by this Dev keynote no no I
46:56 was not underwhelmed by it I was
46:58 completely overwhelmed by it because
47:00 there's so much there's so much to
47:02 unpack so so let me go back so sorry
47:04 about the the the delay let me let me
47:06 just go back and reset a little bit
47:09 so there's there's there's two main
47:12 things that that Sam talked about in in
47:15 the uh in the keynote today one was the
47:18 developer stuff this is a developer
47:20 conference so most of what he talked
47:21 about lives over in this developer
47:23 playground most of what we talk about on
47:26 this channel is not that now we we are
47:28 going to talk a little bit about that
47:30 cuz it's the only thing I have access to
47:33 so we can go play with some stuff over
47:35 here um but he did also talk about stuff
47:39 over here in GPT so um the custom gpts
47:43 that was the thing that he was just
47:44 showing you how to build so you're going
47:47 to be able to make these things and
47:49 publish them as your own so so you'll be
47:51 able to make your own version of chat
47:52 GPT that's specialized on some something
47:56 that that you're interested in or or
47:57 that you think other people are going to
47:58 be interested in you're going to be able
48:01 to do um Advanced Data analysis within
48:03 it so if you if you've got um let's say
48:08 you've got sales data for your company
48:10 and you want to do a GPT for people to
48:14 ask for reports for you know sales you
48:17 know historical sales data and and run
48:20 analysis on it you can create a bot that
48:22 does that and then retrieval is you can
48:24 upload your own data and you can either
48:26 do data analysis on it or you can
48:28 retrieve it so if you have a list of
48:32 phone numbers for all the different
48:33 offices that could be up there and you
48:36 could say hey what's the phone number
48:37 for the Madrid office and it'll just
48:39 pull that in right because it can
48:41 retrieve your data and then they're
48:43 going to have this GPT store and within
48:47 the GPT store if your Bot is popular and
48:52 gets used a lot they're going to do rev
48:55 sharing for that bot so so this is the
48:59 this is the you can make money
49:03 with uh let's see for Grins and Giggles
49:06 can we ask GPT to decipher the YouTube
49:09 video and
49:12 explain um real world use cases um we
49:16 can
49:18 although it's probably too long for we
49:23 we may have to go to Claude ironically
49:25 enough let's let's see if they've got
49:27 the
49:28 uh the transcript up yet yeah they
49:33 do no results
49:36 found oh so they're not showing their
49:38 transcript oh that was evil of
49:42 them maybe it's just not ready
49:48 [Music]
49:55 yet
50:17 okay let's
50:20 see um so I can't there's no I I don't
50:23 have the transcript to this right now so
50:25 I can't do anything there
50:30 um just think you can connect Bots the
50:32 key is Simplicity and
50:43 reliability thank you so much Nomad all
50:47 right
50:54 fantastic
51:02 harod D what's happening how are you
51:05 what is whisper so whisper is uh open AI
51:09 uh
51:10 transcription uh tool that you'll be
51:12 able to take something like this and
51:14 transcribe it I use whisper in my
51:17 companies I've got a video company
51:18 called storyvine and we use whisper
51:20 technology to transcribe videos and
51:23 they're coming out with whisper version
51:25 three whisper version two was really
51:28 good so version three is probably going
51:29 to be quite spectacular like really
51:32 amazing I I like if you've been around
51:34 software for a while like transcriptions
51:36 have been laughably bad for decades and
51:38 just all of a sudden in the past two
51:40 years they're good Kyle there's some
51:42 text summaries floating around on
51:45 Reddit and I wonder if perplexity could
51:48 hit it
51:50 possibly let's go see
51:54 uh
52:01 um
52:05 focus
52:07 Reddit
52:09 um open AI Dev
52:13 Day video
52:24 transcript
52:26 no couldn't find it couldn't find it oh
52:29 well all right there's some okay so
52:32 let's let's go back to this thing so so
52:34 that's all of that stuff it going true
52:38 multimodal and they said That's supposed
52:39 to happen today it was supposed to
52:41 happen today at 1 I assume this is just
52:44 it's going to take a while for them to
52:45 roll it out that's why we don't have it
52:47 yet April 2023 cut off and then this
52:50 this whole idea of custom GPT so these
52:53 custom bots that we're going to anyone's
52:54 going to be able to make their own I
52:56 mean why I'm excited about these they
52:58 can take action so you can connect them
53:00 to other services you'll be able to
53:03 create
53:04 gpts for yourself personally for your
53:07 friends and your family for your company
53:11 um and then just generally publish them
53:13 out to the world so we're all going to
53:15 be able to create these applications
53:16 without having to know any programming I
53:19 feel like I need to converse with AI
53:21 like I need to teach it and I need to
53:25 learn from it that's so Joseph Baker
53:27 that's a really good point that that
53:30 these things do need teaching they've
53:32 memorized a lot but they don't know who
53:34 you are and what you want and what these
53:37 gpts will allow you to do is say okay
53:41 I'm going to teach it how I like to
53:43 write like I was thinking I've written
53:45 seven featurelength screenplays in my
53:47 day I was thinking I could create a bot
53:50 where I upload my screenplays into it
53:53 and I create a bot that just writes
53:55 screenplay dialogue like I write
53:57 screenplay dialogue now that's something
53:59 I could share with the world or that's
54:00 something I could just use so I could
54:03 just pop into the my little GPT there
54:05 and say hey give me a little scene about
54:07 this that or the other or let's do a
54:09 little outline here and now write me
54:10 some dialogue and it'll just take you
54:13 know training from my seven screen plays
54:17 and use that as part of its training
54:19 data so that's that um okay so now over
54:23 here on the developer side I'm going to
54:24 go through this stuff quick because it's
54:27 really geeky there there's there's not a
54:29 lot of uh of fun stuff to show here but
54:33 I am going to go show you some stuff in
54:34 the playground we're going to make an
54:36 assistant which is basically one of
54:38 these
54:38 gpts but here's a couple of things they
54:41 talked about with GPT turbo so GPT 4
54:44 Turbo
54:45 128,000 token context
54:48 window Claw claw. is a 100,000 token
54:53 context window so it's about 3/4 of that
54:56 amount of of that amount of tokens is is
54:59 words on average so uh whatever that
55:03 would be half of that would be 60 so so
55:08 probably you
55:10 know 85 or 990,000 words 300 Pages
55:15 basically 300 p a 300 page
55:18 book it can have in its memory it's also
55:22 got um threading which which means it
55:25 can remember stuff longer than its
55:27 context so so basically you know how if
55:30 you've ever been using chat GPT and it
55:32 just kind of loses its mind you start
55:34 out working with it you tell it some
55:35 things and it's going really good and
55:37 then all of a sudden it just seems like
55:39 it forgot everything you talked about
55:41 it's literally because it just forgot
55:43 everything you talk about it's only got
55:44 a 3,000-word window and and that's
55:47 that's going up by a lot to like 75 to
55:51 990,000 words from
55:53 3,000 and then they're doing this thing
55:56 where it's also going to create threads
55:58 where it'll have memory of the
55:59 conversation where it can go back and
56:02 and you know augment its memory with
56:05 with that
56:07 thread um it's the the API calls are
56:11 almost three times cheaper on average
56:13 they're 2.75 times cheaper on average um
56:18 than gp4 so so it's a much more capable
56:21 it's GPT
56:23 4.5 128,000 toen context window better
56:28 faster um and almost three times cheaper
56:31 so so from a development standpoint this
56:33 is a big deal they've also added true
56:36 Json as output and then function calling
56:39 so function calling is like being able
56:41 to call like the map they were they were
56:43 showing the the map when I came back
56:45 from dealing with the doy um and and
56:48 Json is basically it can output things
56:52 in structured data format that make
56:55 makes it really easy to use with other
56:58 computer systems basically um higher
57:00 rate limits just basically means it's
57:02 it's two times higher rate limits so so
57:05 uh you can you can do more for less
57:07 money you can do fine-tuning you can do
57:10 this retrieval augmented generation
57:12 which is basically pulling documents so
57:14 a lot of this stuff is stuff that
57:17 um that uh what's it called
57:24 um
57:26 oh I can't remember
57:28 it well it doesn't matter um whisper B3
57:31 is the text transcription and then the
57:34 API will also have access to text to
57:37 speech GPT vision and Dolly 3 so
57:41 developers are going to be able to
57:43 develop applications that can have
57:46 Vision can talk can listen can generate
57:49 images can do all this stuff can call
57:52 actions so all of the stuff that people
57:55 people have been talking about
57:56 autonomous agents all the raw materials
57:59 for being able to easily build
58:01 autonomous agents is in our hands now or
58:04 is going to be in our hands in the next
58:06 week or two all
58:08 right um where do all the calculations
58:11 occur
58:13 um if you're talking about Advanced Data
58:16 analysis those calculations when it's
58:18 writing an executing its own python code
58:21 it's doing that within a container
58:23 within your chat session
58:25 but it is likely also sending calls to
58:28 open AI so some of the calculations are
58:31 being done on open AI servers some of
58:33 them are being done within the
58:36 container um he did say something in in
58:39 the thing about um for anything
58:43 Enterprise they're not they're not
58:45 training any of their stuff and they're
58:46 not like they he basically want one of
58:49 the things he also said in the uh in the
58:52 uh in the keynote was they're going to
58:54 pay your legal bills if you get sued for
58:56 any kind of copyright things or things
58:58 like that any sort of you know security
59:01 or copyright things so um you know
59:04 they're they're they're aware of those
59:07 issues and they're aware that big
59:09 companies are not going to use them if
59:12 uh if they don't get that stuff dealt
59:15 with um okay so let me go show you so so
59:20 now I'm I'm in the GPT playground
59:26 and it's changed a lot so GPT playground
59:29 used to sort of start you out in this
59:31 Mode called um called
59:34 chat and let me see close that you kind
59:38 of have your your system message here
59:40 where that's where you tell it how to
59:42 act and then you basically have your
59:43 chat session here and then over here
59:45 you've got things like temperature and
59:47 maximum length and and all that and
59:50 there's now this gp4 1106 preview so
59:54 that's the
59:55 gp4 turbo preview I
59:59 think um so you can now you can now
1:00:01 start to play with that so you can make
1:00:03 your own kind of chatbot and and prompt
1:00:06 for that but there's also this thing
1:00:08 called
1:00:09 assistance at the top and that's
1:00:11 basically these
1:00:12 gpts so I created one and then there's a
1:00:16 on the left hand side there's a little
1:00:17 menu here called
1:00:19 assistant and I've made one of them 70s
1:00:22 muscle car expert so I just have in a
1:00:24 very short little prompt but if I go to
1:00:27 the playground The Prompt is you're an
1:00:30 expert on 1970s muscle cars from America
1:00:33 no matter what anyone asks find a way to
1:00:36 turn it around to talk about a car it
1:00:38 reminds you of and create a list of
1:00:41 trivia to catch their attention then
1:00:43 start a dialogue about what their
1:00:45 favorite cars of the era are and we're
1:00:48 going to use the gp4 turbo I can also
1:00:51 add in functions so if I know how to
1:00:54 call Api
1:00:55 I could you know have this go out
