aiLL 7/13/23 - The AI News Roundup Strikes Again!

Description

Chapters

Transcript

0:00 [Applause]
0:01 foreign
0:35 welcome welcome welcome
0:38 see if I got my Internet working any
0:40 better that was a big long uh
0:44 keep going we're letting more viewers
0:46 know you're alive
0:47 yay thanks Tick Tock you're the best
0:51 oh Tick Tock I've always wanted a social
0:54 media service like you
0:56 There's No One Like You Tick Tock you're
0:59 the best
1:02 looks like my numbers are updating a
1:03 little more real time so maybe I'll
1:06 step stutter a little less
1:09 invite another host
1:12 more likes can attract viewers keep it
1:14 up keep whacking that like button let
1:17 people know we're here
1:19 and then I'll get started holy crap the
1:24 like the amount of news the amount of
1:26 enjoy your content thank you very much
1:27 yeah welcome everybody my name is Kyle
1:30 Shannon this is the AI learning lab
1:33 uh we get together pretty much nightly
1:35 lately because I'm you know
1:38 sociopath
1:41 and uh rather than Sim Racing for two or
1:44 three hours a night I've been going live
1:45 for two or three hours a night talking
1:46 about Ai and we we have fun hey Tobias
1:49 what's shaking we have fun here uh I
1:52 like to cuss a lot so if you don't like
1:54 cussing I wouldn't stay
1:56 um I get some Army I get a little smart
1:58 assy if you're gonna troll me at least
2:00 be creative uh
2:02 if you can dial in your trolls to be
2:04 somewhere between a compliment and a put
2:06 down those are the best ones those are
2:07 the those are the ones that make me
2:08 laugh
2:10 um I'm an entrepreneur I'm the founder
2:13 and CEO of a company called storybine
2:16 hello Stacy what's shaking hey I vampire
2:20 I know yeah for anyone on the East
2:21 coaster certainly Europe I guess Europe
2:24 is on the verge of waking up at this
2:26 point
2:27 um there's a lot going on if you've been
2:30 here before you'll notice my little
2:32 wall oh URLs has grown
2:36 um it's uh
2:38 there's a lot happening so I'll talk
2:41 about all that in a second
2:43 um
2:44 I've started 12 or 15 companies in my
2:46 day
2:48 I
2:50 was a was a an entrepreneur in the early
2:54 days of the World Wide Web I started one
2:55 of the first digital agencies company
2:57 called agency.com
2:59 and was Chief creative officer there and
3:01 so I am
3:03 I love being like right right as a
3:06 technology goes from you know the Geeks
3:09 territory to the rest of us territory
3:12 um that's when I like jumping on board
3:14 and so this generative AI stuff
3:16 November 30th 2022 that was when chat
3:20 gbt launched in my opinion that marks
3:22 the inflection point
3:23 from
3:25 it was for the Geeks before it's for the
3:27 rest of us now and the Geeks still get
3:30 to play don't get me wrong
3:33 um but holy crap this stuff is coming
3:35 fast and it's coming in a deeply
3:38 profound way so that's what this channel
3:40 is about if you have questions about AI
3:41 pop them in the questions below let me
3:44 explain all these Earls
3:51 there we go
3:54 um
3:57 Okay so
3:59 if you're new to this generative AI
4:01 stuff
4:02 if you've not played with chat GPT if
4:05 you if you haven't heard of chat gbt
4:06 that's okay a lot of people haven't
4:09 go to that top URL chat.openai.com you
4:12 can ignore these other
4:14 pile down here the four below it right
4:16 now I'll talk about those in a second
4:17 but go to chat.openai.com go there and
4:21 and start playing with it until you've
4:24 had what I call a Kevin McAllister
4:27 moment this moment from Home Alone uh
4:30 and when you have that moment that's
4:32 when your Journey Begins and just think
4:34 of it as an adventure
4:36 and if you start to feel things like
4:39 existential dread and like holy crap is
4:41 this thing going to take my job and is
4:42 this better at doing what I do than I am
4:44 then you're doing it right
4:48 and uh if if you have a simultaneous
4:52 feeling of oh my God this is the most
4:54 exciting thing I've ever seen and holy
4:56 crap I'm terrified
4:58 that's the appropriate feeling
5:00 um Okay so
5:02 for the past I don't know month month
5:04 and a half I've I've only had three
5:06 sites here chat GPT Bing and Poe I've
5:11 added claude.ai and bard.google.com for
5:15 those of you that have been here before
5:16 you know that the reason I didn't have
5:18 Bard on this list before is because it
5:20 sucked they released a new version of it
5:22 today or yesterday I don't know when but
5:23 it's good now it's like like
5:26 it's good it's just it's it's not the
5:28 same it's Google but it's actually got
5:30 good interface it's oh it'll analyze
5:33 images uh it does coding really well it
5:36 does tables really well some of the
5:38 language stuff I played with was pretty
5:40 solid
5:41 claude.ai
5:43 that is anthropics Claude 2 model which
5:46 is their competitor to gpt4
5:49 it's got a hundred thousand token
5:51 context window which basically means
5:54 somewhere between 50 and 75 000 words
5:56 someone told me today they they tried
5:58 uploading uh 75 000 words and it broke
6:02 but it didn't break it 50 so they might
6:04 have a different how they deal with
6:06 tokens might be different but it's a lot
6:08 you could you know the Great Gatsby was
6:10 50 000 words so you can upload an entire
6:12 novel to that thing and interact with it
6:15 bing.com is Microsoft's version of gpt4
6:19 it's you can generate images there and
6:22 then poe.com added the ability to upload
6:25 files so all five of these right now are
6:28 viable options for this large language
6:30 model stuff Claude is free
6:33 gpt4 here cost you 20 bucks a month
6:37 now up here
6:40 chat.openai.com you get access to this
6:42 thing called code interpreter which is
6:45 [ __ ] mind blowing However the fact
6:49 that you can now upload files to Claude
6:51 and to Bard and to Poe
6:54 you can do a fair amount of what uh code
6:58 interpreters doing but code interpreters
6:59 still leagues above what these other
7:02 things are doing with files but holy
7:04 crap
7:05 what a day then
7:08 oh wait so let me let me talk about two
7:11 other things so prompts.chat
7:13 is
7:15 um just a it's like a it's a Google doc
7:17 basically and it will teach you if you
7:20 haven't played with chat GPT or any of
7:22 these tools it will teach you about
7:23 prompting and what how the prompting
7:25 Works how these large language models
7:27 work and it also gives you a whole list
7:30 of examples of personas that you can
7:32 tell these things to act like like act
7:34 like an accountant act like a script
7:37 supervisor during the writer's strike
7:40 it's probably a little insensitive but
7:43 but yeah but you know writer's gonna
7:45 write uh but but you'll you'll really
7:48 get a good sense of how deep and
7:50 Powerful these tools are and that
7:52 they're Way Beyond
7:54 a Google search
7:56 um they're profoundly beyond that and
7:57 then futurepedia.io that's just a big
8:00 old list of uh
8:02 uh whatever you call it uh AI tools
8:05 because I can't keep up with them
8:06 anymore I'm just just trying to keep up
8:08 with where you can actually go start
8:10 playing with these things like all five
8:11 of these are viable options right now
8:14 and I and I haven't mentioned pi and I
8:17 haven't mentioned Elon musk's x dot AI
8:20 which there's something coming for that
8:21 tomorrow from that tomorrow so anyway if
8:25 you have questions about AI pop them in
8:27 the comments below uh you can ask the
8:28 people that are here I really do try to
8:30 answer them all I will get behind uh and
8:33 it will seem like I'm not answering them
8:34 and people are like I don't think he's
8:36 really live I think he's the bot I think
8:40 this is just a pre-recorded situation
8:43 people get very high and mighty in here
8:45 with theories about whether I'm live or
8:48 not
8:49 but if I weren't live could I do this
8:54 no
8:55 okay
9:02 [Music]
9:07 why is it called chat GPT
9:10 well because it's kind of a chat bot I
9:14 hate so I think chat gbt is a bad name
9:17 but everyone knows it now so tough I
9:19 think we're stuck with it I think it's
9:20 going to be like Xerox or just that's
9:22 what people call these things
9:25 um
9:26 AI chat Bots for the past decade have
9:29 been the bane of my existence because
9:31 everyone's like oh chat Bots are gonna
9:32 be so big and chat Bots are exciting and
9:36 chat Bots were just like phone voicemail
9:39 trees that you typed into
9:42 because I I don't know what it is about
9:45 I have a particular talent for the
9:48 minute I'm typing into a pre-programmed
9:50 tree I I break it and it's like I don't
9:52 know how to answer you so chat Bots were
9:56 a [ __ ] disaster they were they were
9:57 incredibly labor intensive to create
10:00 you know they basically had to
10:02 anticipate all of the scenarios and
10:04 write all of the copy for all of the
10:06 scenarios they were incredibly
10:08 um intensive to write now with chat GPT
10:11 you can actually have those
10:13 conversations so I think that's why they
10:14 called it that the GPT part is
10:17 generative pre-trained Transformer
10:19 Transformers the technology that Google
10:21 invented that everyone else is using to
10:23 do all this remarkable [ __ ] pre-trained
10:26 means it's pre-trained in this in the
10:28 case of these things on essentially all
10:30 of the internet all of the public
10:32 internet and generative means it's
10:34 generating stuff so I wish it had a
10:36 better name but it doesn't what are your
10:38 thoughts on the news agency like
10:40 Associated Press partnering with the
10:41 open AI so I've got like 14 tabs on my
10:45 browser just as people are talking about
10:48 this that's a huge [ __ ] deal so so if
10:51 you so if you don't know uh open Ai and
10:54 the Associated Press announced a
10:56 partnership today I think it's a huge
10:59 deal
11:00 open AI removed
11:04 Chachi PT's connection to the internet
11:06 five days ago now six days ago and they
11:09 haven't put it back and they removed it
11:11 because
11:13 uh it was pulling content from behind
11:16 paywalls
11:18 so it was it was effectively publisher
11:21 publishing
11:24 um
11:26 uh you know content that should be paid
11:30 for for free so they pulled it down
11:33 and then today they announced a deal
11:35 with the AP I would not be at all
11:36 surprised if they announced a deal with
11:38 time Inc and probably some other big
11:40 Publishers my my suspicion here is that
11:46 um
11:48 this is total speculation
11:50 my suspicion is Microsoft is running
11:52 interference with that
11:54 and they're saying that they're
11:55 basically saying to open AI hey we love
11:57 your technology we're transforming our
11:59 company around it but you know let's not
12:02 go ahead and piss off Microsoft's
12:04 biggest clients so why don't we start
12:06 cutting some deals with our major
12:09 clients and major Publishers uh who have
12:12 a vested interest in this so my my
12:14 instinct is this AP deal is the first of
12:17 many
12:18 um
12:20 I think companies that cut these kind of
12:23 content deals are going to be in a cat
12:26 bird seat right because if chat GPT is
12:29 the place where you can go chat GPT
12:32 official AP stories then lots of people
12:35 are going to trust that lots of people
12:37 are going to use that so I think it's
12:38 absolutely fascinating
12:41 um if you go look at the history of the
12:44 Edison cylindrical record player and the
12:47 flat one
12:48 Edison's was technologically Superior
12:52 but he didn't cut content deals the flat
12:54 one technologically inferior but he cut
12:57 content deals with like the most famous
12:59 opera singers of the day and that's why
13:02 we have flat record players rather than
13:04 cylindrical ones probably it was also
13:06 probably more convenient but whatever
13:09 um so content's really important so I
13:11 think this is a huge huge deal yeah
13:13 that's one of the many things like just
13:15 the news today and and the fact that we
13:18 went from essentially
13:20 two major players and kind of a third
13:22 one [ __ ] around where you kind of had
13:24 Claude models at poe.com but you had to
13:26 go to Poe which didn't make any sense
13:28 now the fact that they rolled that out
13:30 and then Bard all of a sudden just
13:32 doesn't suck today
13:33 something whoa
13:36 good Lord I checked out Bard when they
13:39 announced it it's greatly improved I
13:41 know right Jenna it's like like I mean
13:44 every time I went to use Bard before
13:45 where they're like there's a new version
13:47 of Bart I'm like awesome what's oh
13:49 that's a piece of stinky [ __ ]
13:51 and uh not today not today so
13:58 I don't I don't I can't I say
14:02 I can't keep up with this [ __ ] anymore
14:04 all right are people just too
14:06 disconnected with themselves they need
14:09 an AI to talk to well
14:11 wait carp carpetrum carpetrum carp
14:18 carpe diem
14:19 I don't know whatever your name is
14:22 um
14:24 no so so these are not just chat Bots
14:27 like you if you haven't played with one
14:31 go there now go to go to prompts.chat
14:33 and read about what prompting is and and
14:36 how these things work and go to go to
14:38 chatgpt or Claude or bard.google.com go
14:41 to any of them and start start playing
14:44 with them
14:45 this is not just like hey Fred how you
14:48 doing oh I'm great Kyle how are you boy
14:50 you know I'm really lonely but now with
14:52 my fake chat bottom even better it's not
14:54 that I mean
14:56 there's going to be digital
14:58 companionship fueled by these things
15:01 but these things write screenplays these
15:04 things which is one of the reasons The
15:06 Writer's Guild is striking right now
15:07 these things code these things write
15:10 marketing plans business plans marketing
15:13 copy they're they are remarkable they
15:16 are remarkably powerful tools and
15:20 um
15:21 so so I would strongly encourage you to
15:24 go start playing with them start to get
15:26 educated about what they are and what
15:28 they make possible and and once you
15:30 start to get your head around it take
15:31 the whatever you're good at and have it
15:34 start doing that for you or being an
15:37 assistant for you to do the kind of
15:39 stuff you know about and that's when
15:41 your mind will probably go
15:45 and that's the moment when the adventure
15:47 begins where you're like holy [ __ ]
15:49 really
15:52 is this gonna take my what do I how do I
15:56 I've been doing this for 30 years
15:59 what do I do now
16:03 that's what we're all having to deal
16:05 with
16:05 all right in the time you were gone Bard
16:08 planned my anniversary trip with great
16:11 extras impressed that's awesome Silver
16:13 Fox oh my God that's hilarious yeah so I
16:18 I think I I mean this is this is really
16:22 really fascinating stuff because you
16:24 know chat gpt's been in a cat bird seat
16:27 for for seven months now
16:30 December to July
16:32 and
16:35 while there were other things out there
16:37 there they just seemed on the edge and
16:40 then all of a sudden there were a couple
16:41 of chinks in the armor right there there
16:43 were you know they pulled back
16:46 browsed with Bing they pulled back the
16:48 connection to the internet there was
16:50 another lawsuit that was announced
16:53 um
16:55 and then
16:56 um inflection comes out
16:59 which is that's Pi dot AI if you want to
17:02 go play with the inflection version of
17:03 these things
17:05 and that's Reed Hoffman who's the
17:07 founder of LinkedIn and and the guy I
17:09 forget his name who was one of the
17:10 founders of deepmind created inflection
17:12 but Microsoft also invested big into
17:16 inflection so so Microsoft not only is
17:19 invested into open AI but they're
17:21 invested into inflection so all of a
17:23 sudden it's like oh you know maybe it's
17:25 not such an exclusive deal you know
17:27 maybe they're not so hitched at the hip
17:29 and Microsoft's hedging its bets so it's
17:32 just like
17:33 it's just amazing like the the I feel
17:35 like we're in the middle of a power
17:37 swing right now where
17:39 we're gonna find out
17:42 a month or two from now like this list
17:45 in my opinion probably goes back to
17:47 three I've only had three on there now
17:49 I've got five
17:50 my gut tells me this list goes to six or
17:54 seven
17:55 and and it's just boiling for a while
17:57 and then it'll kind of you know it'll
17:59 sort of drop back to a top three but
18:02 we'll see that's amazing
18:04 okay let's see what else you got here
18:07 you're not live definitely not Scott
18:09 thickner
18:10 759
18:13 well Mr thickener
18:16 it's Scott thickner
18:19 thicker than you know
18:21 759. oh yeah
18:24 yeah no I'm definitely not live Scotty
18:27 definitely not live and certainly not
18:29 qualified we've talked about that for
18:32 days in a row now
18:37 okay
18:46 hello Apocrypha I know what that means
18:50 now
18:51 you're not live yes I am Scott please
18:55 stop antagonizing me it's very hurtful
18:59 I'm so triggered
19:01 I hate being triggered by Scotty
19:05 Scotty's triggered me
19:07 um
19:10 I use Claude a f i for the first
19:13 time to give it a PDF of my syllabus and