1:00:58 to I don't know one of the car auction
1:01:00 sites and pull in car data for that
1:01:03 particular car I could use code
1:01:05 interpreter to do math on you know
1:01:08 calculating the cost of buying a car um
1:01:11 retrieval I could retrieve data but now
1:01:14 if I go in here and I say
1:01:20 um what's the
1:01:23 best what's the best
1:01:25 climate to plant
1:01:29 avocados and so I'm going to say add and
1:01:31 run so I didn't ask it anything about
1:01:41 cars and it should here we go so it says
1:01:46 Ah avocados th thrive in warm climates
1:01:48 with mild Winters such as those found in
1:01:50 sou Southern California or Florida
1:01:53 however speaking of Florida it remind
1:01:54 reminds me of one of those 1970s muscle
1:01:56 cars the 70 Hemi Cuda its iconic
1:02:00 performance and design could definitely
1:02:02 heat up the conversation just like a
1:02:04 warm climate that avocados need to grow
1:02:07 and then it gives you a little uh little
1:02:09 trivia about the Hemi the Hemi engine
1:02:12 The Limited production the Barracuda
1:02:14 Evolution um color for personality and
1:02:19 then it says so now that we've dipped
1:02:20 into the uh into some horsepower hype
1:02:23 Nostalgia so so I'll be able to now we
1:02:27 we'll be able to now publish these like
1:02:29 if you just wanted to put out a cool
1:02:30 thing about you know' 70s muscle car
1:02:33 trivia you'll be able to just knock it
1:02:35 out really really quick quickly and put
1:02:36 it out in the
1:02:38 world do you work with llms for a living
1:02:41 no I'm I'm a founder and CEO of a
1:02:43 company called Story vine which is an
1:02:45 automated video storytelling platform
1:02:47 and we've Incorporated a lot of AI into
1:02:49 the platform but it's an 11 11 and 1/2
1:02:52 year old company so most of what we do
1:02:54 is not AI um but but we've rolled it in
1:02:57 and the product road map has a lot of it
1:02:59 in there and I am just I I was a an
1:03:03 entrepreneur in the early days of the
1:03:05 worldwide web in the mid 90s I started
1:03:07 one of the first digital agencies and so
1:03:10 I've been through one of these Cycles
1:03:11 before and so here we are again and I'm
1:03:14 like I'm just the the this generative AI
1:03:19 stuff is like nothing I've ever seen and
1:03:21 so I've spent the last two years just
1:03:23 being AAG man obsessed with it and I
1:03:26 started this
1:03:27 channel um a year ago the first week of
1:03:31 December so I'm coming up on the one-
1:03:32 year anniversary of this channel the
1:03:34 oneyear anniversary of the AI Salon in
1:03:36 fact let me talk to you
1:03:39 about that sweet Sher Sherry D no I
1:03:42 don't I've got one I just need to use
1:03:45 it um let's see watch the whole opening
1:03:48 I think today still confused
1:03:53 about
1:03:55 chat versus
1:03:57 API okay I can I can talk to that do we
1:04:01 know how much we can upload yet we do
1:04:05 not we do
1:04:07 not it seems to be
1:04:11 about what was
1:04:14 um oh that was something they said today
1:04:18 oh thank you very
1:04:19 much one of the things they said today
1:04:23 is
1:04:28 historically if you upload a file that's
1:04:31 larger than the context window so so
1:04:34 chat GPT had a context window of four or
1:04:37 8,000 tokens
1:04:40 fored and so if you uploaded a
1:04:44 document that was 15 times that size you
1:04:48 would have to do what's called chunking
1:04:50 where you would have to take that single
1:04:51 file and break it into chunks and then
1:04:54 and send them all individual basically
1:04:56 embed them all individually and just get
1:04:58 it into sort of chat GPT kind of
1:05:01 language you you had to do that
1:05:03 manually um that's now all being handled
1:05:06 automatically with with this gp4 turbo
1:05:09 so people people are going to be able to
1:05:11 upload oh he said 128k well
1:05:15 128k is
1:05:17 the 128,000 tokens is the number of
1:05:21 tokens I don't know what the file size
1:05:23 uh limit is is there a file size limit
1:05:26 when you upload something the 128,000 is
1:05:29 is uh is the token
1:05:31 limit um maybe I can retrieve some of my
1:05:35 money back from cryptocurrency yeah I
1:05:37 know tell me about it I got I got
1:05:39 something got flushed down the toilet
1:05:41 there
1:05:43 too luckily not a lot oh
1:05:48 man
1:05:53 um
1:05:55 all right let's
1:06:01 see this is this is a lot of this right
1:06:04 now is really frustrating I know because
1:06:07 because we can't actually go play with
1:06:09 these tools so maybe what we should just
1:06:11 do is go play with chat GPT um as it
1:06:14 currently exists because not for nothing
1:06:15 it's still pretty [ __ ] remarkable
1:06:17 even in its old ancient four- week old
1:06:21 features like like uh Dolly 3 and GPT
1:06:25 Vision um so maybe we'll do that U hey
1:06:28 Kyle on assistant is it doing persistent
1:06:31 data retrieval like persistent knowledge
1:06:34 it so so I don't I I'm
1:06:39 I'm I'm I'm a few hours into this too
1:06:42 and I haven't really built anything on
1:06:44 it yet so but what they said today is
1:06:47 yes is that it is going to have because
1:06:50 it's got
1:06:51 retrieval Lang chain Lang chain is the
1:06:54 thing I was thinking about before so
1:06:56 basically what they're creating is even
1:06:59 in the consumer Bots these GPT things as
1:07:01 well as the API for the
1:07:04 developers a lot of the tools that are
1:07:06 in Lang chain they're now making
1:07:08 available within the directly within the
1:07:12 API and it's you're going to be able to
1:07:15 to do all that stuff big warning though
1:07:17 if you play with this make sure you have
1:07:19 a spending cap on your settings yeah for
1:07:22 sure although um it's gp4 is now
1:07:27 2.75 times cheaper than GPT 4 so but
1:07:30 it's still it's still more expensive
1:07:32 than GPT 3.5 if you're the only time
1:07:35 you're going to really if you're just
1:07:37 doing um sort of low volume little chat
1:07:40 Bots and things like that it's not going
1:07:42 to cost you much if you're starting to
1:07:43 do big scale automations and things you
1:07:46 know automations from within um zappier
1:07:50 or or you know you're manually coding
1:07:52 your own things then yeah you could you
1:07:53 could run into some some decent Siz
1:07:55 bills I've never run into any so but I'm
1:07:57 not I'm not sending lots and lots of uh
1:08:01 things it's cheaper now but but not not
1:08:03 cheap it adds up fast yeah
1:08:07 exactly and that's that's another thing
1:08:09 is you know if you start getting into
1:08:11 the development side of this you know
1:08:13 understanding the costs so using things
1:08:16 like GPT 3.5 when you can um there's a
1:08:19 token count tool that will tell you how
1:08:21 big your file is oh that's cool that's
1:08:23 nice but it sounds like they're handling
1:08:25 all of the chunking automatically now
1:08:27 that's that's what um one of the
1:08:30 developers said when Sam brought him
1:08:35 out all right I never saw okay hey Kyle
1:08:38 Kyle here hey Kyle Boyer have you used
1:08:41 chat
1:08:42 GPT for code refactors or fixes yeah we
1:08:46 have um we had we had in in my my
1:08:51 company's app we had a really obscure
1:08:54 bug that was only happening on one
1:08:57 device and it was just crashing out that
1:09:00 one device and um we could not crack it
1:09:06 and it was after like three weeks of
1:09:07 trying to troubleshoot this I just said
1:09:10 just grab the code where it's breaking
1:09:12 and pop it into chat GPT and see what it
1:09:14 says and he had two he had two
1:09:17 characters inverted and he fixed them
1:09:20 and it fixed the bug so I don't know
1:09:21 what it was doing in the code it it
1:09:23 wasn't affecting any other devices but
1:09:25 that one device um but that was an
1:09:27 obscure bug that we never would have
1:09:29 caught right or you know or it would
1:09:31 have taken another couple of weeks for
1:09:32 someone to stumble on it and you know
1:09:34 fix fix that
1:09:36 inversion um so yeah uh and and
1:09:41 refactoring do yeah doing things like
1:09:45 refactoring I've I've heard that that
1:09:48 gp4 was really good at coding and then
1:09:52 it got worse and I don't know if this
1:09:54 new stuff's going to be much much better
1:09:56 I assume it is um but I would say that's
1:09:58 something that you're going to have to
1:09:59 play around with will the marketplace be
1:10:02 lucrative for people creating gpts um
1:10:06 good ones I mean I don't I don't know I
1:10:09 mean I know that one of the things that
1:10:12 Alman talks about is is universal basic
1:10:15 income and his if you go read or or
1:10:18 watch videos on worldcoin his his
1:10:21 identity and cryptocurrency that's all
1:10:24 about Universal basic income so I know
1:10:27 he's
1:10:27 about um you know providing some kind of
1:10:31 equity for
1:10:33 people realistically I don't think
1:10:36 you're going to make a fortune on those
1:10:40 gpts
1:10:42 but I would think that if say Disney
1:10:45 wanted to make a Disney bot or Disney
1:10:49 GPT um and it got used a lot or you know
1:10:53 or George Lucas wanted to make a Star
1:10:55 Wars GPT if they get used a lot they'll
1:10:58 probably generate a fair amount of
1:10:59 Revenue so the it this this is almost
1:11:02 identical to the early days of the App
1:11:04 Store like some some people made big
1:11:06 fortunes um launching relatively shitty
1:11:09 iPhone apps because they just did
1:11:11 something clever that caught people's
1:11:13 attention there's going to be that kind
1:11:14 of opportunity shortterm and then if you
1:11:17 do something really remarkable there's
1:11:19 going to be opportunity long term check
1:11:22 mod Discord all right all
1:11:32 right
1:11:34 um
1:11:35 no
1:11:38 fine thanks have
1:11:50 [Music]
1:11:53 a
1:12:04 uh all right thank you very much jrc
1:12:08 appreciate
1:12:11 that I'm all for Ubi I think Ubi is
1:12:14 interesting we'll we'll see we'll see
1:12:17 how it's GNA play out I don't think it's
1:12:19 I think I don't think we're going to
1:12:20 need it long term but who knows it's
1:12:23 it's just it's way too early to know
1:12:24 what the impact of all this stuff's
1:12:26 going to
1:12:26 be 3.5 turbo is better at coding yes
1:12:30 I've been using Lang chain to automate
1:12:31 code fixes yeah that's that's cool um I
1:12:34 you know I would say definitely play
1:12:36 around with four but if you're doing a
1:12:39 lot of stuff you know it might be too
1:12:40 expensive to use for today was about
1:12:46 4.9 uh file uploads seem interesting
1:12:50 considering I've had to manually chunk
1:12:52 yeah yeah the fact that they're saying
1:12:54 and it's it's doing Auto
1:12:56 chunking and it's doing persistence and
1:12:59 it's doing threading there's a thing
1:13:00 called threading they're doing which can
1:13:02 basically maintain conversations over
1:13:05 time um again I don't because I'm not
1:13:09 developing a lot of applications on the
1:13:12 developer side I don't know a lot of the
1:13:14 implications of this stuff but I know
1:13:17 this is what people are talking
1:13:22 about
1:13:27 [Applause]
1:13:32 [Music]
1:13:33 App Store analogy having reminiscent for
1:13:36 the same thing let's see wasn't today
1:13:39 just about having more
1:13:41 access like nothing about the AI itself
1:13:47 um no I
1:13:50 mean going from an 8,000 or
1:13:54 32,000 token context window to
1:13:58 128,000 that's about that's about the
1:14:01 functionality of the AI itself right
1:14:04 that's about its memory getting better
1:14:07 um it's
1:14:09 faster
1:14:12 um regular human beings who are not
1:14:16 Engineers are going to be able to
1:14:18 essentially do function calling
1:14:21 and um rag
1:14:24 retrieval augmented
1:14:26 generation in in these gpts and they're
1:14:29 like like the non-developers are not
1:14:32 going to know what this shit's called
1:14:33 but they're going to know I can upload
1:14:34 some of my documents and I can make a a
1:14:38 one of these
1:14:39 gpts that answers the questions my
1:14:42 employees
1:14:43 have and so it's it's uh no I think this
1:14:48 is way more than than access this is
1:14:51 this is way more than access this is
1:14:54 this is fundamentally
1:14:58 shifting all right so here's a good
1:15:00 example so right now with chat
1:15:04 GPT you've you've got all of these
1:15:06 people you can make money with chat GP
1:15:09 making um prompt libraries and and you
1:15:12 know take my course and I'm going to
1:15:14 give you all these prompts and I'll give
1:15:15 you access to this prompt library and
1:15:17 stuff like that and then you have to go
1:15:20 find their database of prompts and you
1:15:22 have to sift through it and find things
1:15:24 that are relevant to what you want right
1:15:25 now and then you have to go to your own
1:15:27 version of GPT and you have to put in
1:15:29 your prompt and then if the prompt isn't
1:15:31 quite right you got to modify
1:15:35 it as opposed to that same person that
1:15:39 wrote all those prompts could now write
1:15:41 a bunch of bots and there might be
1:15:42 here's the gardening bot and here's the
1:15:45 small business bot and here's the pizza
1:15:47 marketing bot and here's the make cool
1:15:51 Instagram images bot I use it to make
1:15:54 lesson plans as a teacher yeah here's
1:15:56 the lesson plan making bot and then
1:15:58 those those Bots those gpts can get
1:16:02 published and if I'm looking for if I'm
1:16:04 a teacher looking for lesson plans I I'm
1:16:06 not going out to find a lesson plan
1:16:08 prompt so I can put it in chat GPT I can
1:16:12 just go find the GPT that's specifically
1:16:14 designed to do lesson plans for the
1:16:16 fourth grade in Southern Illinois that
1:16:19 follows the school board's
1:16:21 recommendations for the curriculum in
1:16:23 Southern Illinois someone's going to
1:16:26 make that GPT and I could just go get it
1:16:29 so so I I think it changes a lot it you
1:16:33 know it it changes who can make an
1:16:36 application it changes how you
1:16:39 use GPT in the first place there's a
1:16:42 tremendous amount of people right now
1:16:44 that don't use chat
1:16:46 GPT it's like north of
1:16:49 80% haven't used it once none so 80% so
1:16:56 I don't know how many in here there's
1:16:57 176 people in here right now if you
1:17:00 haven't used chat GPT you should go
1:17:02 there and use it but there's probably a
1:17:03 decent percentage in here that haven't
1:17:05 used
1:17:06 it and so I think I think the
1:17:09 announcements today the implications of
1:17:11 it are over the course of 2024 there's
1:17:14 going to be this whole world sort of
1:17:17 opened up to people um that that are
1:17:21 going to make it much more access
1:17:24 ible so let's see just got here where's
1:17:28 the GPT for
1:17:31 genealogy
1:17:32 um I don't know there's probably if you
1:17:35 go to Future
1:17:43 pedia.com GPT thing I think you could
1:17:47 probably make one um I need this Pro
1:17:52 service litigation in defense against
1:17:55 government oh I need this for proservice
1:17:58 litigation in defense against
1:18:02 something Kyle it's going to be like the
1:18:05 iPhone and the App Store yes exactly
1:18:08 Jonesy it is what do you think about
1:18:11 gpts and the marketplace of
1:18:14 assistance that's that's what we're
1:18:15 talking about right now I think it's
1:18:17 absolutely flipping huge it it's
1:18:22 huge huge huge huge I ask my friends and
1:18:25 many of those who haven't used it they
1:18:28 say they will wait until it's
1:18:31 easier yeah and it's listen this stuff's
1:18:33 not going to get super easy for some
1:18:36 time just because the nature of what
1:18:40 these tools are and what they're doing
1:18:42 is is profoundly powerful and they're
1:18:45 they're they're all kind of separate
1:18:47 tools right now and the announcement
1:18:50 today says that they're going to start
1:18:51 rolling them together so so hopefully
1:18:54 over the next week or so we're going to
1:18:55 start seeing that sort of
1:18:57 stuff
1:19:00 um yeah
1:19:06 I I don't know that people have as much
1:19:08 time as they
1:19:10 think I I think it's just I think it's
1:19:13 really imperative on as many of us as
1:19:15 possible to just get people to try it
1:19:18 get a little bit educated about it
1:19:19 because it's moving so damn fast that by
1:19:22 the time it gets easy enough for
1:19:24 everyone to use it like a lot of the
1:19:26 opportunity is going to pass people
1:19:29 by um it can't be easier it's literally
1:19:32 like talking to a person well it's so
1:19:35 what what can be easier what's your name
1:19:38 Tyree logston um what can be easier is
1:19:42 what do you use it for right and that's
1:19:45 where I think these agents these these
1:19:48 uh assistants these gpts are going to
1:19:50 help because right now you go to to chat
1:19:53 GPT and it's just like this empty box
1:19:55 and you don't know what it's for and I
1:19:57 think these agents it's going to be
1:19:59 easier to Market them to people hey do
1:20:01 you run a pizza shop here's the pizza
1:20:04 shop marketing
1:20:06 GPT oh I need to Market my pizza shop
1:20:09 let me start talking to oh that's really
1:20:11 cool oh that was easy so yeah it's it's
1:20:13 easy to talk to them but it's not easy
1:20:15 to know what to do with them and I think
1:20:17 that's that's one of the things that
1:20:19 this is going to help a lot so so here's
1:20:22 the deal all of us anyone who's talking
1:20:25 at all about this anyone who's
1:20:26 experimenting at all with this stuff is
1:20:29 going to be able to
1:20:32 um like have a have a a head start on
1:20:36 everyone
1:20:38 and generate some of these Bots and put
1:20:40 them in the world and get noticed for it
1:20:43 if everyone can Dev with it what
1:20:44 prevents copycats from making identical
1:20:47 Bots and apps J Hamilton that is the Jay
1:20:50 Hamilton
1:20:51 1776 um
1:20:54 that is the billion dooll
1:20:56 question so if you want to see an
1:20:58 interesting there's an interesting
1:21:00 article called we have no moat M
1:21:04 oat that was written by a Google
1:21:07 engineer talking about the fact that he
1:21:09 was talking about large language models
1:21:11 that basically the open source Community
1:21:14 is starting to
1:21:15 make their small inexpensive large
1:21:18 language models behave you know perform
1:21:20 almost as good as the really expensive
1:21:22 ones and he said there's no way we can
1:21:24 Pro protect against
1:21:27 this nothing is going to prevent people
1:21:29 from you know copying those Bots
1:21:33 now it's not just the prompt that's in
1:21:36 that bot there's also the documents that
1:21:39 you upload right so if you upload a
1:21:42 bunch of documents that are really good
1:21:44 proprietary data you might have a GPT
1:21:47 that even though it's the same on the
1:21:49 surface is doing things way better than
1:21:51 someone else's GPT that doesn't have as
1:21:53 good of data I think private data is
1:21:56 going to be the place where software and
1:22:00 companies differentiate themselves using
1:22:02 tools that are essentially
1:22:04 identical use Bots for what exact
1:22:07 exactly custom mobile apps yeah custom
1:22:09 mobile apps
1:22:11 um personal assistance um travel
1:22:15 shopping
1:22:22 Bots
1:22:35 uh uh uh
1:22:38 uh nothing that's why there's 28
1:22:41 Different official chat GPT Bots before
1:22:44 open
1:22:45 AI yeah yeah well those were all just
1:22:48 copycats trying to copy
1:22:51 it what was that video called the video
1:22:54 is the official open AI Dev
1:22:58 day opening
1:23:05 keynote from open
1:23:12 Ai and again there's there's kind of
1:23:15 there's kind of two things that he talks
1:23:17 about in there he starts out talking
1:23:19 about GPT for Turbo which is the base
1:23:23 the base
1:23:24 model and I that applies to both chat
1:23:27 GPT and to the
1:23:29 developers then he talks about chat GPT
1:23:32 in particular with this this GPT maker
1:23:36 and then the whole rest of the keynote
1:23:38 is for developers basically so the bulk
1:23:41 of this conversation is for the
1:23:43 developers so if you're not a developer
1:23:46 just watch the just watch the middle
1:23:48 section where he's talking about chat
1:23:49 GPT and these things called gpts um cuz
1:23:54 that's the stuff that we'll have access
1:23:55 to apparently today but I still don't
1:23:59 have access to it let me go see if I get
1:24:01 access to it
1:24:06 now
1:24:08 no option
1:24:12 reload no no I got nothing I got
1:24:21 nothing
1:24:25 will standard filters applied to
1:24:28 personal gpts I don't know what standard
1:24:31 filters I don't know what you mean by a
1:24:32 standard filter I on
1:24:34 China any tips for better prompt writing
1:24:37 I've had GPT ignore instructions or
1:24:41 rules
1:24:46 um my general recommendation on prompt
1:24:49 writing
1:24:51 is think of prompts more like
1:24:54 conversations than um than writing the
1:24:59 perfect prompt so so don't just think
1:25:02 like oh I've got to write the perfect
1:25:03 prompt and get the perfect answer back
1:25:05 that's sort of the Google model of of
1:25:08 what Google sort of trained us to
1:25:10 do and with chat GPT it needs a lot more
1:25:14 context so start out and you know tell
1:25:17 it what role you want it to play and
1:25:19 then talk give it some background about
1:25:22 what your about to do and then ask it if
1:25:25 it understands and say things like take
1:25:28 a deep breath and and think through this
1:25:30 step by step before answering me
1:25:33 surprisingly those kind of things make a
1:25:37 measurable difference in the quality the
1:25:39 the output and then when it does answer
1:25:41 you tell it if it got it right tell if
1:25:44 it if it got it wrong and then just keep
1:25:47 iterating and iterating and iterating
1:25:49 that for me that's that's really where
1:25:51 the art of promp comes in um creating
1:25:55 bots you know until the until and always
1:25:58 be polite that's that's a good one Herod
1:26:01 um until the uh until this thing shows
1:26:03 up for us you can go practice on po.com
1:26:06 make yourself a bot there publish it and
1:26:09 start interacting with it and if it
1:26:11 starts giving you shitty answers go
1:26:12 modify your prompt and then go back to
1:26:15 it and you know modify it just keep keep
1:26:17 iterating on Bots and it'll it'll help
1:26:20 your prompting a lot what your take on
1:26:24 boot
1:26:25 camp I don't know boot camp let me see
1:26:29 let me go look let me go look is that
1:26:32 new is this open AI or something
1:26:36 else boot camp
1:26:51 AI
1:26:57 I don't know I don't know
1:27:00 what what you're talking about with boot
1:27:03 camp Google is keywords totally
1:27:05 different mindset yeah exactly Earth
1:27:07 rocker it's it's it's just a completely
1:27:09 different thing the thing about it
1:27:10 though
1:27:11 is we've got 15 or 20 years of
1:27:14 conditioning for how to write a keyword
1:27:17 prompt and so I think people make the
1:27:19 mistake with chat GPT of thinking it's
1:27:21 the same thing and it's not
1:27:23 tell chat
1:27:24 GPT that it's a mock
1:27:27 situation and it will be what you
1:27:30 want yeah well it that that's a way to
1:27:34 get around it depending on what you want
1:27:35 to do it's still it's safety guard rails
1:27:37 are still going to come
1:27:39 in uh uh uh
1:27:43 um hang on one
1:27:51 sec
1:28:02 all right let's see don't forget to like
1:28:05 and follow oh okay yeah so thanks Silver
1:28:08 Fox for reminding me of that so do me a
1:28:10 favor here
1:28:12 um one of the things we can do is we can
1:28:15 go in and start playing with chat GPT um
1:28:19 although we're kind of the the world is
1:28:21 discombobulated right now cuz they made
1:28:23 all these big announcements and we don't
1:28:25 have access to them so it's hard to
1:28:26 actually show you show you what they are
1:28:29 without you know totally geeking out
1:28:31 right now um you know and going going
1:28:34 doing stuff that looks more Cody it's it
1:28:36 it doesn't demo well when when you're on
1:28:38 the development side um but if you would
1:28:42 um follow my channel I'm I'm like 350
1:28:45 people away from a a big milestone of
1:28:47 30,000 followers and November 30th is