19:16 update the dates oh wow did it do it oh
19:19 that's so cool did it let you redownload
19:22 it
19:23 that's pretty slick that's it see that
19:25 there's a really slick uh use case that
19:27 Professor Sonia put in there upload
19:30 something that has a bunch of dates in
19:31 it say change the dates to this year
19:33 starting on
19:35 whatever out the cub
19:38 that's super cool moving so fast it's
19:40 moving so fast I'm an information
19:42 hoarder and love chat GPT I'm addicted
19:44 that's that's so I'm I'm a total data
19:47 hoarder I've got somewhere here can I
19:50 have it nearby oh yeah right back here
19:53 I've got just just a suitcase full of
19:56 hard drives and I'm never gonna throw
19:58 them out
19:59 will I ever plug them back in I don't
20:01 know with these tools now we're going to
20:04 be able to take all the data we've ever
20:05 hoarded and just dump it into the data
20:09 bucket and start interacting with it
20:11 I took a uh I took a screenplay I wrote
20:14 years ago
20:16 and I had it in in PDF on on my hard
20:19 drive and I uploaded it to Claude today
20:23 and I just uploaded the screenplay
20:27 and then said um write a log line a
20:30 synopsis and a character sheet and it
20:34 just went and then it was perfect it's
20:37 like
20:38 it's just amazing like it's just I
20:43 they act as an assistant yeah that's
20:44 that's a really good point uh LR Patrick
20:47 it's um
20:49 a good way to think about these tools
20:51 someone was asking before like are they
20:53 chat Bots are they virtual companions
20:54 what what it's more like what what these
20:57 are very much like is like you've got an
20:59 idiot savant as your personal intern or
21:03 your personal assistant or like a
21:04 creative collaborator the idiot part is
21:07 it's completely dumb it'll just do
21:09 whatever you tell it and if you give it
21:11 shitty instructions it'll give you
21:13 shitty results right like a new intern
21:15 that doesn't know your business or
21:16 doesn't know who you are
21:18 so you got to tell it what to do you're
21:20 like okay intern Now You're Gonna Act
21:23 Like A Stanford mathematics professor
21:25 and it's like okay
21:26 and we're gonna I'm gonna put a formula
21:29 in front of you and you're gonna figure
21:31 it out for me okay
21:33 we'll figure it out for you
21:36 um so it's very much like that
21:38 and and I think I think where these
21:40 things go is toward being assistance for
21:44 us so I think that's actually where
21:46 they're headed
21:47 where we don't have to like right now
21:49 you have to do this thing called prompt
21:51 engineering and and I think one of the
21:53 things that you're going to start to
21:54 experience with these tools especially
21:56 is these guys start to compete is
21:58 they're going to make it easier and
22:00 easier and easier for people to put in a
22:03 minimum amount of input for a maximum
22:05 amount of output
22:07 like for example these things are are
22:09 very shortly going to start to remember
22:11 who we are what our preferences are
22:14 and and probably even
22:16 the history of all of our conversations
22:19 so that we can refer to conversations we
22:22 had two months ago
22:24 uh is is likely where these things are
22:27 going and probably pretty quickly what's
22:29 the best way to generate generate
22:31 t-shirt graphics
22:33 um you just need to start playing around
22:35 with
22:37 um
22:38 with your prompting and and using things
22:41 like
22:42 um
22:44 like graphic illustrations flat Graphics
22:46 things like that let me show you some
22:53 um
22:57 so go to openart.ai
23:04 prompt book
23:06 and
23:08 you'll see this thing here it's the
23:10 stable diffusion prompt book but you can
23:12 use this for mid-journey you can use it
23:14 for stable diffusion you can use it for
23:16 Adobe Firefly this is really just more a
23:18 primer
23:19 um
23:20 and it will this document is the best
23:24 thing that I've seen like this this
23:26 thing came out I don't know probably I
23:29 don't know when it came out but I
23:30 discovered it kind of the end of 2022
23:32 and I still haven't really seen anything
23:34 as clear as this so it's like a 96 page
23:37 PowerPoint presentation that just talks
23:40 about
23:41 these image generation tools it talks
23:44 about prompt engineering defines it you
23:46 know gives you a whole bunch of things
23:48 to do here's an intro here's how it
23:50 works here's prompt formats so like do
23:54 you want a photo or painting what's the
23:56 subject you want special lighting do you
23:58 want an environment do you want a color
23:59 scheme Point of View background is there
24:02 a specific style so this for you if
24:04 you're talking about doing uh T-shirt
24:07 graphics you want you know simple
24:09 graphic flat colors you can describe
24:12 that so as I go through here
24:16 there's modifiers things like it
24:18 understands photographic terms like so
24:20 there's a table of a bunch of different
24:22 photographic terms like you know tilt
24:24 shift and Polaroid and long exposure and
24:27 color splash and GoPro
24:30 it understands lighting it understands
24:33 art mediums so chalk graffiti
24:35 watercolors clay sculptures it
24:38 understands different artists and you
24:41 know the styles that they paint in
24:43 um
24:45 mixed media
24:48 um
24:49 so here you get into illustration right
24:51 so the different illustration types
24:53 and so these start to be things like
24:56 like there was one
24:59 this one here you know comic book style
25:02 you know starts to look a little
25:04 t-shirty and just you know so start
25:06 learning all of these different terms
25:09 that that that give you those examples
25:11 if you're using mid journey and I would
25:13 say I would suggest mid-journey
25:16 um
25:18 go to the mid-journey Discord that's
25:20 where you have to make the images anyway
25:21 and there's prompt libraries in there
25:23 and there's all sorts of people in there
25:25 I would imagine that there are you know
25:28 sub communities within the mid Journey
25:30 server that that are doing t-shirts and
25:33 can probably help you with that
25:35 so hope that helps and by the way
25:40 if you're new here do me a favor if you
25:42 would share the live with your folks if
25:45 you uh if you think it's worth sharing
25:47 and maybe we can get some more folks in
25:49 here
25:52 and we'll get this party still at it
25:55 I'd say we're decently started at this
25:57 point we're going we're rolling
26:00 we're rolling rolling rolling
26:17 got any predictions or thoughts
26:18 regarding the ftc's investigation into
26:21 open AI That's yet another news story
26:23 that dropped tonight
26:26 um not yet like it's just they've just
26:28 announced an investigation I listen I
26:35 when someone like open AI pulls a major
26:37 feature like browse with Bing something
26:39 that's co-branded with Microsoft and
26:41 they pull it
26:42 there's there's some [ __ ] boiling behind
26:45 the scenes that that we're not privy to
26:47 there's some big players
26:49 tossing around big lawyer money
26:53 saying we're gonna
26:55 [ __ ] some [ __ ] up and so um so it
26:58 doesn't surprise me at all that the FC
26:59 FTC is investigating them
27:02 um
27:04 I think what
27:06 what we're going to see for we're kind
27:09 of entering we're we're entering but
27:12 what I assumed we would see right about
27:14 now
27:16 um we're entering a kind of panic phase
27:17 about Ai and I think what's if you
27:23 listen to The Screen Actors Guild
27:27 um the speeches from them starting their
27:29 strike today
27:32 um one of the big things that they
27:34 talked about was
27:36 the studios came back and they said okay
27:39 they want anyone who's an extra in a
27:42 movie background actors ones without
27:44 lines
27:45 they want to be able to pay them for one
27:47 day Rate One Flat day rate and then use
27:50 their likeness in perpetuity for any
27:52 project in the future never have to tell
27:55 them never have to pay them again
27:58 um
27:59 so that's a fundamental
28:02 existential threat
28:05 to that business model so I think what
28:08 what you're going to start to see in in
28:11 all sorts of businesses from Consulting
28:13 to advertising agencies to creative arts
28:17 to and and everything in between and
28:19 everything
28:21 on the outside of that
28:23 is companies are going to start
28:25 realizing holy [ __ ] our business model
28:29 might not exist in a year or two
28:32 and one of the reactions to existential
28:35 threats is defensiveness right sue them
28:38 shut it down make them turn it off
28:42 so so
28:44 I think this is very powerful players
28:48 um
28:51 exercising their their
28:54 United States granted rights to sue the
28:57 [ __ ] bejesus out of everyone
29:00 um
29:01 as as ways to
29:04 hedge against the future they they
29:06 probably know is coming
29:09 um
29:11 I think that open AI is going to be in
29:14 good shape to survive it because they've
29:16 got Microsoft as a partner right this is
29:18 this is a major player in the economy
29:21 you know they've got major connections
29:24 they've got major lobbyists so so what I
29:26 suspect will happen is open AI will
29:29 start cutting
29:31 big licensing deals with big Publishers
29:33 and pay them some amount of the profits
29:37 um and then at some point they'll just
29:38 cut it off and say if you're below that
29:39 line you can come sue us but we've cut
29:41 all of our big deals so that's what I I
29:44 feel like is going on but I that's one
29:47 thousand percent speculation
29:50 is there a list of the best AI sites
29:52 yeah futurepedia.io well it's so it's
29:55 not the best so my list of sites right
29:58 here is this is if you want to get
30:00 started with
30:02 um generative AI just start with chat
30:06 GPT or or any of these other four
30:10 um
30:10 these are all viable options to to
30:13 really start to get your head around
30:14 what's possible
30:16 futurepedia.io is a list of just like
30:19 all the AI tools that are out there
30:21 there's probably four thousand at this
30:23 point they're like 3 700 at the
30:25 beginning of the week
30:27 um they're they're adding to it like an
30:30 insane amount of tools a day
30:32 um
30:34 there are some
30:36 um
30:37 like top 10 kind of kind of
30:39 services and content sites I don't
30:41 really follow them all but um I would
30:44 just go there and go for the top rated
30:46 ones the ones that have been there the
30:47 longest the ones that you know there's
30:50 gonna be a lot of stuff there that's
30:51 just crap but
30:53 you know you'll be able to find some
30:54 stuff there
30:55 what's the criteria for your list there
30:57 those are just the ones you're
30:58 discussing on live so so my criteria is
31:02 it used to be just three of them and it
31:04 was really just three different ways to
31:05 play with chat chat gbt
31:08 um because the the the top one and Bing
31:12 and Poe all have gpt4 in them but you
31:16 know Bing is free but it's not really as
31:18 good Poe has a bunch of other things
31:20 like Claude built into it and it's got
31:22 Bots which are really interesting and
31:25 then open AI is the inventor of chat GPT
31:29 so it's like it's just the best right
31:30 now and then literally in the past uh
31:34 like day and a half
31:36 cloud.ai launched and it's good it's
31:39 really good it's like close to chat gpt4
31:42 and it's the site's good the interface
31:45 is good and then Bard which was horrible
31:49 for a long long time now six months all
31:52 of a sudden they relaunched
31:54 bard.google.com and it's good it's
31:56 really good it's like so all of a sudden
32:00 there are now major competitors to chat
32:02 gbt so that's why this list is here
32:04 prompts.chat is just a place to teach
32:06 yourself prompting and then future PD is
32:08 just a big long list of tools so so
32:10 that's all I this is just here as if
32:13 you're wondering where to get started
32:15 it's why I put these here I'm not
32:17 affiliated with any of them I don't give
32:19 a [ __ ] if you use them I don't give a
32:20 [ __ ] if you buy subscriptions
32:23 um at some point someone some of them
32:25 should knock on my door and go hey you
32:26 talk about us a lot that's well here's a
32:28 Starbucks gift card but they haven't
32:30 done that yet
32:32 qualifications I don't need those
32:34 thinking qualifications
32:36 I know nothing I know nothing about AI
32:40 you stinky people nothing
32:45 is AI going to make our brains more lazy
32:48 after a while you know Rain Ninja by the
32:50 way good name good name uh good Avatar
32:54 um
32:56 I don't think so I think it goes the
32:57 opposite direction I've been
33:01 compulsively obsessed with generative AI
33:03 since probably
33:06 about this time last year it started
33:08 with a project called pixelmind then
33:10 stable diffusion then chechi BT came out
33:13 in November so I've been in this I have
33:16 never in my life been more creatively
33:19 alive and inspired and activated
33:24 and and it what and the reason is this
33:26 it's there's there's a counter-intuitive
33:28 thing that happens with these so
33:32 what they do and what people talk about
33:35 them doing is oh it wrote the paper
33:37 right so the human's just a lazy [ __ ]
33:40 and the Machine wrote the paper
33:43 it's it's not really how it works a
33:47 human being said I have a thing I want
33:50 to write and I want you to write it and
33:52 then lazy human beings just hand off
33:55 that piece of paper
33:57 but all the creatives that I know that
33:59 are using these tools are using them
34:00 more like a collaborative tool so what
34:03 I'm finding is that rather than having
34:05 writer's block or rather than having to
34:07 spend two hours getting to a first draft
34:09 I can get to a first draft in about 30
34:12 seconds that's about 80 percent of what
34:14 I want and as you get better at
34:16 prompting like it when I first started
34:18 working with this stuff I was getting
34:20 things that were consistently at sixty
34:22 percent of what I wanted and it was
34:23 really frustrating and it took me a
34:25 while and as you get better at prompting
34:27 you're like oh I have an idea for
34:29 something and you you figure out okay I
34:31 gotta write these four or five elements
34:33 into the prompt and it's weird because
34:34 you think like
34:36 I'm I'm writing as much copy into the
34:39 prompt is what I'm going to get out of
34:40 it but but you sort of you start to
34:42 figure it out
34:43 and then you get it right and what comes
34:46 out is like 80 85 90 of what you want
34:49 and then you just go in and tweak it and
34:51 edit it
34:52 so what it's doing for me is
34:56 you know how in any given day you've got
34:59 some amount of hours of creative output
35:02 and then the brain just shuts down and
35:04 for me it's probably somewhere between
35:06 four and six if I'm particularly hyper
35:09 focused on something I can go longer but
35:11 I've generally got like four to six
35:13 hours a day where I can I can crank out
35:15 some [ __ ] a lot of days it's a lot lower
35:19 before when I would work
35:22 I might spend three and a half of the
35:24 four hours I had for that day just
35:27 getting to the starting line
35:29 and what's happening now is I get to the
35:32 starting line in 10 or 15 minutes and
35:34 then I have like the rest of those four
35:36 hours
35:38 to like refine it or come up with new
35:40 ideas so so I'm finding that in the
35:43 space of the time I would have had
35:44 gotten one idea to a decent level of
35:47 completion I've now got two or three
35:49 things completed and like 10 10 new
35:52 things started so so it's completely
35:57 transforming my relationship with
35:59 writer's block and with creative block
36:03 and doing block you know as an Adder it
36:05 might you know I have zero executive
36:07 function I have a block of cheese for my
36:09 frontal lobe
36:10 um and so it's kind of taking over a lot
36:13 of that function and it's it's like a
36:16 [ __ ] magic trick it's amazing so so
36:18 no I think it goes the other way I think
36:20 we get more creative more self-expressed
36:23 more expert
36:26 but we're still the pilot it's the
36:29 co-pilot
36:30 so we still get to drive the thing can
36:32 other people get your files you uploaded
36:35 which files
36:37 which files that I upload
36:40 I don't know what you're talking about
36:41 Tobias let me know and I'll I'll figure
36:42 it out oh my God I have so many hard
36:44 drives going back to Windows 95 and
36:47 Macintosh though I know I do too I've
36:49 got a I've got
36:53 this was this was the uh the online
36:56 magazine I created in 1994 and this for
37:00 you gen xers
37:02 is a side quest drive and for those of
37:05 you that are too young to know what that
37:06 was I hate you but that is 45 megabytes
37:12 of hard drive platter goodness that you
37:15 would shove into a sidequest drive
37:18 and that's my uh
37:21 that was my
37:23 web magazine back in the 90s so yeah I
37:26 got scads of that [ __ ] am I getting rid
37:30 of it no
37:31 I might need it someday
37:34 you know that outline I wrote for that
37:35 article that we never published in 1995.