1:28:50 coming which is the oneye anniversary of
1:28:52 of chat GPT so so follow that and then
1:28:55 also if you'd be so kind share the live
1:28:58 um so we can get some more people in
1:28:59 here and
1:29:02 um and what else um you know what I want
1:29:05 to talk about I want to talk
1:29:08 about so so this live is part of a
1:29:12 larger
1:29:14 thing a community that I
1:29:18 started um last December the the the
1:29:22 week after chat GPT launched I created
1:29:25 this thing uh with with uh this woman
1:29:28 Leah faston who's a professional
1:29:29 photographer out of Boston called the AI
1:29:32 salon and if you go to the salon. um
1:29:35 it'll walk you through
1:29:39 um it's a link tree the first link will
1:29:41 walk you through what the salon's about
1:29:44 what our values are um and there's a
1:29:48 Meetup so we have a Meetup tomorrow so
1:29:50 Cindy [ __ ] who's here in the in the the
1:29:52 uh she's been in the channel tonight
1:29:54 she's speaking she's a futurist she was
1:29:57 she was on the live stream uh when when
1:29:59 we looked at the uh the keynote today
1:30:02 from open Ai and she was just like oh my
1:30:04 God my mind's blown which a lot of
1:30:06 people were and um so she's going to be
1:30:09 speaking tomorrow night and she's got a
1:30:11 really cool uh thing she put together
1:30:14 that she's going to share with everyone
1:30:15 and we're going to do kind of an
1:30:16 interactive thing so that meeting's
1:30:18 tomorrow night so the information for
1:30:20 the Meetup is the second link on this
1:30:22 page and then the third link is the link
1:30:25 to the Discord and so over on the
1:30:27 Discord introduce yourself there are
1:30:31 guilds um it's it's a really cool
1:30:33 community of people check out the values
1:30:36 uh at at the AI Salon you know the
1:30:38 values are they're going to seem like oh
1:30:40 yeah those are nice but like really take
1:30:42 them in because I while I want to grow
1:30:45 this community I want to I want to
1:30:46 retain what's really good about it so so
1:30:49 like just really quickly the values for
1:30:51 the salon the first one is curiosity
1:30:53 like we want people that are curious we
1:30:55 want people that are generous so as they
1:30:57 learn stuff they share stuff um we want
1:31:01 people that are adventurous and wanting
1:31:03 to explore and not you know not
1:31:05 operating from a place of fear but
1:31:07 operating from a place of Adventure uh
1:31:09 we want people that are collaborative
1:31:11 like hey let's let's do something you
1:31:13 know I got a barn I'll make the costumes
1:31:15 let's put on a show um that's the spirit
1:31:18 we need right now just in general with
1:31:20 all this stuff cuz things are moving so
1:31:22 fast
1:31:22 inclusion is another one people are are
1:31:25 coming from all walks of life into this
1:31:27 right now cuz nobody knows what's
1:31:29 happening nobody if anyone tells you
1:31:31 they're an expert they're not nobody
1:31:34 knows what's going on with this stuff
1:31:36 and then the final one is empathy just
1:31:38 understand that we're
1:31:40 we're we're in a very volatile and shaky
1:31:46 um world anyway right
1:31:49 now and then you know you have what's
1:31:51 going on in in the Middle East and
1:31:54 you've got um what's going on here with
1:31:57 politics in this country and and now
1:31:59 we've got this AI thing that's upending
1:32:02 a lot of jobs and and sort of generating
1:32:05 a lot of insecurity and there's a lot of
1:32:07 fear out there so um so empathy is a
1:32:11 really big part of the salon as well so
1:32:13 if you resonate with those values please
1:32:15 please join the salon and uh and come
1:32:17 hang out here the the whole point of
1:32:19 these things I've been doing these these
1:32:21 lives
1:32:23 nightly seven nights a week for the past
1:32:26 5
1:32:27 months
1:32:29 so why am I doing
1:32:32 that you know I'm not I don't really
1:32:35 have an endgame right now I why I'm
1:32:37 doing it is I think it's very very
1:32:38 important and and you know when we
1:32:41 watched that thing live today as a
1:32:43 community it it was really it was really
1:32:45 quite amazing cuz it's a lot of people
1:32:47 that have been in here sort of kicking
1:32:49 the tires on this AI stuff and they knew
1:32:51 what all of the things that were
1:32:53 announced meant even if they didn't know
1:32:54 what they meant technically they they
1:32:57 knew what was going on they knew that
1:32:58 something different was happening so
1:33:01 anyway that's what that's about mods
1:33:03 please pin Kyle how did you wait how did
1:33:07 you realize to jump on this in five
1:33:10 days so so Veranda I
1:33:17 um so I discovered I I discovered St
1:33:22 diffusion before chat gbt so so for
1:33:26 about a year before chat gbt I was
1:33:28 playing around with a an nft project
1:33:31 called pixel mind that was built on top
1:33:33 of stable
1:33:35 diffusion and then I thought I got to
1:33:36 teach myself stable diffusion because I
1:33:38 want to I want to know how this stuff
1:33:39 works and so I failed nine times to get
1:33:43 stable diffusion up and running on
1:33:44 Google collab and it was really
1:33:46 frustrating and I felt like a [ __ ]
1:33:47 idiot I'm like I'm just not good enough
1:33:50 I'm not technical enough and and then I
1:33:52 finally got it up and working and I
1:33:55 figured out how to make dream Booth work
1:33:57 and that was another four or five
1:34:00 misses and
1:34:02 so probably
1:34:05 in October of
1:34:08 22 a good friend of mine Chris Needum
1:34:10 said hey will you talk to my friend Leah
1:34:12 she's a
1:34:13 photographer and she saw some of your
1:34:16 artwork I created this project called
1:34:18 Kyle Shannon dreams so if you go to Tik
1:34:20 Tok Kyle Shannon dreams or if you go to
1:34:23 instagramy Shannon dreams that's this
1:34:26 art project I did of just these kind of
1:34:28 stable diffusion wacky self-portraits
1:34:31 and and um personas and it was just a
1:34:35 fun project and so I was talking to Leah
1:34:38 and she's a professional photograph
1:34:41 photographer and she's like well what's
1:34:42 going to happen with photography and we
1:34:44 just got to talking and we were like we
1:34:46 should put together a group that's
1:34:47 trying to figure this [ __ ]
1:34:49 out and so that was kind
1:34:53 of early November mid November and so
1:34:57 she and I were already talking about
1:35:00 um doing something you know really just
1:35:03 for
1:35:04 artists and then chat GPT came out and I
1:35:08 just said [ __ ] it we got to go and so I
1:35:11 literally just just we we set up
1:35:17 um I think we set up a salon we did our
1:35:20 first salon four days after we decided
1:35:23 to do it like it it was just we
1:35:26 just we just went it's it's one of my
1:35:30 it's one of my fatal flaws and and I
1:35:35 don't know I guess it's a good thing I
1:35:37 just decided [ __ ] it I'm going to go do
1:35:39 it like it just felt important to me
1:35:41 here's why I knew that I needed to do it
1:35:44 so in
1:35:46 1994 when I started learning about the
1:35:49 worldwide web and I started teaching
1:35:51 myself HT ml I was like there's got to
1:35:54 be people that know more about this than
1:35:56 I
1:35:57 do are you our Professor
1:36:02 X
1:36:05 um you know I know there's got to be
1:36:08 people that know more about this than I
1:36:09 do and so I put together this thing
1:36:11 called the worldwide web artist
1:36:13 Consortium and um what I discovered is
1:36:17 that nobody knew
1:36:20 anything everybody was trying to figure
1:36:22 it out and let's
1:36:26 see there's a cool
1:36:29 one and so I thought well this AI
1:36:33 stuff's going to be the same way cuz my
1:36:35 big Epiphany with chat GPT was in 90
1:36:39 minutes never having written a line of
1:36:40 python I wrote a a fully functional
1:36:43 python
1:36:46 application and
1:36:48 uh I just knew that the world had just
1:36:51 [ __ ] changed and so I just put the uh
1:36:54 put the salon together so I don't know
1:36:56 and I started so I started a newsletter
1:37:00 this Tik Tok Channel and the AI Salon in
1:37:02 a 3-we period and back in 1994 I
1:37:06 started um an online magazine called
1:37:09 Urban desires um the worldwide web
1:37:12 artist Consortium which was that group
1:37:14 and then agency.com which was one of the
1:37:16 first digital agencies I did that within
1:37:18 a month I did those three things within
1:37:20 a month so so once once I have one of
1:37:24 those moments where I something kicks in
1:37:26 that oh [ __ ] this is this is something
1:37:28 you need to pay attention to I just go
1:37:30 so I don't know that's how I did it
1:37:33 that's how I did it all right fantastic
1:37:36 so many people in here it's great having
1:37:37 y'all in here it's really great um all
1:37:40 right let's let's uh first of all what
1:37:43 the what the what the hell's going on
1:37:45 with the hair here
1:37:47 people hang on I need a little I need a
1:37:50 little quaffing
1:37:52 forever hooked thank you appreciate
1:37:58 that all
1:38:01 right can I instruct it to replicate
1:38:04 this video script and then create a
1:38:06 monetizable version using AI
1:38:09 sure um go to go to the YouTube channel
1:38:13 so if you go to
1:38:15 um AI learninglab dtt on YouTube all the
1:38:19 transcripts are there yeah go turn it I
1:38:22 was thinking about that today that I
1:38:23 could take the
1:38:25 um I could take the transcripts of all
1:38:28 the all the videos of these that I do
1:38:32 and put them up there but it's all it's
1:38:33 all historical so I don't know it's me I
1:38:36 mean those scripts are me just rambling
1:38:38 like like it's it's Chad ad sometimes
1:38:41 I'm just cracking jokes sometimes I'm
1:38:42 demoing [ __ ] Kyle is AI Jack Black I'll
1:38:46 take it I'll take it I like you yeah
1:38:49 [Music]
1:38:50 woo
1:38:53 oh
1:38:54 man hair growth geek going on about my
1:38:57 hair I need some product hair growth
1:38:59 geek we got some well the product I need
1:39:01 is [ __ ] scissors thank you Sher bear
1:39:04 have you been a part of a hacker space I
1:39:06 have not um I've been in a lot of
1:39:09 hackathons um this the salon in a great
1:39:13 way is is kind of turning into a a a big
1:39:16 hacker space the guilds so what we did
1:39:18 within the salon is it used to just be
1:39:21 would meet every other week we had this
1:39:22 Discord Channel and there there were
1:39:24 kind of sections in there to have
1:39:26 discussions but
1:39:28 um we we added guilds and so like we've
1:39:32 got an AI 101 Guild for beginners we've
1:39:35 got a an AI mechanics Guild for geeks
1:39:39 we've got um an Art Guild a writing
1:39:41 Guild a business Guild an entrepreneur
1:39:43 Guild a healthc care Guild um we're
1:39:46 looking for someone super passionate
1:39:47 about education to head up an education
1:39:49 Guild I think it would be great to have
1:39:51 people that are passionate about
1:39:53 education trying to figure this [ __ ] out
1:39:55 rather than trying to figure out how do
1:39:56 we block it how about we figure out how
1:39:59 to incorporate this stuff into the
1:40:02 classroom since this stuff's going to
1:40:03 transform the world that students are
1:40:05 going to be entering
1:40:08 into right Kyle mentioned his hair have
1:40:12 a drink oh are we playing Kyle bingo so
1:40:17 don't you also have to drink when I do
1:40:18 this you can make
1:40:20 money
1:40:22 I called it the salon so there's a it's
1:40:25 not hair salon although I should call it
1:40:28 that um there there's a tradition of
1:40:30 salons in the in the late 19th century
1:40:34 um of people just getting
1:40:35 together and you know bringing in
1:40:38 speakers and and talking about you know
1:40:41 talking about things so so it's really
1:40:42 just that idea of having a group of
1:40:45 people that come together in a curated
1:40:48 kind of way um to learn and to explore
1:40:52 and to experiment uh and things like
1:40:54 that so just look up the history of
1:40:55 salons in the in the you know
1:40:58 late late 1800s and early 1900s and then
1:41:02 there were some some also some in the in