37:39 yeah I'm gonna need that someday
37:43 total digital hoarder
37:46 did you try prompting to pitch it to
37:48 next Netflix did I try
37:51 prompting what oh to pitch to pitch my
37:54 screenplay to Netflix because there's a
37:56 writer's strike I should I got seven of
37:58 them
37:59 [Laughter]
38:01 I don't I don't think I should be a scab
38:03 I've got too many people that are
38:05 that a wga and sag members I'm SAG
38:09 eligible I did I did uh I had a small
38:11 role in New Jack City back in the in the
38:14 80s is that 80s late 80s early 90s
38:19 um kind of like Google did for search
38:21 engines uh I think I know what you're
38:23 talking about there
38:26 uh
38:30 well welcome everybody we've got some
38:32 new people here my name is Kyle Shannon
38:33 this is the AI learning lab this list of
38:35 URLs right here is just there has been
38:38 an explosion of competition in the in
38:41 the chat gbt space in the past two days
38:44 so this went this used this was a list
38:46 of three sites the top part of it for I
38:49 don't know a month and a half and in the
38:52 past two days I've added two more sites
38:53 to it so go play with any of those the
38:56 top one is the official chat GPT website
38:59 um Bard is Google's version of it
39:01 claude.ai is anthropics Bing is chat GPT
39:05 but it's it's Microsoft's flavor of it
39:09 I watched the clip you shared on the AI
39:11 agency created by teens yeah to build a
39:15 product for free isn't that insane I I'm
39:17 telling you I've had two I mentioned
39:20 this in that video
39:21 but in the past week
39:24 I've had two senior people at Major
39:27 pharmaceutical companies like as I'm
39:30 talking about Ai and what we do at
39:31 storyvine and you know my thoughts on on
39:35 what's coming
39:37 have volunteered to me
39:40 we've gone to our agencies and said hey
39:42 this stuff's coming what are you going
39:45 to do about it and and the agency's
39:47 response was we'll come teach you how to
39:49 use it
39:50 but when the client said to the agency
39:53 well what are you going to do with it
39:54 they're like well
39:56 we we do what we do and you know you
39:59 don't need to be bothered with that
40:00 we're just going to keep charging you
40:02 for the 50 Junior people writing you
40:05 know social media copy at 150 an hour
40:10 clients are not going to put up with it
40:11 so there's like I just I just feel like
40:13 this this Reckoning coming
40:15 where where
40:17 it's just it's
40:20 it's every [ __ ] every time there's a
40:24 new technology like this
40:25 you know it first gets derided oh that's
40:28 stupid it's over hyped it's gonna kill
40:31 us it's dumb
40:34 I don't know man let's just
40:40 there's well I mean here so here's the
40:42 positive side of it there is such
40:44 opportunity right now to obliterate
40:48 big companies
40:50 who are gonna sit back and say I'll take
40:52 a wait and see attitude to this pesky AI
40:55 stuff
40:57 we've got a 30-year relationship with
41:00 that pharmaceutical company so uh that
41:03 relationship's not going in what what do
41:05 you mean we were what do you mean we're
41:06 putting review what what what's
41:08 happening
41:09 that's coming like and that's coming
41:12 everywhere so you're either working for
41:14 one of those companies that gets smacked
41:16 upside the head or you're one of the
41:19 companies that smacks them upside the
41:21 head
41:22 you know
41:23 Kyle can't believe you started without
41:26 me I know Troy right it was so rude of
41:29 me although I started like three times
41:31 tonight there was the first time the
41:33 thing but the app just crashed the
41:36 second time I wasn't ready I had to
41:38 change my shirt then the internet
41:39 crapped out then yeah I don't know just
41:42 you know so my apologies Troy this is
41:44 this really is your channel and uh I
41:47 should I should be much more uh
41:49 respectful in the future and I will I'll
41:51 do my best Cheers
41:54 have you messed around with asking a
41:56 question having AI write it four
41:58 different ways and getting four
41:59 different results each yes yes
42:04 um
42:05 having AI write it four different ways
42:07 and getting four different results each
42:08 yes absolutely finding a like content
42:11 and delivering the best ranked out of
42:13 those 16.
42:14 um that's a that's actually a really
42:15 good approach we did a thing last night
42:17 on the live judge on the live
42:20 judged
42:22 um judged as his his or her name judge
42:24 gear
42:25 um we did a thing last night where I
42:27 wrote a prompt uh because someone said
42:30 Can can Chachi BT write a joke and I
42:33 said no like like it understands what
42:36 jokes are but it ain't funny and so so
42:38 we went and and we took the exact same
42:41 prompt and did it in all these different
42:43 models and Chachi PT was the only one
42:46 that wrote an actual joke that you could
42:48 actually kind of have edited and turned
42:50 into a joke the worst one was Bing uh
42:54 and then the rest were just bad and and
42:55 then some were really disturbing like
42:58 the one joke was
43:00 um
43:00 because because the premise of the joke
43:02 was write a joke about how avocados go
43:05 from being unripe to rotten you know in
43:08 this really short window or whatever it
43:10 was
43:11 and and like I I which one was it was it
43:15 Claude
43:17 I don't know I think it was cloth no it
43:19 was clutter Sage I think it was Claude
43:21 And it said um
43:24 avocados are are just like children they
43:27 start out rock hard and I was like no no
43:30 it's like it's like it sort of had the
43:33 right setup and the right punch line but
43:35 the middle of it was just wrong and
43:37 disturbing and like weird and like it
43:39 was just bizarre
43:41 um but Chachi BT actually wrote a pretty
43:43 good joke but yeah that's that's
43:44 actually a really good idea and and the
43:46 other thing you can do I assume you've
43:48 done it judged you can have it write
43:50 those four results in a table that you
43:53 can then just copy and paste into a
43:54 Google sheet
43:56 and it'll it's it's already you know
43:58 formatted for the table
44:00 have you figured out how to make money
44:02 with AI yet did AI playing with AI for
44:06 months you can make money
44:09 um yeah I figured out how to make money
44:11 with AI I've got it
44:13 um rolled into my my company right now
44:15 we're out there pitching it
44:18 um I've been doing some paid speaking
44:20 gigs I've been doing some paid workshops
44:24 um
44:26 right now I'm not
44:29 I'm not looking to start a business
44:30 around AI because too much is changing
44:33 too fast and I have my company to run
44:36 um but I think it's critically important
44:39 at least for me
44:40 to be as educated as I can and keep and
44:43 just keep building building building so
44:45 I'm building content generation engines
44:48 um I'm helping people that are like hey
44:51 I'm starting something can you give me a
44:52 hand here I'm doing that kind of stuff
44:55 with it the the trick with so so here
44:57 here's my my
45:00 my little soap box for you can make
45:03 money with Chan gbt
45:06 making money with an AI tool is no
45:10 different than making money with
45:11 Photoshop and it's no month different
45:13 than making money with a shovel they're
45:16 all tools right even these are tools now
45:19 these happen to be the most powerful
45:21 tools in the history of humanity so
45:24 there's there's some things you can do
45:25 with these tools that are kind of a
45:27 Quantum Leap above right 10 20 30 x the
45:32 results that you would get if you did it
45:33 without these tools
45:35 but you still need to solve a problem
45:37 and that's the business how you make
45:39 money with tools is you say hey
45:42 I know you how to use Photoshop so good
45:45 that if you're trying to make a poster
45:48 for your music festival I'm your guy
45:50 come see me I'll make you music posters
45:53 and then that's not just making money
45:56 with Photoshop that's making money with
45:58 my knowledge and my life experience and
46:00 my design skills and this and that and I
46:01 happen to be using Photoshop as the tool
46:03 same thing with these so what I would
46:06 lean into if I were you is learn the
46:09 [ __ ] out of these learn what they make
46:10 possible start playing with mid-journey
46:12 and stable diffusion and Adobe Firefly
46:15 and generating images start learning
46:17 what's possible on the with the music
46:19 tools and with the animation tools and
46:20 with the film tools just everything you
46:23 can get your hands on get
46:26 you know I don't know curiously ravenous
46:33 and then start talking about it
46:36 post it on your LinkedIn
46:38 tell your friends about it at dinner
46:40 parties even if they think you're a
46:41 [ __ ] weirdo mine do
46:44 you don't think they think I'm a [ __ ]
46:46 weirdo look at the hair look at that
46:48 that's not normal
46:52 I've got no credentials
46:56 zero qualifications
46:59 um but start talking to people about it
47:01 because we are entering a phase right
47:03 now especially now that now that there's
47:06 like really viable competitors to chat
47:09 gbt I think we're gonna quickly see an
47:13 acceleration where these tools start to
47:15 proliferate into all of the tools that
47:18 we're already using
47:19 and so if you can be one of the people
47:23 that when someone goes I need to figure
47:25 this [ __ ] out who should I talk to about
47:27 oh you know who I should talk to
47:29 let's go let's go Brandon
47:32 let's go Brandon
47:35 cute
47:36 um
47:38 be the guy they come to
47:42 because
47:44 as there's a guy named Paul Rachel from
47:46 the marketing AI Institute and he says
47:49 AI is not going to take your job but
47:51 someone using AI will
47:54 so
47:56 be that person
47:58 the video was used by dopeconsulting.ai
48:01 wait the video which video Miss Birch
48:04 what are you talking about wait oh
48:05 you're talking amongst yourselves Talk
48:07 Amongst yourselves
48:09 I love when you talk amongst yourselves
48:11 you entertain yourselves because I don't
48:12 get to the questions soon enough do I no
48:15 I don't you don't you never got to my
48:17 foot I don't even think he's live do you
48:18 think he's alive I don't think he's live
48:20 I think this is a recording you know
48:22 what I think it is I bet he's chat GPT
48:24 five
48:26 he could be six he could be a spy
48:29 Betty's Chinese we're on Tick Tock I bet
48:31 he's a Chinese AI Avatar spy
48:35 and we're on tick tock
48:37 I don't know that we should be on Tick
48:39 Tock Marge I don't
48:44 oh by the way I'm an idiot
48:47 my workflow is very much like yours not
48:50 as frustrating as it was pre-planning uh
48:52 so much easier yeah exactly
48:54 exactly
48:57 it's just it I mean it's just something
49:00 it's just something this stuff is
49:03 something thank you for changing your
49:05 shirt you're welcome good evening
49:06 Emilio's wife what's happening good to
49:08 see you
49:10 thank you for the heads up on Elon I
49:13 could pop over you so let me pop over
49:16 and and show you all the tabs I opened
49:19 up just for all the news and I and I've
49:21 missed like there's news I don't have in
49:23 here
49:26 all right let's see let's see if they
49:28 brought back browse with Bing I don't
49:30 think they have
49:32 where's my mousy mouse
49:36 beta features no browse with Bing's not
49:39 back okay so there's chat GPT this was
49:41 King of the mountain
49:42 and then all of a sudden here's bard
49:45 looks a lot like Poe they it looks like
49:49 they did a little even interface Steven
49:52 from Poe
49:53 but let's see uh create a table
49:57 comparing Elon
50:00 musk and Mark
50:04 Zuckerberg
50:07 um
50:09 from a
50:12 business
50:13 and martial arts perspective
50:17 [Laughter]
50:21 let's see what it does one thing I've
50:23 noticed about bart that's interesting is
50:24 it doesn't it doesn't show you like like
50:27 the other one sort of show you that
50:28 they're writing and Bard just sort of
50:30 thinks it's got like this little twirly
50:32 twinkle thing and then it just
50:35 delivers your thing so it says sure here
50:38 martial arts experience Elon Musk
50:40 trained in Judo karate and Taekwondo as
50:44 a child Mark Zuckerberg no formal
50:46 training that's not true actually he's
50:48 been to a lot of doing a lot of
50:49 Taekwondo or a lot of Jiu Jitsu lately
50:52 business experience
50:55 net worth 215 billion 67 billion 51
50:59 they're both one's 51 one's 52 same age
51:02 notable achievements style strengths
51:06 so that's barred Bard's good now you can
51:10 upload files to Bard you can upload
51:12 images to Bard and have it analyze them
51:14 which is really interesting you can't
51:16 upload csvs to Bard but I assume that'll
51:20 come so that's barred
51:22 then you've got your Bing you got your
51:24 Bing which is gpt4 if you go to Bing and
51:26 then click on chat you've got more
51:29 creative more balanced more precise you
51:31 can analyze images here you can
51:35 um it's connected to the internet you
51:37 can say to Bard
51:40 um make me an image of
51:44 musk and Zuck
51:49 as cartoon
51:51 characters in a
51:55 cage match in Vegas
52:01 and it should pop up a thing to say I'll
52:03 try to make that for you I can help you
52:05 with that here's your generative image
52:08 oops try another prompt looks like some
52:11 words might be blocked jerks
52:14 oh fight okay wait oh cage match what
52:18 what's blockable in that
52:21 all right we'll just do
52:23 Zuck and musk as cartoon characters
52:29 and it looks like these guys have
52:30 changed their interface for the image
52:34 no I guess you can't they don't want you
52:37 doing those people fine oh they're
52:39 trying to prevent deep fakes yeah we're
52:40 coming into election uh uh make me an
52:45 image of a 70s
52:51 muscle car
52:53 as a cartoon
52:58 so this interface has changed so so it's
53:00 like these these things are evolving
53:02 daily oh no there it goes that's that's
53:04 what we saw before
53:06 that's the same interface never mind
53:12 so those are pretty cool all right so
53:14 that's that's Bing so then you got
53:16 Claude and Claude is very much like gpt4
53:20 you can upload up to five files of 10
53:23 megabytes each you can say write me a
53:28 poem
53:29 about uh let's see
53:32 explaining
53:34 quantum mechanics
53:39 as
53:40 an m m
53:43 rap
53:45 and the reason I do this one is
53:48 to see if it can do uh rhyming rhyme
53:52 schemes listen up to the story I'm about
53:54 to tell about quantum physics and how
53:56 elect electrons dwell in superposition
53:59 which means they can be in so many
54:02 places simultaneously G
54:05 that ain't bad
54:07 Einstein called it spooky action at a
54:09 distance uncertainty the heisen
54:11 Heisenberg name the more you know a
54:15 particle speed okay that was bad
54:18 completely missed it there so you know
54:21 again not as good as gpt4
54:24 but really powerful then you have Pi Pi
54:29 dot a i and this one is the one from a
54:33 company called inflection that just
54:34 raised 1.3 billion dollars series B 1.3
54:39 billion dollars to buy 20 000 gpus and
54:44 create a massive super computer cluster
54:48 to train all this [ __ ] and this one is
54:52 let's see
54:53 um explain
54:56 um
54:58 quantum
54:59 physics uh
55:02 uh as if I am seven
55:08 okay I'll try to keep it simple Quantic
55:10 quantum physics is about how very very
55:12 tiny things behave things that are
55:14 smaller than anything in your body so
55:16 it's doing its thing
55:18 um this thing tends to be really really
55:21 friendly they've got it programmed to be
55:23 friendly so that's that one
55:25 then you've got po
55:27 and the reason I'm going through these
55:29 is like all of these have changed in the
55:31 past they've either appeared out of
55:33 nowhere or or they've changed and
55:35 updated so Poe if you remember if you've
55:38 been here before there's they've got
55:39 Sage they've got a Claude model gpt4
55:42 Claude instant and chat gbt well they've
55:45 now got Claude 2 and Claude instant so
55:48 Claude 2 is like gpt4 but it's a hundred
55:51 thousand token context window so 50 75
55:54 000 words then you got GPT for
55:57 you've got all the bots so pose really
55:59 interesting the other thing I noticed
56:01 about Poe the other night if you go to
56:03 explore
56:04 you have all these Bots like like one of
56:07 the Bots are AI Bots and the AI Bots are
56:10 things like mid Journey Dolly prompt
56:12 genius
56:14 um so things that help you with
56:15 prompting but what I noticed is that
56:18 there's there's official pobots
56:21 so you've got all the ones that are
56:23 listed here those top six but you've
56:25 also got chat GPT 16k
56:29 chat gpt4 32k so on Poe you can play
56:34 with chat gpt4 with the 32k context
56:37 window I don't think that's available
56:39 anywhere else you can also play with
56:41 Google's palm palm model here
56:44 so Poe's really interesting for that
56:46 reason
56:48 what's this oh yeah this is the this is
56:51 the story about open AI cutting a deal
56:54 with the Associated Press which is a
56:57 huge deal that's zapier zapier is a big
57:00 deal this is
57:03 Elon musk's new startup right so they've
57:05 got chat GPT as the as the the house and
57:09 x dot AI as the as the tornado that's
57:12 coming so I think something's going to
57:15 be announced on that tomorrow it says
57:16 breaking Elon musk's new startup just
57:19 launched their official Twitter account
57:23 the slogan is to understand the true
57:26 nature of the universe so
57:29 um so that's coming
57:31 what else do I have here
57:36 oh
57:37 gpt4s
57:39 infrastructure was leaked and it's
57:43 insane
57:45 it's not 175 billion parameters it's 10
57:48 times that it's 1.7 trillion or
57:50 something like that
57:52 um
57:54 there's there's
57:56 I think a dozen or 16 different expert
58:00 systems that when you when you do
58:04 whatever you do when you ask gpt4 it's
58:07 sort of calling on these different
58:09 expert systems
58:11 so so anyway so this this leaked
58:15 so that's out there now
58:17 oh then meta
58:21 meta Facebook said they so so all of the
58:26 open source models right now are based
58:29 on an accidental leak from meta they're
58:33 they're llama models which they said for
58:35 for research purposes only so they're
58:37 not commercially viable so what
58:39 Zuckerberg has now said is he thinks
58:42 he's going to release their competitor
58:44 to chat GPT out into the wild with all
58:47 the weights and Commercial commercially
58:50 licensable and not charged for it
58:53 so
58:54 holy [ __ ]
58:56 and then I think that's it is that it
58:59 yeah that's it I mean good
59:02 God
59:05 I I I I I like
59:10 I pay attention to this stuff a lot and
59:13 I I mean I had to I had to like I
59:15 literally was taking notes today just
59:17 trying to keep up with it I'm like I
59:19 gotta talk about that tonight I gotta
59:20 talk about that tonight Emilio's wife
59:22 sent me a DM hey did you hear about Elon
59:24 I'm like no I haven't heard about Elon
59:33 quite amped up enough
59:45 when
59:47 you upload a PDF or documents to chat
59:50 gbt
59:53 to AI to manipulate can anybody else get
59:56 them
1:00:00 chat GPT I'm pretty sure anthropic says
1:00:04 it I'm pretty sure Bard says it
1:00:06 they all say we may train on your data
1:00:09 open AI has give you given you the
1:00:12 option to opt out of that but I actually
1:00:15 have no idea
1:00:16 if you upload a PDF if it
1:00:20 is being stored in anything other than a
1:00:23 cache you should assume
1:00:26 the worst right I wouldn't put client
1:00:29 data there
1:00:31 um if you really want to start training
1:00:33 and fine-tuning on your own data then
1:00:35 start learning Lang chain
1:00:38 so you can create like I did for my
1:00:41 company storyline we just started a lang
1:00:43 chain instance so that we can start
1:00:46 training on our own data and we can
1:00:47 start playing with customer data in a in
1:00:50 a way that is safe
1:00:55 um
1:00:57 I can't imagine that's going to be for
1:00:59 long also the nature of how
1:01:02 how I understand code interpreter to
1:01:05 work is that it's creating
1:01:08 kind of a standalone virtual machine
1:01:11 that it's execute at writing and
1:01:13 executing all that python code within
1:01:15 just for your chat instance and then it
1:01:17 wipes it because it wipes it like 10
1:01:19 minutes after you use that it it wipes
1:01:21 it and you lose all your work so
1:01:24 I'm guessing that they're not
1:01:26 storing that but I I don't I don't know
1:01:30 I I don't know like I would assume the
1:01:32 worst
1:01:33 I mean that that's the the problem with
1:01:35 all of these tools right now is is
1:01:37 they're so new there's like they're all
1:01:39 like shiny new and like the engineers
1:01:42 are like yeah we're good to launch let's
1:01:43 get that thing out there
1:01:45 you sure we checked everything I put an
1:01:47 extra roll of duct tape on it we're good
1:01:48 to go oh here let me polish that with my
1:01:50 elbow yep that's that's uh that's launch
1:01:53 ready technology so so it's all like
1:01:57 that right now
1:01:58 this is
1:02:00 a little concerning you know it should
1:02:02 be good it should be fine yeah it's
1:02:04 probably not gonna
1:02:05 torpedo your your systems or your
1:02:08 company Trade Secrets probably not
1:02:10 probably probably it's probably gonna do
1:02:12 that you're probably gonna we're we're
1:02:14 all gonna die oh God oh God
1:02:19 all right can I search what a competitor
1:02:22 could have prompted
1:02:25 like if they if they
1:02:28 there's a way to do it I don't know how
1:02:30 to do it but someone did it because
1:02:32 someone from Samsung put their trade
1:02:34 secrets in the chat and they left their
1:02:37 chat history there and someone hacked
1:02:40 them through their chat history so I
1:02:42 yeah like I I don't think they're
1:02:44 encrypting what's being stored or or
1:02:48 what's being sent and I assume it was
1:02:51 done through some sort of packet
1:02:53 sniffing situation but I don't [ __ ]
1:02:55 know I don't I ain't no hacker I ain't
1:02:57 no hacker I just I just pretend
1:03:00 everything's gonna be okay I just live
1:03:02 in Oblivion
1:03:03 you look like the guy from Modern Family
1:03:05 I've gotten that thank you I appreciate
1:03:06 that I've gotten him I've gotten uh
1:03:11 Tom Arnold
1:03:13 I've gotten skinny John Goodman so so I
1:03:16 got got a lot of Roseanne connections
1:03:18 apparently I've gotten Rosie O'Donnell
1:03:20 you look like Rosie O'Donnell which is
1:03:22 again that's a solid troll because it's
1:03:25 is it a compliment hmm okay
1:03:28 sorta yeah hmm yeah uh occasionally if
1:03:33 I'm a if I'm a bit angry and and ragey I
1:03:36 get Nicholson so yeah so anyway thank
1:03:38 you appreciate that
1:03:41 I want to know how Pharmaceuticals are
1:03:43 using it for research well so so listen
1:03:45 here's the thing about all this machine
1:03:46 learning and AI [ __ ] it's been around
1:03:48 for decades so so pharmaceutical
1:03:50 companies are using
1:03:53 they've been using AI for for research
1:03:56 and all this stuff for for just years
1:03:58 and years and years
1:04:00 the thing that's different is the
1:04:01 Transformer model this thing that
1:04:04 um
1:04:05 Google invented in 2017 just completely
1:04:08 changed the nature of how these tools
1:04:10 work and the scale of data that you can
1:04:12 train them on so when deepmind created
1:04:15 Alpha fold
1:04:16 and they generated 400 000 you know
1:04:19 protein models in what was it like three
1:04:23 days or a week or some ridiculously
1:04:25 short amount of time and I think it I
1:04:27 think historically it would take you
1:04:29 like two weeks to do one and they did
1:04:31 like 400 000 of them and they you know
1:04:33 they were largely correct
1:04:36 um that's the new stuff and so so I
1:04:39 would imagine that that um
1:04:42 they're using them in all sorts of ways
1:04:44 what what
1:04:45 where we're now starting to hear our
1:04:48 clients the pharmaceutical companies
1:04:50 talk about this stuff is in everyday
1:04:52 operations in marketing in particular
1:04:55 they spend a tremendous amount of money
1:04:58 on marketing with these big agencies
1:05:01 um so so they're trying to figure out
1:05:04 how can we do this they're trying to
1:05:05 figure out
1:05:07 um well rather than having to do a photo
1:05:09 shoot could I just do personas and then
1:05:11 create images that look like people
1:05:13 right and this is this is where things
1:05:16 like
1:05:17 um you know existential threats to
1:05:19 things like you know photography start
1:05:22 to show up where if an internal group at
1:05:26 a Pharma company can just generate
1:05:27 Persona images that look like
1:05:29 photographs they don't need to hire that
1:05:31 photographer to go do that photo shoot
1:05:33 to have you know those pictures of
1:05:37 fake patients you know who who are fake
1:05:39 patients anyway like the Pharma company
1:05:41 doesn't really care they just want
1:05:43 something that's believable for that
1:05:45 particular Persona so they can talk
1:05:47 about a concept or or a particular piece
1:05:49 of research
1:05:51 um
1:05:52 so yeah but I you know I I would assume
1:05:56 in their medical divisions they're using
1:05:58 it in all sorts of profound ways I think
1:06:00 we're going to see a lot
1:06:04 medical Innovations in the next five
1:06:06 years
1:06:08 I'm writing a story right now about AI
1:06:10 Tech on the farm
1:06:12 oh Chris that's hilarious that's solid
1:06:16 that's that's that's really a dad joke
1:06:19 you know not to be sexist but shouldn't
1:06:22 I be saying [ __ ] like that I think I
1:06:24 should
1:06:28 just got hello
1:06:31 just got on oh hello
1:06:33 O'Donnell
1:06:36 Juan O'Donnell is that me oh like as in
1:06:39 Rosie I get it
1:06:45 does anyone else hear a slight Echo I
1:06:48 think you you you have weird computer
1:06:50 speakers and microphones they're
1:06:52 Emilio's wife
1:06:54 I think I think it's time for an upgrade
1:06:56 I think you gotta talk to Emilio Emilio
1:06:59 and say listen I need a more
1:07:01 professional setup
1:07:03 we're doing lives we do them regularly
1:07:05 we're on the tickety talk
1:07:07 and uh we gotta upgrade the speaker
1:07:09 situation but anyway if I am echoing let
1:07:11 me know
1:07:13 all clear here see
1:07:14 his AC's going on in the background yeah
1:07:17 it's hot here I can turn it off
1:07:19 hang on
1:07:27 see these are the kind of sacrifices I
1:07:30 make for you
1:07:31 you think this is easy no
1:07:35 we're all in on AI at my gig for a year
1:07:39 now oh you're in an agency retro Punk
1:07:40 aren't you
1:07:41 they get it however it's Pharma so also
1:07:45 slow oh yeah there's there they're also
1:07:47 very slow yeah listen I
1:07:50 Pharma is not going to jump all over
1:07:51 this but I think what Pharma is going to
1:07:53 do is they're gonna get really smart
1:07:56 about it
1:07:57 um and understand it and then they're
1:07:59 gonna start demanding that their vendors
1:08:02 have a plan
1:08:04 and
1:08:06 the I.T groups like I talked to I talked
1:08:09 to a senior guy who found me on LinkedIn
1:08:12 from a big Pharma company today
1:08:15 and he's been actively like generating
1:08:18 personas generating uh marketing copy
1:08:21 generating all sorts of stuff he he's
1:08:23 within what used to be their I.T group
1:08:25 and it's now a big Innovation Center of
1:08:29 Excellence
1:08:30 and um
1:08:33 like
1:08:35 he was pretty flipping educated like he
1:08:38 was he was up there so and and he was
1:08:41 also smart enough to know we're not
1:08:42 going to use any of this [ __ ] until
1:08:44 we we know what's going on from a
1:08:46 security standpoint
1:08:48 um
1:08:49 but once they have that figured out I
1:08:51 think they're going to be moving pretty
1:08:52 fast and I think they're like listen
1:08:53 these companies spend a lot of money
1:08:58 and if they could spend a hundred Milli
1:09:00 100 million less
1:09:02 on Advertising fees and get the same
1:09:04 amount of work or better or more
1:09:08 they're going to push for that the
1:09:10 agencies aren't going to have a say in
1:09:11 that
1:09:12 so anyway but I mean listen I can
1:09:16 congrats to you for working for a place
1:09:18 that gets it
1:09:20 because
1:09:22 sounds perfect now good what about AI oh
1:09:24 xai yeah xai is coming I I it sounds
1:09:28 like there's going to be something
1:09:29 tomorrow right they launched their
1:09:31 Twitter Channel today they made some
1:09:33 announcements like I saw a couple of
1:09:35 news stories come through after you
1:09:36 mentioned it and then
1:09:38 um it sounds like they're gonna do
1:09:40 something tomorrow so I don't know
1:09:43 I just noticed Bard made the list in the
1:09:45 background I know Vicky right look at my
1:09:48 list it's [ __ ] this has been three of
1:09:51 these poor little sites for
1:09:52 month and a half now now there's five
1:09:55 and I think it's gonna go to I think I'm
1:09:58 gonna have to find a new way to display
1:10:00 it because because it's getting stupid
1:10:03 but I think we're gonna have to put Pi
1:10:05 dot AI on here if it gets like one level
1:10:07 better like right now it's a little
1:10:09 simple tinny
1:10:11 um and then I think xai is gonna show up
1:10:14 here and then if meta if meta launches
1:10:18 something that's as good as GPT 3.5 like
1:10:20 they're claiming it's better than GPT
1:10:22 3.5
1:10:23 and if they release it commercially
1:10:26 available as open source I think that
1:10:29 makes the list so that would be seven or
1:10:32 eight
1:10:33 I don't know what to do
1:10:36 but you know I think they'll Whittle
1:10:38 their way down but it's just it's crazy
1:10:40 it's crazy crazy crazy crazy crazy crazy
1:10:42 crazy crazy crazy crazy
1:10:44 endlessly curious that's me that is the
1:10:47 best so the the I I think the the super
1:10:50 power for the next year to 18 months is
1:10:54 curiosity
1:10:55 and a sense of adventure just think of
1:10:59 all of these tools everything you're
1:11:01 learning everything you're building just
1:11:04 think of it as a curiosity exercise and
1:11:06 an exploration and an adventure because
1:11:09 almost anything that any of us build in
1:11:12 the next year is going to be rendered
1:11:14 obsolete by some update of something
1:11:19 it doesn't mean you shouldn't build it
1:11:21 it just it just means you shouldn't be
1:11:24 too attached to it because in building
1:11:26 it you're going to learn a ton right you
1:11:29 should be playing with automations with
1:11:32 tools like zapier or make or bubble if
1:11:36 you're a little more technically