1:41:04 the middle of the century that were
1:41:06 pretty amazing too I'm the director of
1:41:08 innovation for my company your channel
1:41:10 has made the light click for me awesome
1:41:13 fantastic Jay Hamilton a lot that's
1:41:16 great that's great listen here's here's
1:41:18 the whole thing this is so that's that's
1:41:21 why this Channel's here
1:41:23 is it's very easy to be cynical about
1:41:26 this stuff it's also if you're if you're
1:41:28 director of innovation for a big
1:41:32 company there's no inherent rush you
1:41:35 know this there's no inherent rush in
1:41:38 big companies to actually innovate
1:41:42 right the problem with this AI stuff is
1:41:46 it's so powerful and it's so
1:41:48 transformative that what's going to
1:41:50 happen is
1:41:52 a lot of organizations are going to get
1:41:54 caught sleeping where smaller companies
1:41:56 are going to come
1:41:58 in like out of nowhere really
1:42:03 nimbly that historically wouldn't have
1:42:05 been able to compete with a big company
1:42:08 and they're going to come in and they're
1:42:09 going to kick the [ __ ] out of the big
1:42:10 company and and the big companies are
1:42:13 are like literally going to be like wait
1:42:15 that's not possible how's that even
1:42:17 possible like there's going to be a lot
1:42:18 of companies that get blindsided by
1:42:22 small organizations that have gone all
1:42:24 in on AI and have been doing it for a
1:42:26 year or two or three so to the extent
1:42:30 that big companies are kind of sitting
1:42:32 back going ah we've got some years it's
1:42:34 not secure yet it hallucinates like they
1:42:36 know enough about generative AI to know
1:42:38 all the bad [ __ ] about it but they're
1:42:41 not actually curious about it and
1:42:42 exploring it so so Jay I I would
1:42:45 encourage you to join the salon um hang
1:42:47 out on the Discord and like share what
1:42:49 you're
1:42:50 learning you know you know ask questions
1:42:53 there it's that's that's the whole idea
1:42:54 is to make a very safe space where
1:42:56 people can just geek out on this stuff
1:42:58 without you know fear of being around
1:43:01 people that are just trying to [ __ ] all
1:43:03 over it
1:43:05 um creative shop education Guild would
1:43:08 be autum uh would be awesome yep I would
1:43:11 love to see
1:43:13 [Music]
1:43:16 that and by the way the way we do guilds
1:43:20 so
1:43:22 there's lots and lots of guilds we could
1:43:24 have all of the guilds that we have they
1:43:27 exist because someone naturally emerged
1:43:30 as a leader within the community that
1:43:32 was really passionate about that thing
1:43:35 and so um so we're starting with the
1:43:37 people right back to that idea of
1:43:39 empathy and inclusion we're starting
1:43:41 with who's who's showing up and who's
1:43:43 showing up in a way that they they
1:43:46 really want to talk about a particular
1:43:47 thing like Roger and Pate in who who
1:43:51 co-lead the mechanics Guild they were
1:43:53 both really passionate about explaining
1:43:56 the how the technology of large language
1:43:59 models work to a non-technical audience
1:44:02 and so they made a guild you know with
1:44:04 with that as the imperative and that's
1:44:07 that's the thing I want is I want people
1:44:09 that have some sort of vision for how
1:44:11 they want to connect with other people
1:44:13 and how they want to help other people
1:44:14 and what they want to accomplish Sher D
1:44:17 who started the uh the healthcare Guild
1:44:20 with Joker and
1:44:22 jrc um similar sort of thing like they
1:44:25 they've got passion for it and they want
1:44:27 to do it so anyway how do I lead an
1:44:29 education Guild guide on the Discord so
1:44:32 just me and the trees so just jump in
1:44:35 and start
1:44:36 participating um introduce yourself if
1:44:39 you have a different Discord name than
1:44:41 just me and the trees like let us find
1:44:44 let give us a way to connect your Tik
1:44:47 Tok handle and who you are in Discord
1:44:50 and then just jump in and and start
1:44:51 participating you'll that you'll get
1:44:53 welcomed in there and start
1:44:54 participating and if it's if it's a
1:44:56 community that you like and you know
1:44:58 come to some Salon meetings and it's a
1:45:00 kind of thing where you want to spin up
1:45:01 a community within it then then let's
1:45:03 talk and we can we can make it happen um
1:45:06 I uploaded a basketball game
1:45:08 play-by-play list and GPT wrote a sports
1:45:10 article with stats perfect now I hope
1:45:13 you Bob Hamilton also went back and
1:45:16 checked the stats because sometimes
1:45:19 it'll give you stats you think are right
1:45:21 cuz you uploaded the data but they're
1:45:22 not really right so just so make sure
1:45:24 you confirm that but assuming you did
1:45:26 yeah it's it's great for stuff like that
1:45:29 I can send you the Tik Tok I did not
1:45:31 okay
1:45:33 fantastic all right let's
1:45:36 see Kyle when are you switching from
1:45:39 Discord to the other platform hey
1:45:41 Barbara so we're going to announce um
1:45:46 we're going to
1:45:48 announce tomorrow that we're making
1:45:50 making the
1:45:52 switch
1:45:53 um we're we're going to ask oh I've got
1:45:56 to set this up Jesus I got I got so much
1:45:58 to do um we're going to have people um
1:46:03 donate in dollars how much they hate
1:46:06 Discord from like1 to
1:46:09 $20 and and then uh because um because
1:46:14 we we're switching to Mighty networks
1:46:16 and uh and we've got it mostly built out
1:46:19 so we're going to announce it tomorrow
1:46:21 then probably middle of the month
1:46:24 um we'll probably start to move the
1:46:27 guild members over so there's some
1:46:28 activity in there and then on our
1:46:31 one-year anniversary on December 5th um
1:46:35 we're going to fully roll out Mighty
1:46:37 networks the new the new um the new
1:46:41 community platform and it's it Mighty
1:46:44 networks is really good it's it's we
1:46:46 we've got it set up and there's a decent
1:46:48 amount of people in there now kind of
1:46:49 playing around and kicking the tire and
1:46:51 everyone seems to be liking it so I
1:46:53 think it's going to be quite good um I
1:46:55 hate the other more sorry it's okay
1:46:57 Joker so you know that's we're going to
1:47:00 the Discord will likely stay alive for
1:47:03 you know playing with Bots playing with
1:47:05 mid Journey there'll probably be some
1:47:08 you know sub community that really does
1:47:10 like Discord the problem with Discord is
1:47:12 so many people haven't used it it's it's
1:47:14 really been a hindrance for um for the
1:47:17 growth of the community
1:47:19 so cuno $20,000 she hates she hates
1:47:26 Discord uh forever hooked I got 100
1:47:29 bucks for you to match awesome that's
1:47:30 great yeah it's it's going to cost us
1:47:32 116 a month um for the for the level
1:47:35 that we need for the community so so at
1:47:38 some point we'll we'll figure out how to
1:47:39 do I think what we might do at some
1:47:41 point is is maybe it's like 99 cents a
1:47:44 guild and then I don't know $299 for all
1:47:47 the I don't know we'll we'll figure it
1:47:48 out we'll figure it out um I just
1:47:51 started learning Discord uh will Mighty
1:47:54 networks be better um yeah so so mighty
1:47:57 networks is web-
1:47:59 based um it's got a feed kind of like
1:48:02 Facebook so you log in and like um your
1:48:05 feed is generated from any activity in
1:48:07 things you've joined um you can join
1:48:11 guilds you can join channels there's
1:48:14 chats there's um like one of the
1:48:17 problems in Discord is you're you're
1:48:19 having a chat right right and someone
1:48:21 says something important and then lots
1:48:23 of people start talking about it and the
1:48:25 important thing they said just just like
1:48:27 goes away cuz there's all this you got
1:48:29 to like scroll back to see what people
1:48:31 are talking about um within Mighty
1:48:34 networks you can have different kind of
1:48:36 pages that are not really chats but
1:48:37 they're more like a post like I post
1:48:39 here's an image I made and then people
1:48:41 can vote on that and comment on it it's
1:48:44 got calendaring you can go live within
1:48:47 it so so one of the things we encourage
1:48:49 is people teach teing what they know um
1:48:52 so you're going to be able to just
1:48:54 basically if you're in a guild you can
1:48:56 just go live and have people join your
1:48:58 live right right from within
1:49:00 it
1:49:04 [Laughter]
1:49:07 [Applause]
1:49:19 um um let's
1:49:22 see I can start a private chat server
1:49:25 for free yeah I I I mean we could do it
1:49:29 for free it's it's just this is this is
1:49:32 one of those things where it's just
1:49:33 going to take a while like like you
1:49:35 could spin your wheels on this forever
1:49:37 there there are 40 different ways to do
1:49:39 this Mighty networks has been around for
1:49:41 like 15 years they're they're decently
1:49:43 sophisticated so I'm I'm happy with it
1:49:45 so far um the other thing that it's got
1:49:48 that I'm excited about is people are
1:49:50 going to be able to make um courses so
1:49:54 you can make a course within it and then
1:49:56 you'll even be able to charge for those
1:49:58 right and so so individuals are going to
1:50:00 be able to make money and the the the
1:50:03 salon will be able to you know um
1:50:05 sustain itself that way is mighty
1:50:08 networks more access accessible for
1:50:10 screen readers or closed captions in
1:50:13 live I don't know I think the lives I
1:50:16 think the lives will probably still go
1:50:18 live with Google meet
1:50:21 um which has closed captionings in it
1:50:23 but I don't
1:50:25 know
1:50:30 uh harod D how can I make M with it g to
1:50:34 going to need a new
1:50:39 button oh
1:50:42 [Laughter]
1:50:45 man have you heard about chat jpt
1:50:47 releasing a Bots Builder yeah we were
1:50:49 talking about that earlier then I had to
1:50:51 run off um yes the Bots Builder they're
1:50:55 called
1:50:57 gpts you don't have to know programming
1:51:00 with them it will they will actually
1:51:02 help you write themselves and then you
1:51:04 can do things like connect them to other
1:51:07 services and um upload your own
1:51:10 documents to them it's going to be
1:51:12 incredible uh it's it's going to be
1:51:14 incredible you lost me at Myspace I mean
1:51:17 web- based
1:51:19 oh yeah
1:51:21 I mean Discord is webbased it's just a
1:51:23 different kind of browser but that's
1:51:25 it's it's basically a web all Discord is
1:51:27 is a fancy web browser so
1:51:32 technically
1:51:35 [Music]
1:51:41 uh my company tells us that the reason
1:51:45 wait my company tells us all the reasons
1:51:47 it's dangerous I ask
1:51:49 them
1:51:51 how are we getting to yes yeah yeah
1:51:55 there there's there's an infinite number
1:51:57 of ways to say no to AI
1:52:06 absolutely what was that about Kyle
1:52:09 laugh type continue talking oh I don't
1:52:12 know the mods are just chatting about
1:52:14 stuff so I've got I've got my iPad off
1:52:19 to the left there d
1:52:25 oh
1:52:26 cool digital
1:52:28 gods made a
1:52:30 transcript of the AI
1:52:49 conference
1:52:53 download all right so
1:52:59 let's I have a feeling this is going to
1:53:01 be way too big for the current chat GPT
1:53:05 to deal
1:53:18 with oh that's it okay
1:53:33 okay um can
1:53:36 [Music]
1:53:39 you
1:53:42 summarize the main points
1:53:46 from
1:53:49 this no that's too
1:53:52 big though we can probably go over to
1:53:57 Claude which Claude is not happy right
1:54:00 now anthropic is probably pretty pissed
1:54:02 off right now so anthropic if you
1:54:04 haven't played with Claude claud's big
1:54:06 claim to fame was that it's got 100,000
1:54:09 token context window so you could take
1:54:11 say a transcript of a one hour talk and
1:54:14 just paste it in here see how it just
1:54:16 pastes it in as a document and now I can
1:54:19 say
1:54:21 um
1:54:24 oh
1:54:25 please
1:54:29 summarize the main
1:54:32 announcements and categorize
1:54:36 them
1:54:39 into consumer facing and
1:54:42 developer
1:54:47 announcements let's see how this does if