1:11:37 oriented which is sort of low code
1:11:39 rather than no code
1:11:41 um
1:11:42 and and just start understanding what
1:11:45 these tools make possible and how they
1:11:47 change what you what you know about how
1:11:49 do they transform what you already do or
1:11:52 what you already know
1:12:01 yeah by the way give me a follow if you
1:12:03 want to I've got a bunch of content on
1:12:05 my tikitaki and if you follow me I think
1:12:07 you get notified when I go live
1:12:10 um do you do I have to pay mid-journey
1:12:12 to generate a picture you you can
1:12:14 generate I think like 20 before you have
1:12:17 to pay
1:12:19 um Mid Journey's the one image
1:12:20 generation tool I pay for
1:12:23 um although
1:12:24 uh you know we we pay for the Adobe
1:12:27 suite you know my business pays for that
1:12:29 for me so so I've also got the uh Adobe
1:12:31 Firefly and generative fill in Photoshop
1:12:34 beta
1:12:35 um but but mid journey is the one that I
1:12:38 feel is worth paying for if you really
1:12:41 don't want to pay for it and you're
1:12:43 decently geeky Beer Hunter
1:12:46 um then start playing around with stable
1:12:47 diffusion go go to the subreddits for
1:12:50 stable diffusion and just start learning
1:12:51 about Google collab and you know setting
1:12:54 up automatic 11-11 and installing the
1:12:57 extensions because
1:12:59 um stable diffusion is by far the most
1:13:02 flexible and Powerful
1:13:05 um
1:13:06 it because it's open source and the
1:13:08 developer Community is just pouncing on
1:13:10 it
1:13:11 um but mid-journey for me is the best
1:13:12 one it's just the quality is just
1:13:14 consistently
1:13:16 jaw-dropping consistently drop dropping
1:13:20 [ __ ] insane
1:13:23 let me see
1:13:24 I'll show you what I did oh that's one
1:13:26 this is that one that's a beautiful
1:13:28 image
1:13:31 this is this is one that I did that's a
1:13:33 mid-journey image
1:13:35 and it's just stunning look at those
1:13:38 things
1:13:40 that one I post this one I posted to
1:13:42 LinkedIn the one with the VW bus
1:13:46 I mean
1:13:48 look at that
1:13:53 that was a cool one that was one where I
1:13:55 had one of the things you can have
1:13:56 mid-journey do is you can have it
1:13:58 describe another image and then you take
1:14:01 its description and use that as a prompt
1:14:04 and so I took a description of
1:14:07 um cruise liners and these are these are
1:14:10 some of the images that came out of the
1:14:12 cruise liner one but then for just some
1:14:14 random [ __ ] reason it made a VW bus
1:14:17 so so I got I got you know these cruise
1:14:21 liners which are stunning like don't get
1:14:22 me wrong those are beautiful but that
1:14:24 one is just magical
1:14:26 it's just like you know like I want to
1:14:28 be surfing there that's like California
1:14:29 Dreaming baby you know
1:14:32 so like that that modern art image this
1:14:34 is mid-journey
1:14:37 these are some ones from from the AI
1:14:40 Salon that I run
1:14:41 these are all mid-journey that were made
1:14:43 in the server here
1:14:46 so yeah
1:14:48 cool stuff
1:14:53 [Music]
1:15:01 chat go okay
1:15:05 go
1:15:06 I don't know what you mean what does
1:15:08 check go mean
1:15:10 is that is that another chat GPT
1:15:12 competitor that launched today I can't
1:15:14 take it anymore I don't have room left
1:15:15 on my billboard
1:15:17 please tell me there's not a chat Go
1:15:19 please
1:15:21 I wish I could organize my life it
1:15:24 probably can I in fact I know it can you
1:15:27 can
1:15:31 if you if you've if you're paying for
1:15:33 chat GPT Plus
1:15:35 you could upload a bunch of
1:15:37 disorganized data and tell it to parse
1:15:41 it and organize it into a spreadsheet
1:15:43 that you can download
1:15:45 tell it how to order it tell it to tell
1:15:48 you what it found
1:15:50 suggest ways to categorize it and
1:15:53 organize it and then tell it to put it
1:15:54 in a CSV file so you can download it and
1:15:57 it'll do that
1:15:59 so now if you're like me if you want AI
1:16:04 to help you organize the crap around
1:16:05 your desk then if you're anything like
1:16:08 me we we need we need the robot we need
1:16:10 elon's robot to do that [ __ ]
1:16:14 the digital stuff it can do the physical
1:16:17 stuff huh
1:16:22 chat EBT couldn't come up with a do like
1:16:26 that no exactly that's that's that is
1:16:31 like a [ __ ] raccoon lost a fight
1:16:34 that's bad that's bad love Kai Kyle in
1:16:39 character half the reason I watch every
1:16:40 night keep up the good work Kyle keep up
1:16:43 the overacting Kyle you've chewed the
1:16:45 scenery beautifully tonight
1:16:48 you can take a man off to stage
1:16:51 but you can't help him from overacting
1:16:55 isn't that how the saying goes pretty
1:16:57 sure that's it
1:16:58 oh
1:17:01 all right meta Victory what do you have
1:17:03 to say you added Bard to the board I
1:17:05 know I've been bad-mouthing Bard for a
1:17:07 [ __ ] month longer than that I've been
1:17:09 bad mouthing Bard ever since that French
1:17:12 did you see the um
1:17:14 that when when they did the first one
1:17:16 open AI did their thing and like you
1:17:18 know all of a sudden like everyone was
1:17:19 like holy [ __ ] chat apt is the thing and
1:17:22 and Google was like we're having an
1:17:24 event in Paris and uh we're going to
1:17:27 announce all of our AI that we are we
1:17:29 have magically done it all and then this
1:17:32 one woman gets up to demo something and
1:17:34 she's like okay here I am going to demo
1:17:36 this with the demo phone it that
1:17:39 wait where's where's the demo for what
1:17:41 what oh we don't we lost it oh okay I
1:17:45 guess we're not gonna demo it like it
1:17:46 was [ __ ] embarrassing
1:17:48 and then they did they did another one
1:17:50 like two months later and it was fine
1:17:51 but they didn't really announce anything
1:17:53 like it was just it was just like a
1:17:55 whole event of we do AI too uh here's a
1:17:58 tool with AI this has always had AI but
1:18:00 we're adding more
1:18:02 and then you know they launched Bard and
1:18:03 it sucked and then they relaunched Bart
1:18:06 after that and it sucked and so I've
1:18:08 just been [ __ ] bad mouthing it for
1:18:10 months and months and months and people
1:18:12 would come in life and buy it on the
1:18:14 list I'm like because it sucks
1:18:17 but I also said to my credit don't count
1:18:21 Google out because there's some smart
1:18:23 [ __ ] with a lot of money and
1:18:24 they invented this stuff and some of the
1:18:27 tech that they've showed like style drop
1:18:29 is really impressive so
1:18:32 today they dropped it and it's good so
1:18:35 there you go it's this is again we're
1:18:38 really early but but this is here's the
1:18:41 deal for for all of us just regular Folk
1:18:45 all us chillings out here just waiting
1:18:47 for these guys to come up with more
1:18:48 remarkable tools to play with
1:18:51 um this is really good news
1:18:53 when you have five major competitors and
1:18:55 it's likely that there's gonna get three
1:18:57 more added to that list and there's
1:18:59 probably
1:19:01 three to five more we don't even know
1:19:03 about that are in stealth right now
1:19:08 this is going to be some insane [ __ ]
1:19:10 if you think this is just hype and this
1:19:13 is blowing over and you've got three
1:19:15 years to figure it out
1:19:18 nope
1:19:19 what's your thought on Google bard I
1:19:21 just gave it uh I thought it was a giant
1:19:24 steaming pile of poo and now it is a
1:19:27 legitimate competitor to all the all the
1:19:28 others
1:19:32 uh
1:19:34 never mind you're using it already
1:19:38 Zuckerberg is 39. well these things
1:19:41 hallucinate RM all right I'm Walter yes
1:19:44 by the way if if you're new to this
1:19:46 stuff and you're playing with this and
1:19:48 you're like hey wait a minute
1:19:49 Zuckerberg's 39 yeah I didn't I I
1:19:50 thought that seemed old but I'm like
1:19:52 okay whatever yeah he is 39 that's right
1:19:55 um so these things do what's called
1:19:58 hallucinating and hallucinating is
1:20:00 basically just it'll outright lie I like
1:20:02 to describe it as mansplaining as a
1:20:04 service I didn't invent that but it's
1:20:06 it's my favorite uh it's my favorite
1:20:08 name for for these uh for these large
1:20:11 language models
1:20:12 um the way they do what they do is quite
1:20:16 remarkable
1:20:19 so when you write a prompt it basically
1:20:21 turns it into a mathematical weighting
1:20:23 and says okay based on that prompt I'm
1:20:26 gonna like run the math
1:20:28 and based on all that math the next most
1:20:31 probable word as a response is the
1:20:34 and then it's going to take all of that
1:20:36 plus the word the and it's going to run
1:20:38 a new piece of math and say the next
1:20:40 most logical word after the is Pony and
1:20:44 it'll write Pony so when you see it
1:20:46 right three or four paragraphs for you
1:20:48 in you know two seconds it's doing it
1:20:51 word by word
1:20:53 what's remarkable about what it does is
1:20:55 it looks like it's sentient right it
1:20:58 looks like it understands what the [ __ ]
1:20:59 it's talking about but it literally has
1:21:02 no concept of what it just wrote like
1:21:04 all it knows it like it's it's literally
1:21:06 a calculator putting mathematical
1:21:09 probabilities that are tied to a string
1:21:11 of letters that are the word
1:21:13 all in a row
1:21:15 um
1:21:16 there will be new versions of this where
1:21:19 it's conscious of what it said but
1:21:21 they're not right now so what ends up
1:21:23 happening is if you just tell it to give
1:21:25 you a fact it'll give you a fact and
1:21:28 that fact may or may not be a fact so
1:21:31 it's perfect for conspiracy theories
1:21:33 [Laughter]
1:21:39 it's hilarious
1:21:41 he's very ripped these days thank you I
1:21:44 don't oh you meant Zuckerberg damn it
1:21:47 oh dang it
1:21:49 pie torture tensorflow I don't know that
1:21:52 shit's above my pay grade I don't know
1:21:54 enough about mathematics to use
1:21:56 tensorflow I tried to use it I tried to
1:22:00 use it five years ago I'm like I'm Gonna
1:22:02 Learn machine learning and then I was
1:22:03 like I don't have a degree from Stanford
1:22:05 I'm why
1:22:07 and then chat GPT came out and I'm like
1:22:09 I can use words for programming I could
1:22:12 do that [ __ ]
1:22:13 so I don't know
1:22:16 um probably tensorflow I'm just getting
1:22:18 on the tensorflow because it's Google
1:22:20 and I just I feel like if it's from a
1:22:22 technical perspective Google's got their
1:22:24 [ __ ] together
1:22:25 Bing is bad at physics trust me doesn't
1:22:27 surprise me at all
1:22:29 uh what you should play with RM is uh
1:22:33 play with a code interpreter chat gpt's
1:22:36 code interpreter and see if it's bad at
1:22:38 physics because I bet it's not because
1:22:40 what it's actually doing is it's writing
1:22:42 python code to fulfill whatever your
1:22:44 request is and then it's running that
1:22:46 python code within a virtual machine
1:22:48 checking its answer and if it gets it
1:22:51 wrong or it times out or it can't figure
1:22:53 it out it'll it'll come up with new
1:22:55 models to run so my guess is code
1:22:57 interpreter is going to be better at
1:22:59 physics than Bing
1:23:01 Bing is also the when we did the little
1:23:04 joke writing exercise yesterday Bing was
1:23:07 laughably bad it was as if someone at
1:23:09 Microsoft wrote a joke
1:23:13 it was bad
1:23:17 I know there's probably some Microsoft
1:23:19 levers in here but tough
1:23:21 I'm an apple Fanboy and I can bash on
1:23:24 Microsoft all I want I've been like a
1:23:26 [ __ ] ad for them for the past six
1:23:27 months
1:23:29 they they uh they give me Bing give the
1:23:32 official Bing site you know comments on
1:23:34 my videos sometimes I'm like oh yeah
1:23:36 hanging out with the Bros at Microsoft
1:23:38 me and Redmond were like this
1:23:41 can you explain that licensable but free
1:23:44 yes so okay
1:23:46 so
1:23:49 when meta
1:23:51 so so chat GPT did what it did got to
1:23:54 100 million users in six weeks and then
1:23:56 Google freaked out and said we've got AI
1:23:57 too and they made all these
1:23:59 announcements and then their Tech was
1:24:00 really shitty
1:24:01 Meredith jumped on the bandwagon and
1:24:04 said well
1:24:06 I I know
1:24:08 we uh we told you this metaverse thing
1:24:11 was going to be really big and in fact
1:24:13 we renamed our company around it anyway
1:24:16 um let's just put that to the side for a
1:24:17 second anyway we're great at this AI
1:24:19 stuff too I don't know if you knew that
1:24:21 but we've got uh engineers and
1:24:23 scientists and my goodness we there's so
1:24:26 much AI here you probably don't even
1:24:27 know so that was going on
1:24:29 and they took their llama they're
1:24:32 they're weaker llama models they're
1:24:33 seven billion billion and they're 13
1:24:35 billion parameter llama models and they
1:24:38 gave them to the developer community and
1:24:40 said here you can use these
1:24:43 um for for experimental purposes for
1:24:46 non-commercial purposes
1:24:48 and they gave them what's called the
1:24:50 weights so they basically gave them the
1:24:53 secret sauce so right now open Ai and
1:24:55 Google their models they they have not
1:24:58 released the actual core underlying
1:25:01 model that runs it so that you know the
1:25:03 Geeks can get in there and figure out
1:25:05 how they did what they did meta did
1:25:08 but they're not to be used for
1:25:11 commercial purposes so the open source
1:25:14 Community has been building all sorts of
1:25:16 models on top of them but you can't
1:25:19 really use them for commercial purposes
1:25:21 so what Zuckerberg said today is that
1:25:24 they've got a 30 or 35 billion model or
1:25:28 65 I don't know they've got they've got
1:25:30 some decent size I think it's 30 30
1:25:31 billion parameter model
1:25:34 that they claim is better performing
1:25:38 than GPT 3.5 and rather than license it
1:25:42 or launch their own chat gbt he says he
1:25:44 thinks they're just going to put it out
1:25:46 into the world with a free commercial
1:25:47 license so what it means is you've got
1:25:50 the core underlying model you can do
1:25:52 anything you want with it you can build
1:25:53 anything you want on it you don't have
1:25:56 to put any safety guard rails on it you
1:25:58 you can do anything and you can use it
1:26:00 commercially so it means me as an open
1:26:03 source developer could take that model
1:26:05 you know build a whole company like chat
1:26:09 GPT around it and I could say hey we're
1:26:12 going to do the
1:26:13 um if they're going to be safe with AI
1:26:16 I'm gonna I'm gonna give you a chat GPT
1:26:18 where you can do all the evil stuff in
1:26:20 the world and I'm going to build a whole
1:26:21 company around that you could do that
1:26:23 with that and you could monetize it and
1:26:25 it and that would not violate their
1:26:26 terms of service so that's that's the
1:26:29 the thing where the nature of the
1:26:31 license
1:26:32 um
1:26:33 is is the thing that's different where
1:26:35 for research only or for commercial
1:26:37 purposes and and that's and it's a huge
1:26:40 deal because
1:26:42 the open source Community is taking
1:26:45 these weaker models and using some
1:26:47 really creative techniques using gpt4