1:54:50 I'm away from gpts I get the
1:54:56 DTs thank you for uh for getting this
1:54:58 transcription digital
1:55:00 Gods So based on the transcript here are
1:55:02 the main announcements from the open AI
1:55:04 Dev day consumer public facing
1:55:06 announcements chat GPT now now uses GPT
1:55:09 for Turbo with improved capabilities and
1:55:12 the latest knowledge cut off the
1:55:15 annoying model picker is removed this is
1:55:17 actually doing pretty good um
1:55:19 introduction to gpts customized versions
1:55:21 of chat GPT that can combine
1:55:23 instructions extended knowledge and
1:55:26 actions and then launching a GPT store
1:55:28 later this
1:55:31 month um Revenue sharing for GPT
1:55:34 creators right so it it got the it got
1:55:36 the major things and then and then the
1:55:39 developer announcements gp4 turbo model
1:55:43 launched new
1:55:44 assistant API that's basically the
1:55:48 developer version of of the gpts new
1:55:52 modalities launched Vision speech text
1:55:54 to speech now available via API
1:55:57 everyone's been bitching about why
1:55:59 aren't they available via API now they
1:56:01 are um improvements to the fine-tuning
1:56:04 capabilities and custom model training
1:56:06 and then a micro Microsoft partnership
1:56:08 update Azure infrastructure improvements
1:56:11 tools that Sati nadala the the CEO of
1:56:14 Microsoft was kind of a he didn't really
1:56:16 have anything to say that that was
1:56:19 interesting it was good that he was
1:56:21 there but he didn't really have anything
1:56:22 to say so in summary major announcements
1:56:26 um yeah I'm trying to think what else
1:56:32 um can you make a
1:56:38 table of all of the specific feature
1:56:41 updates and the specifications for each
1:56:44 of the upgrades from a technical
1:56:48 perspective let's see how how how it
1:56:50 does with
1:56:57 this unfortunately the transcript does
1:57:00 not contain detailed technical
1:57:04 information however here's a table
1:57:06 summarizing the key technical
1:57:08 capabilities gp4 turbo if you by the way
1:57:11 if you've not seen one of these large
1:57:14 language models do stuff like this it's
1:57:16 pretty it's pretty something GPT 4 Turbo
1:57:19 up to 28,000 token tokens of context 300
1:57:23 pages of a book improved accuracy over
1:57:26 long context so it won't get so lost New
1:57:30 modalities Vision speech and text to
1:57:32 speech assistant API persistent thread
1:57:35 for long conversations that's a big deal
1:57:38 built-in retrieval system that's a big
1:57:40 deal code interpreter with python
1:57:42 sandbox we've seen that before but now
1:57:44 that's available in these little Bots
1:57:47 and then function calling with Json out
1:57:49 output and
1:57:51 chaining all all big deals like all that
1:57:54 all all that development stuff for
1:57:56 developers there there were third-party
1:57:59 projects that did all that stuff like
1:58:02 Lang chain and some of the other ones
1:58:04 but now it's within chat
1:58:07 GPT um fine tuning now available for GPT
1:58:12 3.5 GPT 4 fine-tuning is experimental in
1:58:17 Access and then pricing the GPT 4 Turbo
1:58:21 prompt tokens are three times cheaper
1:58:23 the completion tokens are two times
1:58:25 cheaper GPT 3.5 is also reduced in
1:58:28 pricing
1:58:30 so um let's ask
1:58:34 it who spoke at the conference and what
1:58:39 was their
1:58:48 topic
1:58:51 I was trying to do some photoshop stuff
1:58:53 today and amazed at the step-by-step
1:58:55 perfect instructions yep amazing all
1:58:58 right Sam Alman gave opening
1:59:00 remarks review of open ai's progress
1:59:02 over the this is quite quite quite
1:59:05 accurate Satia
1:59:08 nadela basically said we love you guys
1:59:12 Jessica
1:59:13 Shay demoed the zappier
1:59:16 GPT Raymond Sharma or Ramen Sharma
1:59:20 demo of the new assistant
1:59:22 API and then Sam mman did closing
1:59:25 remarks
1:59:27 yep pretty
1:59:29 swell all right I uploaded my L [ __ ]
1:59:32 prototype into the salon Meetup in case
1:59:35 you're trying to get filthy rich where
1:59:37 did you put it did you put it in uh in
1:59:39 Show and
1:59:41 [Laughter]
1:59:43 Tell let's see oh there's a bunch of
1:59:46 stuff in here this how I feel some days
1:59:49 chasing the AI Dragon quick take Jake
1:59:53 this what it's like trying to keep up
1:59:56 with this thing so so last night if you
1:59:58 weren't here um someone asked we were
2:00:02 doing something in Del and they asked if
2:00:04 if Del could combine an elephant and a
2:00:06 duck and so so I went into Dolly and I
2:00:09 was doing it and then Theus said the
2:00:13 eleu and so we just went off for like
2:00:17 half an hour 45 minutes on Ella the uck
2:00:21 and came up with a bunch of toys and it
2:00:25 was fantastic let's
2:00:27 see I don't see where did you put it oh
2:00:31 this was a this is a good one this was a
2:00:32 response to the elu which I guess was
2:00:35 the the
2:00:40 shant that thing's terrifying here is
2:00:43 there was the L
2:00:45 [ __ ] that was one of them that was
2:00:48 pretty good
2:00:50 um you must have done it
2:00:52 in AI
2:00:58 Irregulars oh whoever just said private
2:01:01 pepper love to beta test the new system
2:01:03 yeah we'll we'll probably start moving
2:01:04 some people over in the next week or
2:01:11 two oh that's really
2:01:15 cool
2:01:18 nice
2:01:21 man there's a lot of activity in the old
2:01:32 cell all right Kyle can you go over the
2:01:34 beginning of the transcript for a
2:01:39 visitor what do you mean for a visitor I
2:01:41 don't know what you
2:01:44 mean what do you mean
2:01:48 Joker
2:01:50 can you summarize the beginning of this
2:01:52 transcript for
2:01:54 someone who wasn't
2:02:04 there oh got to hit the return
2:02:12 [Music]
2:02:18 key
2:02:21 Sam mman welcome to attendees he
2:02:23 provided an
2:02:25 overview why is everyone spamming icons
2:02:27 okay so so for for those of you that are
2:02:30 are new
2:02:31 here so I do these things nightly I I I
2:02:34 tend to go live for two to three hours
2:02:36 every night seven nights a
2:02:38 week because I'm insane and then there's
2:02:41 a bunch of people that come in here
2:02:43 nightly also and they also hang out in
2:02:46 the salon Discord and they hang out here
2:02:49 and so so over the months there just
2:02:52 been a handful of inside jokes like one
2:02:54 of the the first inside joke that
2:02:56 started just started because I'm an
2:03:01 idiot and this
2:03:09 video just cracks me
2:03:13 up back here live at the waterfront
2:03:16 village with my friend the zombie
2:03:18 Jonathan you're looking good Jonathan
2:03:20 just got an awesome face paint job what
2:03:22 do you think I like turtles all right
2:03:25 you're great zombie so here Turtles
2:03:30 became a spirit animal of the of the
2:03:32 salon and then whenever I would mention
2:03:36 llama
2:03:37 2 um who was it it was
2:03:42 um Winston I think it was Winston would
2:03:46 always give me the Llama gift like you
2:03:49 know the gift where the little dude
2:03:50 comes up on screen and it's the Llama
2:03:52 doing this and so llama became the
2:03:55 second spirit animal and then when open
2:03:58 AI launched Dolly
2:04:01 3 um they used a hedgehog as as a as one
2:04:05 of the example drawings for showing that
2:04:08 Dolly 3 didn't suck and so the Hedgehog
2:04:12 became one I forget how we got gummy
2:04:14 bears now we've got the elu so all of
2:04:16 the little animals are just people using
2:04:19 those things catnip vigilante okay I'm
2:04:22 going to leave okay bye catnip VI
2:04:24 vigilante you can
2:04:28 stay um you don't you don't have to give
2:04:32 turtle or or llama things it's just it's
2:04:36 and and we are also insane yeah exactly
2:04:38 they're just nuts they just they're just
2:04:40 like playing here's the deal so if you
2:04:43 came here every night and had to listen
2:04:45 to my ADHD Rants and try to follow them
2:04:49 you get bored so they come up with side
2:04:52 games and they come up with all sorts of
2:04:54 things to do so it's just it's just part
2:04:56 of the community and what's what's been
2:04:58 really cool though is because we've got
2:05:01 the the Discord and we'll soon have
2:05:03 Mighty networks um there's an outlet
2:05:07 where as people are like as most of the
2:05:10 times I'm kind of demoing stuff during
2:05:12 the night and as I'm demoing stuff or
2:05:13 I'm cracking a joke people are
2:05:15 generating images and they're sharing
2:05:16 them on the Discord and there's kind of
2:05:18 a whole thing that happens so so that's
2:05:21 that's what these these meetings tend to
2:05:23 be
2:05:31 um let's see Insurance Toronto coming in
2:05:35 strong with the alien heads I like it
2:05:37 solid gummy bears cuz Kyle is in Denver
2:05:40 all right and he gets them from from the
2:05:42 dispensaries easily no that's what it is
2:05:45 the the the uh I get like Sher bear will
2:05:48 give me gummy bears a lot so so I think
2:05:51 it's just because of of the gummy bear
2:05:52 gifts that people give um that's right
2:05:55 and and yeah that's right oh well and
2:05:57 there's that I wouldn't call The
2:05:59 Elephant our spirit animal although the
2:06:02 El [ __ ] is absolutely our spirit animal
2:06:05 thanks Sher bear gummies is
2:06:11 tasty oh man it's all good it's all
2:06:18 good
2:06:37 I'm not printing my own com oh I'm not
2:06:38 pinning my own comments oh oh yeah well
2:06:41 you know I don't I don't know I just see
2:06:44 what gets pinned I'm not I I don't tend
2:06:46 to pay attention to who pinned it or how
2:06:48 how it all that sort of stuff so I'll
2:06:51 I'll get there eventually I'll figure
2:06:52 out this Tik Tock thing that's one of
2:06:55 the things that we hear pretty regularly
2:06:56 here is is just like oh is this the
2:06:58 elderly teaching each other how to use
2:07:03 computers oh
2:07:09 man all of our hands were up Drew
2:07:12 learning thank you so much sir
2:07:14 appreciate
2:07:18 that
2:07:24 yeah it was busy night tonight there was
2:07:25 there was lots of people in here I have
2:07:26 a comic book character I've designed
2:07:29 with I didn't do that
2:07:31 wait what was the comic book
2:07:35 comment that I've designed based on
2:07:39 them life's tough it's tougher when
2:07:42 you're
2:07:44 [Music]
2:07:45 stupid catnip vigilante I like that name
2:07:48 that's weird rhinos and elevant are two
2:07:50 of my favorite animals very
2:07:55 cool they will all be elderly before
2:07:57 they know it it goes by so fast tell me
2:07:59 about it so I got 23-year-old kids I'm
2:08:03 like oh
2:08:05 God twins based on the elephant
2:08:10 character oh based on the elephant
2:08:12 character oh cool oh that's your that
2:08:15 shark is is from a comic book of yours
2:08:19 wait let me go look at that
2:08:21 again cuz that thing's so cool where did
2:08:25 I just go oh I went to a different
2:08:34 server yeah look
2:08:36 at how cool is
2:08:39 that so that's you catnip
2:08:42 vigilante that's super
2:08:45 cool that one is [ __ ] terrifying
2:08:50 in can you
2:08:53 imagine if that's how Evolution played
2:09:02 out L shark
2:09:05 exactly all the Gen xers get that one L
2:09:09 shark um have you mentioned the change
2:09:12 oops nope oh that's not yours catnip
2:09:14 vigilante who I whose is that that's
2:09:16 cool wait I can I can see
2:09:19 Oh that's
2:09:27 steo oh I see I got
2:09:30 it thanks Joker I got it
2:09:34 now all
2:09:37 right have you mentioned the change I'm
2:09:40 just seeing in chat GPT with the explore
2:09:43 Tab and the new
2:09:45 design um I well we've been talking all
2:09:50 night about the video of Sam Altman
2:09:53 talking about all the new [ __ ] that's
2:09:54 going to be in chat gbt and then it's
2:09:56 not there so so uh we can't really I
2:10:00 want to go play with all the new [ __ ]
2:10:02 but I don't have access to it CU you
2:10:04 know what are you gonna do it just ain't
2:10:07 they oh wait oops something went wrong
2:10:10 that's
2:10:11 promising that's actually promising
2:10:15 because that says to me they might be
2:10:16 updating something so let's
2:10:20 let's let's go back
2:10:24 here ah let see chat.
2:10:31 open.com that's still the same let me
2:10:33 log
2:10:35 out log
2:10:44 in
2:10:47 now still old [ __ ] old as in like the
2:10:50 stuff that was brand new like four weeks
2:10:56 ago oh man can I use AI art app to bring
2:11:01 my comic book illustrations to
2:11:06 life
2:11:08 um to a degree to to the extent that you
2:11:12 want things really really
2:11:15 predictable there's a new there's a new
2:11:18 tool that I just got access
2:11:21 to I can't remember what it's
2:11:28 called if it depends how geeky you are
2:11:31 catnip vigilante like I I would say just
2:11:34 wait a little bit cuz right now one of
2:11:36 the challenging things with image
2:11:38 generation is
2:11:40 consistency of characters across
2:11:43 scenes there's ways you can do it with
2:11:46 do within within chat GPT
2:11:49 um by asking for seeds and things like
2:11:52 that um what about the sketch to Art one
2:11:55 um yeah you could maybe use the sketch
2:11:57 to Art one which is in um clipd drop.
2:12:04 Co
2:12:06 um clipd drop.
2:12:09 Co and within clipd drop. Co there's a
2:12:13 thing called stable
2:12:15 doodle but I think you actually have to
2:12:17 just draw this I don't think you can
2:12:18 upload something here can you
2:12:23 no so like if I did let's
2:12:27 see an
2:12:30 elephant with
2:12:32 a duck
2:12:45 bill and let's see big old elephant
2:12:52 legs all right and then I'm going to
2:12:54 [Music]
2:12:56 say an elephant crossed with a
2:13:03 duck upgrade to Pro do I need to upgrade
2:13:06 to Pro to do that
2:13:13 oh it's exhausting okay well I guess I
2:13:16 can't do that oh wait I need to sign in
2:13:19 am I going to lose my beautiful elephant
2:13:22 here they better not take away my
2:13:27 elephant
2:13:32 no I guess I can't do it um you can do
2:13:35 it there
2:13:37 um if you're geeky you could use control
2:13:40 net and stable diffusion and then you
2:13:42 can get some decent predictability
2:13:43 across things um what's the biggest
2:13:46 thing oh wait what was that what's the
2:13:48 biggest thing that people are missing to
2:13:50 understand with
2:13:57 llms I think most people just aren't
2:13:59 trying them I think that I think the
2:14:03 thing I think the thing most people are
2:14:05 missing with llms is they're not playing
2:14:08 with them long enough to have their
2:14:09 Kevin McAllister
2:14:11 moment that
2:14:14 moment um whatever these sites egg you
2:14:18 on and say to upgrade Pro an angel loses
2:14:20 its wings I know I'm right there with
2:14:22 you Abbott skinny like I feel like they
2:14:24 should all just give you some amount of
2:14:25 tokens a day like like
2:14:47 leonardo.da make their own cartoons oh
2:14:50 don't ask me I disagree with that I
2:14:52 think as soon as as soon as someone
2:14:54 cracks being able to make consistent
2:14:57 characters across across
2:15:00 storylines
2:15:02 um oh I think I I think that Tool's
2:15:05 going to be here the the problem the
2:15:08 problem with almost any piece of
2:15:10 software right
2:15:11 now is that let's say I spent the last
2:15:14 year and a half tweaking stable
2:15:16 diffusion and and just figuring out
2:15:19 perfect models and lauras and things
2:15:21 like that and control net and I built a
2:15:23 whole application around stable
2:15:24 diffusion to do consistent characters
2:15:27 and I get it dialed in and I just I'm
2:15:29 just about to launch it and then stable
2:15:31 diffusion launches consistent character
2:15:34 maker or Del does now we do really
2:15:37 consistent characters without you having
2:15:38 to try that's am I familiar with Rags
2:15:41 yeah um so retrieval augmented
2:15:44 generation is a is a big big deal and a
2:15:47 lot of what what um open AI announced
2:15:50 today is basically like rag for the
2:15:54 masses so so yeah it's it's coming
2:15:58 didn't Dolly do that for us with Larry
2:16:00 it did Silver Fox but like I haven't
2:16:03 gotten it I haven't gotten it to really
2:16:06 consistently do a
2:16:09 character like that's just clearly that
2:16:12 character it it it tends to like Veer
2:16:14 and deviate now what I haven't played a
2:16:17 lot with one of the things you can do in
2:16:19 Del is you can ask for the seed of an
2:16:21 image and then you can say use the same
2:16:24 seed for other images and that's
2:16:26 apparently doing it okay but it's still
2:16:28 not perfect so I just uploaded one of my
2:16:31 characters into Discord
2:16:34 cool in what channel in Show and
2:16:37 Tell
2:16:39 vigilante oh dude that's
2:16:47 beautiful that's super
2:16:50 cool you could do tarot
2:16:54 decks
2:16:57 sweet
2:16:59 amazing V cool that
2:17:02 cool thank you for sharing that
2:17:07 sir yeah Herod D you you guys should
2:17:10 hook up Herod does some pretty cool
2:17:12 stuff too um who's in here all right
2:17:15 what parts of llm apps will all always
2:17:18 need developers or data
2:17:22 scientists
2:17:23 [Music]
2:17:24 um I think I think anyone working on the
2:17:27 frontier models right so if you're
2:17:29 working on building those big base
2:17:31 models you're going to need data
2:17:33 scientists for that I think you're going
2:17:34 to need data scientists
2:17:38 for security infrastructure like AI Ops
2:17:42 is what uh Rachel Woods calls it so on
2:17:45 Tik Tok if you follow the AI exchange or
2:17:47 Rachel Woods she talks a lot about AI
2:17:50 Ops so I think you're going to you're
2:17:52 going to need um people that know what
2:17:55 they're doing about you know which
2:17:57 models do you use how do you use them
2:17:58 how do you fine-tune them how do you do
2:18:00 embeddings how do you do rag right
2:18:04 [Music]
2:18:05 um so so I think within enter within
2:18:08 Enterprise like Enterprises are going to
2:18:11 be struggling for a long time to figure
2:18:13 out how to do this stuff um so so I
2:18:18 think either you're working on the sort
2:18:19 of very foundational parts of machine
2:18:22 learning right so the GPT 5S GPT 6s GPT
2:18:26 7s and then you know Google's version of
2:18:28 that and right so there's all that you
2:18:31 know or you're working in in big
2:18:33 companies I think a lot of
2:18:37 the sort of Base like like like
2:18:42 highlevel application development I
2:18:44 think that's going to get democratized I
2:18:46 think Chad open AI did that that
2:18:49 today these gpts that we're going to be
2:18:51 able to create because we we can call
2:18:54 actions and they can do you can upload
2:18:57 data to them and and we can do Advanced
2:18:59 Data analysis within this
2:19:02 GPT um mere
2:19:05 mortals are going to be able to create
2:19:08 very sophisticated applications and
2:19:10 launch them and at some point so right
2:19:13 now the way it sounds like they're doing
2:19:15 the the GPT store is if your GPT is
2:19:19 popular you'll get some Revenue sharing
2:19:22 I would assume at some point they're
2:19:23 going to make it so that you know
2:19:26 McKenzie or Kyle Shannon as a consultant
2:19:29 I can put a model out there in charge
2:19:31 for it so you know it's private unless
2:19:36 otherwise subscribed to that's that's
2:19:39 not what they're doing right now but I
2:19:40 would imagine that that's
2:19:43 coming do I think neural Nets are the
2:19:45 future of AI or do you think a
2:19:49 more interp wait more interpretable
2:19:52 models will
2:19:54 be what parts won't be democratized I
2:19:57 don't know AI put or AJ
2:20:00 put I think I
2:20:03 think the my my phrase right now is the
2:20:06 democratization of specialization any
2:20:10 specialty
2:20:12 execution um like programming like
2:20:15 writing like illustration like
2:20:18 music composition anything any tactical
2:20:22 execution I think is going to be
2:20:25 democratized every single one of them to
2:20:27 some
2:20:28 degree and so what that means is you're
2:20:30 going to have people who are not experts
2:20:32 in a
2:20:33 field doing things in that field that
2:20:37 are that people in that field are going
2:20:39 to be like you can't do that and then
2:20:41 because they don't know any better
2:20:42 they're just going to do it and they're
2:20:43 going to innovate in that field I think
2:20:45 we're going to see a lot of that it's
2:20:46 hard to know what to focus on with the
2:20:48 CH changing landscape Earth rocker
2:20:50 that's that's the it it was funny when
2:20:53 we were doing when we were doing the
2:20:55 live stream of the open AI thing today
2:20:58 it was exactly that
2:21:01 where I think everyone was like really
2:21:03 excited about all the new stuff but like
2:21:07 a lot of the new stuff is like a
2:21:09 completely different way of thinking
2:21:12 about chat GPT and what you what you're
2:21:14 supposed to do with it like for the past
2:21:17 year chat GPT was pretty simple you
2:21:20 typed in words and you got back words or
2:21:24 if you were fancy you typed in words and
2:21:26 you got back code or you got back a
2:21:29 table
2:21:30 wo and then it was like oh you typed in
2:21:34 words and you got back pictures oh okay
2:21:37 we can do that now oh and now it's like
2:21:40 you don't even need to type in words you
2:21:42 can just upload a picture and have it
2:21:44 describe the picture and then take that
2:21:46 description and make a picture and then
2:21:49 you can put in words and it will
2:21:52 generate a little mini computer program
2:21:54 and write Python and execute like like
2:21:56 so in the past two months it's been
2:22:00 morphing
2:22:01 anyway and then today with this idea of
2:22:04 gpts and and all of the stuff that
2:22:07 developers now have access to I don't
2:22:09 where do you even