1:26:51 with like a hundred bucks in compute
1:26:54 costs they're they're fine-tuning these
1:26:56 things to be as good as GPT 3.5 so it's
1:26:59 just it's nuts it's just nuts like I
1:27:03 like the if if meta does that the speed
1:27:06 of innovation is going to get even more
1:27:08 insane
1:27:09 because now now people are going to be
1:27:12 throwing money at it
1:27:13 like right now you've got a bunch of
1:27:15 Open Source developers doing it for the
1:27:17 Love of the Game
1:27:19 but if meta says here's something that's
1:27:21 as good as GPT 3.5 for free that you can
1:27:24 use commercially the money is going to
1:27:26 come flooding into the market and The
1:27:28 Innovation is going to just
1:27:31 it's crazy it's crazy Crystal Archie
1:27:36 what was the name of the chat GPT
1:27:38 infrastructure article somehow I missed
1:27:40 that today
1:27:41 oh yeah
1:27:44 that is
1:27:52 it's on
1:27:55 semi-analysis.com and it's called gpt4
1:27:58 architecture infrastructure training
1:28:01 data set costs vision and MOA Moe
1:28:08 so just look up gpt4 architecture and
1:28:12 then I think
1:28:15 did I have it here
1:28:19 uh let me go to Twitter I'll go find it
1:28:22 on Twitter
1:28:24 let me see uh it's uh just just look up
1:28:27 go to Twitter and look up GPT for leak
1:28:33 yeah and there's there's like a bunch of
1:28:35 people there's like some tick tockers
1:28:37 doing things on it and
1:28:40 you know parroting what it is
1:28:43 but it was on semi
1:28:46 semi-analysis
1:28:49 gpt4 architecture infrastructure it's by
1:28:53 um Dylan Patel and Gerald Wong
1:28:57 I wish there was a way for me to pop
1:28:59 links into this chat but you know
1:29:03 it's the tick tock
1:29:05 we have international live video
1:29:07 distribution for free what do you want
1:29:10 links
1:29:13 someday
1:29:20 maybe I'll turn this into a a tick tock
1:29:23 store and I'll sell like fidget spinners
1:29:25 for 80d I'll just sit here play with
1:29:28 fidget spinners
1:29:30 I'll be like look I have I have let's
1:29:34 talk about AI oh by the way I'm writing
1:29:37 with this erasable pink magic marker
1:29:39 which by the way you can buy right here
1:29:43 anyone else sick of that [ __ ] already
1:29:45 [Laughter]
1:29:49 you've marked my words three months from
1:29:50 now I I was thinking about uh what's the
1:29:54 thing I was saying here about the
1:29:56 technology it's janky I was thinking AI
1:29:58 learning lab it's janky doing t-shirts
1:30:00 or something like that something about
1:30:02 that we could do a hair t-shirt
1:30:05 thanks for the news Roundup things are
1:30:07 getting spicy they really are it's it's
1:30:09 just nuts how soon will government start
1:30:11 trying to regulate this uh they're
1:30:13 already talking about it the FTC has
1:30:16 knocked on open ai's door and said hey
1:30:18 we need to talk what are you doing
1:30:20 what if you no no FTC has knocked on
1:30:23 their door and said what have you done
1:30:24 what have you done
1:30:27 um
1:30:28 so they're knocking on the door
1:30:30 um the the White House had Sam Altman
1:30:33 the CEO of Microsoft the CEO of Google
1:30:36 and the CEO of anthropic to the White
1:30:38 House about three months ago Sam Altman
1:30:41 has been actively talking about
1:30:43 um wanting uh regulation for companies
1:30:46 that have this kind of size and scale of
1:30:49 all of these guys
1:30:51 um
1:30:52 but not for smaller people which is
1:30:55 which is good for innovation
1:30:58 um
1:31:02 the
1:31:03 the regulation thing is really really
1:31:05 tricky because
1:31:08 um we need smart regulation we need
1:31:10 regulation we need smart regulation
1:31:12 we've got a Congress that doesn't know
1:31:14 what [ __ ] Wi-Fi is
1:31:16 and they don't legislate anymore it it
1:31:19 it's it's they're no longer a functional
1:31:22 government so I what I'm hoping is that
1:31:25 they put in some you know an AI Czar or
1:31:28 or you know
1:31:30 some sort of very smart person like
1:31:32 Jeffrey Hinton who actually quit Google
1:31:35 so that he could talk about the risks of
1:31:37 AI he would be an interesting person say
1:31:39 to put you know in in a powerful uh
1:31:42 position in government to drive smart
1:31:45 regulation because you don't want to
1:31:47 over regulate it
1:31:48 or you're going to kill Innovation for
1:31:51 the United States and this shit's moving
1:31:53 so fast that if you stifle Innovation
1:31:55 for six months it's like the the
1:31:57 racehorses are you know the F1 cars are
1:31:59 gonna be like
1:32:00 right past us and we'll be like uh hello
1:32:04 is anybody here
1:32:06 So It's Tricky I don't know I don't know
1:32:09 enough about that whole world to know
1:32:11 what to do but yeah it's it's terrifying
1:32:15 right now because
1:32:17 we have a non-functional clueless
1:32:19 government that we need to be acting not
1:32:22 clueless and not horrible
1:32:25 and I can't seem to they can't seem to
1:32:27 figure that particular piece out I've
1:32:29 been working with chat gbt on data
1:32:31 obfuscation tools so my idiot co-workers
1:32:34 don't leak info so tired Runner so so
1:32:37 you're talking about a thing called uh
1:32:40 uh Rachel woods from the AI exchange
1:32:43 talks talks about this a lot called ai
1:32:45 ai operations
1:32:47 um
1:32:48 if what you're doing that's going to be
1:32:51 such an in-demand role you're you're
1:32:55 going to be able you'll be working for a
1:32:57 long long time because that
1:32:59 making AI safe for companies data is
1:33:03 probably the single most important thing
1:33:05 in the next three to five years probably
1:33:08 longer
1:33:10 because because one of the things these
1:33:12 tools make possible is
1:33:15 you can take dormant data like you know
1:33:18 we were joking earlier about being data
1:33:20 Hoarders
1:33:21 you can take dormant data and embed it
1:33:24 and put it into you know Vector
1:33:26 databases make it gptable and all of a
1:33:29 sudden this dormant data comes back to
1:33:31 life and and in remarkably powerful ways
1:33:34 so as companies start to get the taste
1:33:37 of that they're going to be like oh
1:33:39 we've got 30 years of customer data
1:33:41 we've always meant to do something with
1:33:43 let's go dig into that and pull the good
1:33:45 ideas out of it and then you know some
1:33:48 Cowboy is going to be like I can do that
1:33:50 in my Google sheet if they dump it all
1:33:53 in and all of a sudden all the you know
1:33:55 proprietary [ __ ] is dumped out into the
1:33:57 world
1:33:58 um I I think that's going to be a huge
1:34:00 huge
1:34:01 um area of demand I think there's going
1:34:03 to be entirely new
1:34:05 um job categories in AI operations the
1:34:08 kind of stuff that you're talking about
1:34:09 so
1:34:10 power on get [ __ ] just lean into it
1:34:14 man you are going to be an in-demand if
1:34:17 you think you're a tired Runner now
1:34:18 tired Runner
1:34:20 you get us a joke about your your name
1:34:23 and how you're because you're probably
1:34:24 busy because trying to keep up with the
1:34:26 idiots in the AI
1:34:28 all right we're moving on now
1:34:31 more importantly Kyle an actor and a
1:34:34 writer's strike I know right and and for
1:34:37 both both of them have major AI issues
1:34:41 right the writers are like
1:34:43 um hey producers you can't write scripts
1:34:46 with AI without us or or at all
1:34:50 and then the producers are like okay or
1:34:53 we can
1:34:55 as and then and then after a really long
1:34:58 delay in the actors strike the producers
1:35:02 came back their first Salvo out of out
1:35:04 of like this
1:35:05 ridiculous delay was hey we want to use
1:35:09 background actors and we want to pay
1:35:10 them for one day and then own their
1:35:12 likeness for in perpetuity where we
1:35:15 never have to pay them again so so
1:35:18 uh yeah I know
1:35:23 shit's changing business models are
1:35:26 changing disruption is coming the Panic
1:35:29 the Panic have you been watching because
1:35:32 there's all these tornadoes lately like
1:35:34 I feel like on Tick Tock about every
1:35:36 seventh video is some like Cloud that's
1:35:40 doing this and then it kind of turns
1:35:42 into a funnel cloud and it's where NATO
1:35:44 and just and then up go the rooms it's
1:35:48 like I feel like
1:35:50 this is one of the like like
1:35:52 all these sites showing up in in like
1:35:55 this 36 hour period and just getting
1:35:58 better and and you know Elon announcing
1:36:00 his [ __ ] and then all these lawsuits and
1:36:03 the AP deal and just all the in the
1:36:05 writers strikes I just feel like we're
1:36:09 looking up it's clearly a storm cloud up
1:36:11 there and and we're starting to see them
1:36:13 start to rotate I feel like that's we're
1:36:16 right at that phase where it's like hey
1:36:17 that's interesting you can see those
1:36:19 clouds move oh is that oh my oh no run
1:36:24 oh
1:36:29 banks are gonna get spicy it's gonna get
1:36:31 oh look what I did to my hair we should
1:36:34 never do that again
1:36:37 well it's not yeah
1:36:40 well quaffing I could use some quaffing
1:36:43 coaching maybe I need to go to Chachi BT
1:36:46 the lyrics to the temu I feel it could
1:36:49 have been AI generated the lyrics are
1:36:51 strange and creepy possibly could be
1:36:54 absolutely
1:36:55 what's going to be interesting
1:36:58 so
1:36:59 artists that I know that are doing
1:37:01 really interesting work that are AI
1:37:03 literate
1:37:05 there's not a clear boundary between
1:37:08 their work and the ai's work
1:37:11 why because they're collaborating with
1:37:15 it so they come up with an idea and then
1:37:17 they start a poem and maybe they have ai
1:37:20 write some lyrics and like oh that's
1:37:21 good and then they take that and they
1:37:23 build on it same with images where the
1:37:26 the people that I know that are doing
1:37:28 image Art like like the stuff that just
1:37:31 rises above the noise where you're like
1:37:33 holy [ __ ] that's brilliant
1:37:35 that's often like a thousand iterations
1:37:38 in you know 16 layers in Photoshop where
1:37:42 they're they're compositing and doing
1:37:44 [ __ ] in Photoshop and using generative
1:37:46 fill and mid-journey and mixing this
1:37:48 stuff almost everyone I see that's doing
1:37:50 anything interesting is is some sort of
1:37:53 hybrid AI you know collaboration
1:37:58 is it the computer or just the sounds
1:38:01 echoing in my head it could be that
1:38:03 Emilio's life you never know
1:38:06 chat GPT seems less smart recently I
1:38:09 agree with that what are some of the
1:38:11 Alternatives you would suggest
1:38:12 [Laughter]
1:38:17 right here
1:38:18 these are the Alternatives I suggest so
1:38:22 um so if you're talking about gpt4 so so
1:38:25 if you're at chat.openai.com and you've
1:38:28 paid for chat GPT Plus
1:38:32 and you're using gpt4 and it still sucks
1:38:36 I don't know that any of these other
1:38:37 ones are better but they're different so
1:38:39 you've got
1:38:42 um gpt4 there Bing is gpt4 but it's
1:38:46 different I think the the writing from
1:38:48 what I've seen at being in creative mode
1:38:52 so there's three modes creative balanced
1:38:55 and precise in creative mode it's not
1:38:57 bad but it's it's slightly too creative
1:39:00 claude.ai is good that's
1:39:04 um Claude 2 which is the gpt4 competitor
1:39:08 from anthropic which is a bunch of X
1:39:09 openai people
1:39:12 um pose got Sage but Poe has a bunch of
1:39:15 bots that could be interesting I'm sure
1:39:17 there's a bunch of writing bots in there
1:39:19 and then bard.google the.com
1:39:22 has historically been laughably bad but
1:39:25 all of a sudden today doesn't suck so I
1:39:28 feel like
1:39:29 you know do some experimenting the nice
1:39:31 thing about Poe is there's six different
1:39:33 models there you can compare and
1:39:35 contrast so what I would do is I would
1:39:37 take if you've got a good solid prompt
1:39:39 that that you've written I would just go
1:39:42 down the line and and start uh comparing
1:39:45 and contrasting
1:39:47 um yeah something something is
1:39:49 definitely going on at open AI where I I
1:39:52 think this is I think this is legal woes
1:39:54 I I think I think they're cutting deals
1:39:56 and they're trying to figure out how to
1:39:58 not get their asses suit out of business
1:40:01 um
1:40:02 and so so so they're probably
1:40:05 fine-tuning the algorithms to be
1:40:09 less generous with the
1:40:14 um
1:40:16 inspiration from copywritten material
1:40:22 oh good lord
1:40:25 I you know and it's funny that's that
1:40:27 you know that's another reason Amelia's
1:40:29 wife you're asking about meta you know
1:40:32 releasing that model
1:40:34 the downside of having
1:40:38 the fine-tuning of an open AI or a
1:40:42 Google bard or an anthropic behind the
1:40:45 firewall
1:40:47 is that we don't know the changes
1:40:49 they're making right we don't know the
1:40:52 changes in the algorithm
1:40:54 and so
1:40:56 we just instinctively go it isn't as
1:40:59 good at writing as it used to be or it's
1:41:01 like like my buddy Greg that I work with
1:41:04 he sent me a note he goes is it me or is
1:41:07 is gpt4 worse at coding
1:41:09 and and this is all within the last
1:41:11 month or so and it's you know it it's
1:41:13 all kind of coinciding with them pulling
1:41:15 their Bing things so I have a feeling
1:41:18 they're just they're just dialing some
1:41:19 [ __ ] back
1:41:21 um and that's why it's not good so
1:41:23 they've got to strike that balance
1:41:25 because they they can't they can't be in
1:41:27 Perpetual lawsuits or they won't they
1:41:29 won't be able to build anything so I
1:41:31 think my instinct is they're just
1:41:32 dealing with
1:41:33 cutting as many deals as they can the AP
1:41:35 deal is a really big [ __ ] deal that
1:41:37 feels very symbolic to me
1:41:39 considering they just pulled internet
1:41:42 access because of the Fortune magazine
1:41:44 issue
1:41:49 all right
1:41:51 smart GPT smaller font
1:41:54 but your pick fits in there just right
1:41:58 what's the best stock for AI don't ask
1:42:01 an actor about for stock picks
1:42:05 it's all of the clues are in the [ __ ] I
1:42:08 have on my screen all the time
1:42:11 there's your Clues
1:42:13 throw inflection in there
1:42:17 let this chat gbt
1:42:19 why does it have a hard time with a PDF
1:42:22 other AIS are better at it
1:42:24 I don't know I think it's quite
1:42:27 so have you is it bad with a PDF using
1:42:31 Code interpreter because I think it's
1:42:33 [ __ ] brilliant with a PDF using Code
1:42:35 interpreter
1:42:37 Isaac or Isaac you or Issa Q
1:42:41 or
1:42:43 The Rose
1:42:46 um can we get jobs in the real world if
1:42:49 you've got talent
1:42:51 can we get jobs in the metaverse no
1:42:53 one's in the metaverse
1:42:55 the metaverse is coming let the
1:42:57 metaverse let Apple figure out
1:43:00 the hardware piece of it they're on
1:43:02 their way to do that give them five
1:43:04 years
1:43:06 the the problem with the current
1:43:09 metaverse is that human beings are
1:43:11 needing to build it right now we
1:43:13 actually need AI to get a lot more
1:43:16 sophisticated so you can just speak
1:43:17 worlds into existence
1:43:20 and everyone's got their 3D AI Avatar
1:43:24 that represents them in all of their
1:43:26 different permutations so like that's
1:43:28 all being built right now so I feel like