2:22:11 begin like like it's going to be
2:22:14 trivial to make an applic with computer
2:22:19 vision that you can just like take out
2:22:21 your phone camera and like shoot [ __ ] in
2:22:23 the world and have this app do some
2:22:26 crazy ass thing with it and tie that to
2:22:29 a Google map and tie that to a travel
2:22:32 site
2:22:35 and so I don't I don't know where to
2:22:37 focus right now I'm I'm right there with
2:22:40 you
2:22:41 like I ain't got no
2:22:44 idea I put up what the new interface
2:22:46 looks like in I news and events in
2:22:48 Discord does that mean you have access
2:22:50 to it should I be jealous I probably
2:22:54 should oh here we
2:23:01 go oh that's interesting
2:23:05 okay so let's see can I zoom
2:23:08 yeah all right so this is new interface
2:23:10 oh wait I can there's probably a better
2:23:12 way to do this hold
2:23:16 on who was it that just put that up Kyle
2:23:19 oh quick take Jake all right thank you
2:23:20 quick take Jake quick take Jake for
2:23:23 doing this okay so here's here's a good
2:23:27 example of um generosity by the way um
2:23:31 here's the new layout of the updated
2:23:33 chat
2:23:40 GPT yeah they they stole that interface
2:23:44 from
2:23:45 um they stole that interface from po it
2:23:49 looks like the explore thing okay so in
2:23:52 the in the new chat GPT that I don't
2:23:54 have access to yet but someone does
2:24:00 so that means create a new chat the
2:24:03 explore button this is these are going
2:24:05 to be my gpts right so that's just like
2:24:08 the explore button on
2:24:10 PO which if we go to
2:24:13 PO see the explore button right here and
2:24:17 we click on explore and now here's all
2:24:19 these Bots right so there's writing Bots
2:24:23 and here's professional
2:24:25 Bots and here's history Bots right so
2:24:29 it's going to be a similar sort of thing
2:24:30 so it looks like they're looks like
2:24:32 they're stealing some of that and then
2:24:34 there's my chat history do we have a
2:24:36 search in the chat history yet please no
2:24:40 h [ __ ]
2:24:43 okay how can I help you today so there's
2:24:45 our little
2:24:48 oneclick prompts you can turn them on or
2:24:50 off web access you can turn it on or off
2:24:54 neat and then your
2:24:59 settings and I guess that's your upload
2:25:07 button and then
2:25:14 oops and then this are here's my
2:25:19 gpts so look create a GPT right so
2:25:23 you'll just create one so this is very
2:25:24 much like Po so if you've made a if you
2:25:27 made a poot recently this is very much
2:25:30 the
2:25:33 model so there's a data analysis GPT the
2:25:36 chat GPT classic look the latest version
2:25:38 of GPT 4 with no additional capabilities
2:25:41 as a standalone
2:25:43 GPT game time I can quickly explain
2:25:46 board games card games to players at any
2:25:48 age let the games begin super cool
2:25:52 creative writing coach The
2:25:55 Negotiator Visionary painter of digital
2:25:57 wonder the cosmic dream so there's an
2:25:59 image generator this is very much like
2:26:02 Poots except these are more these are
2:26:04 more sophisticated and can do can do
2:26:07 fancier fancier stuff if you know what I
2:26:10 mean thanks for sharing that quick take
2:26:12 that's pretty cool that's super cool
2:26:15 that's super cool here we are AI learn
2:26:18 lab let's let's kick ourselves a little
2:26:21 a little field goal shall we walking in
2:26:24 Virtual infrastructure wait walking in a
2:26:27 virtual infrastructure in real world to
2:26:30 understand the problems before
2:26:32 construction yeah that's so a good buddy
2:26:35 of mine named Richard Boyd does does a
2:26:37 lot of simulation stuff hang on champy
2:26:40 found an evil person
2:26:42 hold what's going
2:26:46 on
2:26:48 what's going on what are you barking
2:26:59 [Music]
2:27:00 at all right let's see let's close that
2:27:08 down where are
2:27:10 we I'm
2:27:16 lost
2:27:33 all right let's see the
2:27:36 next page is the chat GPT verified gpts
2:27:40 yeah that that yeah that looks pretty
2:27:43 cool it's a nice simple interface I like
2:27:46 it cuz it was getting a
2:27:48 little it was getting a little Microsoft
2:27:51 wordlike in its overly
2:27:58 complicatedness I click on the gpts
2:28:00 right now and it errors out so I got the
2:28:03 golden ticket to a broken
2:28:05 theme theme
2:28:07 park yeah so we listen we'll um they
2:28:12 they announced that they were giving all
2:28:14 this stuff out today so the fact that we
2:28:16 haven't seen today I guarantee you the
2:28:18 Open Eye Engineers are up right
2:28:21 now trying to trying to get [ __ ] working
2:28:25 um they're probably blowing stuff up
2:28:27 left and right over there so it's
2:28:28 probably going to be up and down all
2:28:30 night so I I would I would guess that by
2:28:32 Thursday or Friday we'll have some
2:28:34 decent
2:28:46 stability I think we need a GPT for past
2:28:49 AI learning
2:28:53 Labs
2:29:00 so depending on what the depending on
2:29:10 what what the YouTube API makes
2:29:13 available or depending on what you could
2:29:16 call about YouTube from
2:29:19 zappier we could probably make an AI
2:29:22 learning lab bot that not only calls
2:29:26 what I was talking about but like you
2:29:29 could ask for a certain thing and then
2:29:31 it would jump you to that part of the
2:29:33 video I bet I bet that's
2:29:36 possible that that I mean that you know
2:29:39 what's interesting that could be an that
2:29:41 could be a really interesting project to
2:29:43 collaborate on in the
2:29:45 salon um
2:29:48 um yeah that could be a fun one just
2:29:51 just to learn how to do it learn what's
2:29:53 possible walking in a virtual
2:29:57 infrastructure the past AI learning lab
2:30:00 along with tonius spreadsheet yeah yeah
2:30:04 exactly yeah that would be really good
2:30:06 yeah here's all the stuff to go check
2:30:08 out and then if I check something out
2:30:11 where were all the times that that was
2:30:12 talked
2:30:14 about I found an uploaded I found and
2:30:17 uploaded the elephant sketch a few
2:30:19 minutes ago FYI I need to finish it
2:30:22 though oh cool let me
2:30:32 see oh that's
2:30:36 cool that's super cool so this everybody
2:30:40 is what actual drawing looks like not
2:30:43 not AI drawing although dude you know he
2:30:46 could could do you could upload all
2:30:48 these sketches you could train a model
2:30:50 train a model
2:30:52 tusil um yeah you could train a model to
2:30:55 to have DOI or mid Journey You Know
2:30:59 sketch like you
2:31:01 sketch learn how to do I will
2:31:05 collaborate yeah so join Vanda join the
2:31:08 uh join the salon if you haven't yes
2:31:11 great
2:31:15 uh relive the battle the GPT no I
2:31:23 don't I want to hide from the battle I
2:31:27 there will be no more battles there
2:31:29 won't be no more battles in the ai ai
2:31:31 learning lab or the AI Salon it just oh
2:31:36 no it was not good was not
2:31:39 good um did you see my kylin champy pick
2:31:42 and Show and Tell No let me go see that
2:31:46 this is cool Ted looks at life look at
2:31:48 that one that's
2:31:49 gorgeous dag
2:31:52 dag Ted slant it where's Champion
2:31:59 oh all the ghost
2:32:01 stories trying to get back to Dev day
2:32:04 presentations but the dog has other
2:32:06 plans that's
2:32:12 it picture hanging in my office oh
2:32:15 that's cool did you do that one lunar
2:32:17 stick is that like actual art medium on
2:32:21 paper that doesn't look like a digital
2:32:26 print oh that's
2:32:28 cool yeah there's cool stuff cool stuff
2:32:31 in the
2:32:33 uh in the Discord it's it's really
2:32:38 cool
2:32:45 wow
2:32:47 oh yeah Peter Kaminsky doing the Orton
2:32:52 effect look at those
2:32:54 images
2:32:57 gous gous gous gous can I make my own
2:33:01 app that learns my style then converts
2:33:05 all my sketches to the
2:33:07 correct
2:33:11 prompt you can so if you go to
2:33:15 leonardo.da you can up you can create
2:33:17 your own model at
2:33:45 leonardo.da
2:33:47 so if you want to hang out tomorrow
2:33:51 night we've got the AI
2:33:52 Salon um go
2:33:58 to they
2:34:07 salon. go to that
2:34:11 URL and
2:34:13 um the second link is the is the
2:34:18 the thing for the Meetup don't kill the
2:34:19 Discord we're not going to kill the
2:34:21 Discord but we are migrating guilds and
2:34:24 most of the convers most of the
2:34:26 community is migrating to Mighty
2:34:28 networks but the discord's going to live
2:34:31 the one of the next things we need to do
2:34:33 over this next week is figure out what
2:34:34 in the Discord is staying and what's
2:34:37 going and what's going to live in Mighty
2:34:38 networks so we're not going to kill the
2:34:40 Discord but it's going to just be it's
2:34:42 just going to be focused like like
2:34:44 you'll be if if you want to like
2:34:48 if we want to start creating things like
2:34:51 gpts that do stuff with Discord and and
2:34:55 do stuff with mid Journey like there's
2:34:56 going to be cool things we can do so
2:34:58 there's a reason to have the Discord
2:35:00 there so I think a lot of the the
2:35:02 geekier parts of the community can stay
2:35:04 on Discord and then the less geeky
2:35:08 people can move to Mighty networks and
2:35:09 things like that and and there'll
2:35:10 probably be some that bridge
2:35:13 both so yes good input thank thank you
2:35:17 um what's your top suggestion for a
2:35:19 daily or weekly summary of AI tools
2:35:21 created by others I I gave up on I gave
2:35:25 up on any I just kind of go to I go to
2:35:29 x.com I've got a couple of lists that
2:35:32 just filter AI [ __ ] I follow a handful
2:35:35 of people I follow here