1:43:31 in three to five years we'll start to
1:43:33 have the beginnings of Ready Player one
1:43:34 and then 10 years out we'll be stuck in
1:43:37 those apartments
1:43:39 I saw a slideshow today of how many llms
1:43:41 there are and some are way larger than
1:43:45 chat GPT I'm so naive I know me too I
1:43:48 know it's like it's like what's coming
1:43:50 is just like I just it's it it does like
1:43:54 it it really does feel like the storm
1:43:56 clouds above us and it's slowly doing
1:43:58 this or it's that tsunami thing where
1:44:00 like the low tide went way out like
1:44:03 miles out and you're like oh it's so
1:44:05 calm and peaceful let's go pick up some
1:44:07 seashells and then you get this rumbling
1:44:10 in the distance
1:44:12 I totally feel like that's where we are
1:44:14 right now I don't think Jonah Hill would
1:44:16 approve of your VW bus against his toxic
1:44:18 boundaries that's awesome
1:44:20 uh uh yeah
1:44:23 oh these these uh these kids these days
1:44:28 huh
1:44:30 all right
1:44:32 I'm winding up my my voice is getting a
1:44:34 little hoarse
1:44:36 I gotta have more coffee because it's
1:44:37 ready I'm almost ready for bed let me
1:44:39 have a little more coffee
1:44:43 what's wrong you gotta go pee what do
1:44:45 you want you wanna go visit mama what do
1:44:47 you want to do
1:44:48 come on
1:44:59 all right
1:45:02 [Applause]
1:45:07 can you use mid-journey for sure can you
1:45:11 use mid-journey images for anything in
1:45:13 business
1:45:13 um here's the
1:45:15 it's it's a very gray area right now
1:45:17 mid-journey has definitely been trained
1:45:19 on copywritten images
1:45:22 what I would say is
1:45:25 um
1:45:27 it depends on where your risk tolerance
1:45:29 is if you've got a decent one
1:45:32 um
1:45:33 I would not use the names of specific
1:45:36 artists who still have copyright claims
1:45:39 on any work that might be in there so
1:45:41 you know don't use a specific artist
1:45:43 name
1:45:45 and
1:45:46 um
1:45:47 because those artist names
1:45:50 um unless you know what you're doing are
1:45:52 going to live in the metadata of the
1:45:54 image and one of the ways
1:45:57 um artists are probably going to come
1:45:58 after
1:45:59 companies or images that are out there
1:46:02 is go
1:46:04 build an AI to go search the internet to
1:46:07 find all the images metadata and pull up
1:46:10 their name anytime someone used a name
1:46:12 so find a way to make images where you
1:46:15 don't use an actual artist name is is a
1:46:17 way to do it
1:46:19 if you want if you're more conservative
1:46:20 don't use mid-journey use Adobe Firefly
1:46:24 and generative fill within Photoshop
1:46:26 because they're both good they're not as
1:46:28 good as mid-journey but
1:46:32 they're different and all of the Adobe
1:46:35 training models were trained on
1:46:39 um public domain and Adobe stock images
1:46:41 where they own all the rights so Adobe
1:46:44 is making generative AI safe for
1:46:47 business so if you really want to be
1:46:48 safe in business use the Adobe products
1:46:51 no Kyle you need goblin.tools to
1:46:53 organize your desk all right I'll go
1:46:55 check that out I don't know what
1:46:57 goblin.tools is or is that a joke should
1:46:59 I know that I don't know what
1:47:00 goblin.tools is is that a thing
1:47:03 should I go there is this going to give
1:47:05 me some sort of malware better not give
1:47:07 me malware Vicky
1:47:11 Goblin dot what is it dot tools
1:47:16 Goblin
1:47:18 dot tools
1:47:22 magic to do
1:47:25 yeah
1:47:28 to-do lists and add do not go well
1:47:30 together
1:47:33 add a new item
1:47:35 um
1:47:38 keep up
1:47:41 with AI
1:47:43 oh I like this so I I do like that this
1:47:46 is just
1:47:47 you just start typing in your tools and
1:47:50 I assume it's using AI to organize them
1:47:52 and sort them and [ __ ] like that I might
1:47:53 actually use this
1:47:56 the ones that drive me [ __ ] crazy the
1:47:57 ones that I can't I just can't use are
1:48:00 our um ones where you've got to like
1:48:03 come up with your own [ __ ] hierarchy
1:48:05 like Trello like all you have to do is
1:48:08 name all your columns and then populate
1:48:10 all your columns and then create
1:48:11 templates to create cards to put in your
1:48:14 columns that you created I'm like I just
1:48:16 like by the time I organize my [ __ ] like
1:48:19 it's going to take me a month to come up
1:48:20 with the infrastructure to organize my
1:48:22 [ __ ] I just need to put my [ __ ] in my
1:48:24 [ __ ] and have it work so if goblin.tools
1:48:27 does that then
1:48:28 you have a convert are you real are you
1:48:31 related to Goblin dot tools there Vicky
1:48:33 if so like my initial instinct is it's
1:48:36 better than any other that [ __ ] that
1:48:38 I've seen every single tool that I'm
1:48:40 like you should go no use notion or you
1:48:42 should go use Trello it's so inspiring
1:48:44 that's all said by like project managers
1:48:47 that have good executive function
1:48:49 [ __ ] all right I have a question
1:48:51 about AI higher can I get that through
1:48:54 to you wait what I have a question about
1:48:57 AI higher
1:48:59 uh yeah you can DM me on Tick Tock you
1:49:02 can also go to my profile and I've got a
1:49:05 link tree there that's got my contact
1:49:06 info in it so who's what's your name
1:49:09 stopping
1:49:10 Waits
1:49:14 stopling
1:49:17 stop in grow wait stop oh stop ingrown
1:49:20 hair
1:49:22 sorry sorry too many words too many
1:49:26 words too close together there my
1:49:27 brain's not parsing I think it's getting
1:49:29 late I'm getting a little sleepy
1:49:32 I'm a little punch drunk predict old
1:49:35 movies will be re-released as generative
1:49:37 filled 360 environments for VR glass 360
1:49:40 theater yeah I think that's a good one
1:49:42 Gregory I also think that um
1:49:46 I think we're going to be able to
1:49:50 um recast movies
1:49:53 like I want to watch an old movie but I
1:49:55 want you know some new star I want
1:49:58 Philip Seymour Hoffman starring in it
1:50:03 plug into this AI thing on your own
1:50:06 program I don't quite know what you mean
1:50:08 but I assume you mean on using your own
1:50:10 data where to start is just start
1:50:14 learning these tools then I would start
1:50:16 looking at automation tools like zapier
1:50:19 where you can start to have it do its
1:50:21 you can pull data from different sources
1:50:24 process it in these AI tools and then
1:50:27 get it back out to other sources
1:50:29 is a way to do it and then if you really
1:50:32 want to start digging into it look into
1:50:35 a project called Lang chain where you
1:50:37 can start training your own these large
1:50:39 language models on your own data but
1:50:41 that requires some geekiness so I don't
1:50:44 know where you are on the geeky scale if
1:50:47 you're on the very low end of the geeky
1:50:49 scale Lang chain might be just like yeah
1:50:52 it's a little much you're going to need
1:50:53 to find someone that knows how to do
1:50:54 that thank you for the hard on the hit
1:50:56 hard on ahead hot on the head I
1:50:59 understand I'm kind of hot on their head
1:51:02 foreign
1:51:05 oh God it's so bad why do people stay
1:51:10 you should leave
1:51:11 [Laughter]
1:51:14 I miss your comments about Elon Musk uh
1:51:19 oh
1:51:21 that he's launching a
1:51:24 well he's already announced that he's
1:51:26 creating this new company called x.ai
1:51:29 they launched their Twitter account
1:51:30 today apparently something is happening
1:51:32 tomorrow they're going to announce
1:51:34 something or there's going to be a Bard
1:51:35 like chat GPT like competitor or there's
1:51:38 going to be some announcement or he's
1:51:40 going to say we've discovered Life on
1:51:42 Mars but you can only access it through
1:51:44 my chat gbt thing but it hasn't launched
1:51:46 yet so wait till I launch the rocket and
1:51:48 then drive your Tesla to the thing and
1:51:50 tweet with the blue check mark to the
1:51:52 thing and then you can get into my chat
1:51:54 GPT but I won't call it chat EBT because
1:51:57 I hate Sam Altman because I invested in
1:51:59 a non-profit and he made it for profit
1:52:01 so something around that I think that's
1:52:03 the announcement tomorrow
1:52:06 pretty sure that's it
1:52:08 and I've got inside information
1:52:11 I'm just saying you know what I mean I
1:52:14 mean listen
1:52:15 don't trust me yeah you know what I'm
1:52:18 saying
1:52:19 don't trust this guy
1:52:21 you want to know what my qualifications
1:52:23 are
1:52:24 zero bookka nothing I got nothing
1:52:29 so but you know
1:52:31 is Elon a friend of mine he could be
1:52:34 he could be
1:52:36 did I talked to Zuck this afternoon
1:52:37 maybe
1:52:39 maybe I did maybe I do have
1:52:40 qualifications you ever think about that
1:52:42 what if I did what if I did have
1:52:44 qualifications what if this channel was
1:52:46 a ruse for someone who was actually
1:52:48 qualified
1:52:49 it's very possible it's very possible
1:52:51 all right carrying on
1:52:54 carrying on
1:52:56 put Kyle in your profile name
1:52:58 so you're easy to find that's not a bad
1:53:00 idea which profile name my my that
1:53:03 profile name
1:53:08 I kind of like this because I had a
1:53:10 comment once we were on the live and
1:53:11 they were like how big is this lab
1:53:14 it was like the comment what are your
1:53:17 qualifications
1:53:18 is this lab you work in I want to know
1:53:21 more about this AI learning lab do you
1:53:25 have computers there
1:53:28 who exactly is partying what is your
1:53:30 role in this learning lab I kind of like
1:53:34 that it's
1:53:37 people are like if it's a lab why is he
1:53:39 sitting in a home office with crap on
1:53:41 the shelves
1:53:45 he seems a little touched is this really
1:53:48 learning
1:53:50 is this really do you think he's a
1:53:52 scientist
1:53:53 where's his white coat oh he doesn't
1:53:56 wear a white coat because of the text I
1:53:58 get it
1:53:59 but he he's a he is a he's a scientist
1:54:01 right he's where are the Petri dishes
1:54:05 I think we're gonna keep it vague for a
1:54:07 while so sick of the overload of sales
1:54:09 and gamer streams oh my God I know
1:54:12 I land in those rooms way too often I
1:54:15 know it's just so painful now
1:54:18 it's like
1:54:20 you know I ordered these cups from The
1:54:23 Tick Tock shop and I you know they're
1:54:26 they're way more fantastic than I
1:54:28 thought they would you know what they
1:54:30 can do they can hold liquid
1:54:33 laughs
1:54:34 oh my God that's so good anyway if you
1:54:37 want to buy those
1:54:39 foreign
1:54:44 I buy
1:54:46 oh
1:54:49 I gotta I could do a t-shirt for for UK
1:54:52 money with chat tvt but it's I'm gonna
1:54:54 have to figure out a way to spell it I'm
1:54:56 I need help from the community on this
1:54:58 one you can make money
1:55:01 so you gotta we gotta respell
1:55:06 we've got to figure out how to spell
1:55:08 Monday you're gonna make money when I
1:55:10 touch your big pie
1:55:12 is it t-a-y with a lot of y's
1:55:16 I don't know we'll figure it out that
1:55:18 would be a good t-shirt actually now
1:55:20 that you now now that you speak of it
1:55:23 AI learning lab you can make money
1:55:26 with like a guy with really bad hair
1:55:29 like
1:55:32 we could also do the
1:55:39 I got my mind blown with chat GPT and
1:55:41 all I got was this stinking t-shirt all
1:55:43 right how did the Samsung thing leak
1:55:46 happen so they they basically just used
1:55:48 so if you leave your chat history
1:55:52 in chat GPT
1:55:55 it's basically not encrypted or whatever
1:55:58 like I don't like I don't know
1:55:59 technically how they did it but but
1:56:01 basically if you want the [ __ ] you put
1:56:04 in chat GPT to be safe you need to go
1:56:07 into your data settings and turn off
1:56:10 your chat history
1:56:12 and uh and and tell them you don't want
1:56:15 your data shared right and so it was
1:56:18 before that option existed so someone
1:56:20 from Samsung just
1:56:22 one of them put it might have been the
1:56:24 same person put
1:56:26 um a transcript from a meeting that was
1:56:28 like a sensitive private meeting and
1:56:31 then they also put in some private
1:56:32 sensitive data into the conversation and
1:56:35 then someone hacked it through their um
1:56:38 through their chat history
1:56:39 so
1:56:41 however they did that
1:56:43 but but you can turn that off if you
1:56:45 want it um safer the deal with that
1:56:48 though is
1:56:49 um if you turn that safety setting on it
1:56:53 deletes your history and you can't get
1:56:55 it back so make sure you export your
1:56:57 data first so you get your chat history
1:57:01 um do you think Google also uses Moe I
1:57:05 forget what Moe is but they probably do
1:57:07 I I don't know I forget what it is
1:57:10 though it's it's I forget what it is
1:57:12 um I don't know
1:57:18 Kyle's loopy it's true we're staying for
1:57:21 the show yeah exactly it's just gonna
1:57:25 watch me just like stroke out and fall
1:57:27 off my chair oh it's a good show
1:57:30 he'll get up
1:57:35 oh I PDF and Claude today to extract and
1:57:39 it was great awesome no Adobe also
1:57:42 trained on all product users images
1:57:46 so they're gonna come after them too no
1:57:48 that's not true
1:57:49 that's not true they've explicitly said
1:57:51 I mean listen they might have lied but
1:57:54 they've explicitly said it's trained and
1:57:57 and I can tell you so so I did a test on
1:58:00 Adobe Firefly in fact Let me let me go
1:58:03 try it again and see if anything changed
1:58:08 firefly.adobe.com
1:58:15 text to image generate
1:58:22 why does my image look all washed out I
1:58:24 have a feeling something happened with
1:58:25 some filter or other I look like some
1:58:27 sort of pasty claymation dude
1:58:30 or maybe I'm just [ __ ] Pasty
1:58:35 that's possible too
1:58:37 hey dang it
1:58:43 um you know it comes who said the day
1:58:48 I'm gonna do a spearmint here to Dolby
1:58:56 all right
1:59:01 1970s muscle car in and
1:59:06 abandoned
1:59:07 [Music]
1:59:09 Factory
1:59:12 I will tell you in a few minutes if I
1:59:15 still think that this is
1:59:18 public domain images
1:59:21 yes so let me tell you why
1:59:25 ours none of these are 1970s looking
1:59:28 cars these are like 19
1:59:31 60
1:59:35 four maybe five kind of cars
1:59:39 that sort of looks like the 70s
1:59:43 that doesn't well the the hood does
1:59:48 but like that's definitely 60s that's
1:59:51 definitely 60s so what this says to me
1:59:54 is that
1:59:55 these are Tran they picked some
1:59:57 arbitrary point for public domain to say
2:00:00 we're not going to train on anything
2:00:02 more modern than it's probably like you
2:00:05 know 1967 or something like that
2:00:08 um and I just stumbled on Leonardo
2:00:17 um
2:00:18 [Laughter]
2:00:21 [Music]
2:00:23 introducing Leonardo Academy I think
2:00:26 I've got some that I produced down here
2:00:29 yeah like there's some 70s muscle cars
2:00:31 in Leonardo
2:00:34 um
2:00:35 where's why can I not see the rest of my
2:00:37 images
2:00:40 yeah like here's the abandoned Factory
2:00:42 ones like like these are that's us
2:00:45 that's a 70s muscle car right
2:00:48 that's a 70s muscle car
2:00:51 so so what it says to me is that the
2:00:56 Adobe
2:00:58 um training model is
2:01:01 you know is time constrained kind of
2:01:04 like chat gbt goes up to September 2021.
2:01:08 so
2:01:09 now is it possible they trained it on
2:01:11 copywritten [ __ ] it's possible but but
2:01:14 what I what I do know is this is that
2:01:16 they what they they claim is public
2:01:18 domain images and
2:01:20 um
2:01:22 Adobe stock images that they have the
2:01:24 rights to and then what they've said is
2:01:26 they're going to create systems to allow
2:01:29 artists to opt out of future training or
2:01:31 opt into future training then they're
2:01:34 going to work on an attribution model
2:01:35 then they're going to work on a
2:01:36 compensation model like like a streaming
2:01:38 model so
2:01:41 um
2:01:41 if they are being nefarious and they
2:01:44 just trained it on copywritten [ __ ] then
2:01:46 um that would suck but that's what
2:01:48 they're claiming
2:01:52 Adobe retroactively changed their terms
2:01:54 of service to the to allow the use of
2:01:56 user work to train their AI
2:02:01 to allow the use of user work is that to
2:02:03 allow the use of user AI generated work
2:02:06 to train their AI because it would make
2:02:08 sense if they were re-ingesting AI
2:02:10 generated stuff but I don't know it's
2:02:13 interesting maybe listen I mean Google's
2:02:17 original corporate motto was first do no
2:02:19 evil or do no evil and
2:02:22 they gave that up so so adobe's a big
2:02:26 company but that's really interesting
2:02:28 I'll look into that because I know
2:02:30 someone over there
2:02:31 I will uh I can actually probably get an
2:02:33 answer to that one I might not be able
2:02:35 to say it but
2:02:37 all right I got I got a few more minutes
2:02:40 I got I got it I don't know what I can't
2:02:44 anymore
2:02:45 do we care what is your qualifications
2:02:47 no we don't queer you are one cool dude
2:02:50 man at least you thank you design your
2:02:54 T-shirt with mid-journey I will
2:02:57 um put it in Spongebob case okay
2:03:02 you can make money watch out every day
2:03:05 oh that's good doing it all as one giant
2:03:08 word and they're like
2:03:11 oh oh you know you know what we'll do
2:03:13 okay I'm starting to get some ideas for
2:03:16 you can make money with chat gbts we'll
2:03:18 do like a fake it'll do like a a
2:03:20 graphicized tick tock frame and then
2:03:23 I'll just have a face in it like it'll
2:03:25 be like you know
2:03:27 GPT dollar marks will be the the thing
2:03:30 on the screen with the face and then
2:03:32 around it'll be you can make money
2:03:35 foreign
2:03:42 we're gonna do a merch store this is I'm
2:03:44 telling you the AI learning lab is going
2:03:46 to turn into a capitalistic cash grab
2:03:49 [Laughter]
2:03:51 it's gonna we're gonna we're gonna put
2:03:52 out an nft it's gonna be awesome just
2:03:54 trust me on this
2:03:56 the uh the AI learning lab crypto coins
2:03:59 gonna make you a [ __ ] fortune I want
2:04:01 that shirt okay
2:04:05 oh y'all is funny all right would you
2:04:08 summarize your 14 Chrome tab since I
2:04:11 missed them yeah so I just showed
2:04:14 so I showed chat GPT Bard Bing
2:04:19 Claude Pi Poe
2:04:22 so so all of the sites that that are
2:04:25 have been hanging out on screen plus pi
2:04:28 dot AI
2:04:30 then
2:04:33 um
2:04:34 I talked about the open Ai and
2:04:36 Associated Press deal
2:04:39 um then I just had zapier there as a tab
2:04:41 because sometimes people want to talk
2:04:43 about that
2:04:44 then I had the Elon Musk
2:04:47 um is launching their
2:04:49 their x dot AI they launched their
2:04:52 Twitter today and it sounds like they're
2:04:53 going to do something tomorrow
2:04:56 and then Bard got good and then oh and
2:05:00 then the gpt4 architecture apparently
2:05:03 has been leaked not not the core model
2:05:06 but like what all the architecture is
2:05:08 like like it's it's ten times bigger
2:05:12 than people thought it was it's got all
2:05:14 these expert AI systems within it it's
2:05:18 it's it's a it's a re the reason gpt4 is
2:05:21 better is that it's significantly more
2:05:24 complicated than they were letting on
2:05:27 um and then oh and then then the other
2:05:29 big one was
2:05:30 um
2:05:31 uh meta
2:05:33 Zuckerberg said today that they've got a
2:05:37 30 billion parameter model that's as
2:05:40 strong as GPT 3.5 and he thinks they're
2:05:44 just gonna release it
2:05:46 into the wild you know with all the
2:05:48 weights
2:05:49 commercially licensed you know as a
2:05:51 commercial license so that you can build
2:05:54 businesses on top of it and that that
2:05:56 would be a huge deal so it's just like
2:05:58 it it's like I added you know I added
2:06:02 two additional things to this list and
2:06:04 then just in what I was talking about
2:06:05 right there there's potentially three or
2:06:07 four more that are going to be added to
2:06:09 this so and and they're all going to be
2:06:11 fighting for our attention right and and
2:06:13 we're gonna we're gonna you know I I
2:06:16 said this before like some someone asked
2:06:18 should I should I buy the annual
2:06:20 subscription for whatever it was and I'm
2:06:23 like no you should be doing month to
2:06:24 month on everything because this is
2:06:26 gonna ebb and flow we're gonna be like
2:06:28 oh that's the best but it cost me 20
2:06:30 bucks a month oh now that's as good as
2:06:32 that but it's free okay I'll use that
2:06:33 one and then we're just gonna be you
2:06:35 know we're gonna be like cheap because
2:06:37 that's what we are
2:06:39 because we don't have imaginations
2:06:42 I'm thirsty have you bought a beautiful
2:06:45 Crystal tumbler from my Tick Tock shop
2:06:47 it's right there
2:06:51 um
2:06:52 oh it's so much better in that glass I
2:06:55 don't know why it tastes better
2:06:58 have you know this this is another pet
2:07:01 peeve on on on Tick Tock is is
2:07:05 these prank channels where they've
2:07:09 clearly written a script and it's like
2:07:11 it's like worse than porn acting they're
2:07:14 they're like
2:07:16 watch and see if if Sally notices if I
2:07:19 sneak up on her and then she's like oh
2:07:22 you scared I'm like really come on
2:07:25 just knock it off
2:07:28 take an acting lesson and a writing
2:07:31 lesson it's like I don't care if it's
2:07:33 tick tock and you can do it for free it
2:07:35 shouldn't suck that bad
2:07:38 it's like God damn it
2:07:41 oh
2:07:42 My secret is out wait what was your
2:07:44 secret oh did someone found out that
2:07:46 you're um that you know AI
2:07:50 I was going to send you a t-shirt with
2:07:52 you oh
2:07:58 [Laughter]
2:08:01 that I'm gonna I'm gonna do that that's
2:08:04 good there's something good there
2:08:07 AI learning lab with Kyle Shannon kind
2:08:09 of like painting with Bob Ross okay I
2:08:10 might do that I could do that I don't
2:08:13 know it's a little I I don't know that
2:08:15 my ego needs any more feeding it's
2:08:17 already healthy it's I got a healthy one
2:08:20 all right
2:08:22 we'll call it chunky dog to come if Kyle
2:08:25 needs help yeah exactly
2:08:27 chunky chunky champy
2:08:30 how does he do the picture in picture on
2:08:33 stream
2:08:35 because I'm magical
2:08:38 I said to The Tick Tock I would like
2:08:41 picture in picture no one else should
2:08:43 have it they gave it to me
2:08:46 because I'm influential
2:08:48 some would call me an influencer
2:08:51 uh it's just dual
2:08:53 you click a button when you go live but
2:08:56 what I do know is that they took it away
2:08:58 from me for like a month for no reason
2:09:00 and then they gave it back for no reason
2:09:01 so if when you go live you see this
2:09:04 thing that says duel and it kind of
2:09:06 looks like two little cameras two little
2:09:07 Instagram logos together that's that
2:09:11 it's just the back cameras you know
2:09:14 looking at my laptop which is sitting on
2:09:16 top of a box
2:09:19 because I'm fancy because I'm
2:09:20 sophisticated
2:09:25 [Music]
2:09:28 any plugins for stock info or maybe in
2:09:31 the future hopefully yes go to
2:09:35 if you if you pay for a chat GPT Plus at
2:09:38 chat.openai.com
2:09:40 and you go into your settings and enable
2:09:43 the beta features and go into plugins
2:09:44 then just go to the plug-in store and
2:09:46 look look for stocks there's there's
2:09:47 lots of plugins in there now are any of
2:09:50 them any good I don't know
2:09:52 I I ain't a stock Trader
2:09:55 because I'm too emotional AI might help
2:09:57 me with stocks because it can do things
2:09:59 not emotionally and just you know just
2:10:00 [ __ ] turn it loose and Let It Go
2:10:03 um but I'm sure there's some stuff there
2:10:05 possible risk for financial advice yeah
2:10:07 I'll give you Financial advice don't ask
2:10:09 me for financial advice that's the best
2:10:11 financial advice you've heard tonight
2:10:14 merch yes okay
2:10:18 oh my God we're starting a merch store
2:10:21 laughs
2:10:25 all right
2:10:26 Metals threads through Instagram wait
2:10:31 oh meta's threads through Instagram
2:10:33 launched yeah and they got to 100
2:10:35 million users in
2:10:38 five days six days
2:10:40 which which was faster than Chachi PT's
2:10:43 hundred million in six weeks but when
2:10:46 you start with a billion Instagram users
2:10:48 and just say click to import all your
2:10:50 Instagram friends that's kind of
2:10:52 cheating
2:10:54 please do summaries like this every
2:10:56 night Kyle thanks I will listen I mean
2:10:59 to be clear the the last month there
2:11:02 really hasn't been a ton of news like
2:11:04 it's there's been little things here and
2:11:06 there but this week all of a sudden like
2:11:09 I just feel like all this stuff must
2:11:11 have just been simmering in the
2:11:12 background and it's all just kind of
2:11:14 like it's all coming out right now
2:11:16 and it makes sense we're kind of six
2:11:19 months if if you think about development
2:11:21 you know of software
2:11:24 um large language models have been
2:11:26 around for a while like GPT 3 was
2:11:29 launched a year and a half before chat
2:11:32 GPT came out
2:11:33 so these companies building the large
2:11:35 language models and and building on the
2:11:37 large language models like they've been
2:11:39 around for a while so chat GPT launches
2:11:42 in November it catches on fire and
2:11:45 everybody sort of panics and then
2:11:47 basically everyone started developing
2:11:48 probably in January
2:11:51 so it makes sense that we're kind of
2:11:53 starting to see
2:11:54 stuff come out of the woodwork six
2:11:57 months later like that feels like a
2:11:58 right time to me to actually go build
2:12:00 something that is decently functional
2:12:02 that that you can push out past an alpha
2:12:06 so that's probably why all this is
2:12:08 happening right now but it's it's weird
2:12:09 like this is a weirdly dense Newsweek
2:12:13 and I don't think it's over like I I
2:12:14 would imagine like depending on what
2:12:17 Elon talks about tomorrow
2:12:19 and like like The Bard thing like Bard
2:12:23 just getting that much better like
2:12:25 seemingly overnight like I didn't see a
2:12:27 lot of talk about it it's just like all
2:12:28 of a sudden it's better I there's some
2:12:30 weird [ __ ] happening so I have a feeling
2:12:32 some stuff's coming and then like the
2:12:34 the AP deal with open AI with them
2:12:37 pulling access to the internet but Bing
2:12:40 not pulling access to the internet I
2:12:42 just there there is some [ __ ] and Deals
2:12:45 going down right now I just I have a
2:12:47 feeling
2:12:49 they're burning the midnight oil in San
2:12:51 Francisco these days
2:12:54 but yeah I'll try to if there's if
2:12:55 there's if there's news I'll try to do
2:12:57 it
2:12:58 I love you Kyle slash Kyle thank you
2:13:01 very much and the bad acting takes 15
2:13:03 minutes to get to the good part exactly
2:13:07 exactly the bad
2:13:09 I come for this I come for the AI I stay
2:13:13 for the bad acting or the hair
2:13:16 uh have you seen the vibe computer
2:13:18 monitor maybe good for your next AI lab
2:13:22 uh uh uh
2:13:27 five computer monitor
2:13:39 oh it's a big interactive touch screen
2:13:41 thing
2:13:45 um
2:13:45 [Music]
2:13:49 that's cool
2:13:51 I remember in the in the 90s all these
2:13:55 uh
2:13:56 smart white boards came out that
2:13:59 projected it was like a a projector in
2:14:03 the room projected on it and then it had
2:14:04 a touch screen and it was all synced up
2:14:06 smart white boards were out that reminds
2:14:08 me of one of those
2:14:10 my dude you're on I am the bear
2:14:13 oh thank you that's very sweet
2:14:16 um smart white boards yeah exactly
2:14:19 all right listen
2:14:24 um
2:14:26 it's 11 o'clock I know where my children
2:14:29 are
2:14:31 um I'm gonna go
2:14:32 it's been a busy day look at my
2:14:36 little piece I can't keep up I can't
2:14:39 keep up I can't do it anymore
2:14:43 um anyway thank you all for hanging out
2:14:44 with me tonight I'm gonna go you're not
2:14:47 bad acting Kyle you're acting is good
2:14:49 those those badly acted real things oh
2:14:51 yeah no I wouldn't call my acting good I
2:14:54 I would say I am well committed
2:14:59 that's honestly the difference between
2:15:01 good and bad acting is committing it
2:15:04 really is
2:15:05 because the minute the the problem with
2:15:07 all those videos all the like the bad
2:15:09 porn acting in those videos is that
2:15:12 they're all self-aware they're like
2:15:13 we're doing a joke look at me trying to
2:15:17 not laugh at how funny this is
2:15:20 oh so painful it's so painful anyway
2:15:23 peace everybody I appreciate all your
2:15:26 time I'll be back tomorrow
2:15:29 and tomorrow's Friday so yeah thanks
2:15:31 thank you for acknowledging uh me
2:15:34 sharing and I'll see you tomorrow yeah
2:15:36 we'll maybe we'll get have some more to
2:15:39 talk about Elon tomorrow
2:15:41 it would be nice if he announced
2:15:43 something of substance but I doubt it
2:15:47 because he just ordered like those 10
2:15:49 000 gpus like two months ago
2:15:52 maybe three months ago
2:15:54 I don't think that's fast enough to
2:15:56 build anything but maybe he poached some
2:15:59 people from open Ai and poached some
2:16:01 other people from Twitter and he said
2:16:03 they're going to integrate it into
2:16:04 Twitter so maybe it's some sort of
2:16:06 Twitter's now got an AI chat bot we'll
2:16:09 see anyway that's speculation peace out
2:16:11 everybody good night