aiLL 7/10/23 - GPT is an Adventure, Not a Threat

Description

Chapters

Transcript

0:06 foreign
0:25 this is not showing me how many people
0:27 are here this is weird
0:30 I don't know what's going on man
0:33 tick tock's being weird tonight
0:36 oh well
0:39 I will be clueless as to who's here wait
0:44 maybe I can do it with this
0:46 no
0:49 weird
0:52 strange
0:55 hey simplex
0:56 I'm here but also in The Matrix I know
1:00 retro Punk aren't we all
1:03 how's agency life
1:06 welcome everyone my name is Kyle Shannon
1:09 they call me Hamilton
1:19 um
1:19 this is the AI learning lab I have no
1:21 idea how many people are here
1:24 um happy Monday
1:27 you got any thoughts questions about AI
1:29 pop them in the comments below I'll do
1:31 my best to answer them
1:33 we'll get oh now they're now now I see
1:35 some people there there's uh 16 of us
1:38 hate Tobias which was shaking
1:40 21 so far okay it's just it's being
1:43 really slow to update on my side so feel
1:45 free double tap the screen get the
1:47 hearts going and then if you'd be
1:49 willing to share the live and maybe
1:51 maybe we'll get in a pile 42 all right
1:53 it's it's climbing I'm only seeing 16
1:56 still so
1:59 whatever whatever I mean you know with
2:03 these platforms like they don't charge
2:05 us anything so what do you want for free
2:08 if the product's free you're the product
2:11 [Laughter]
2:14 oh good lordy welcome to Monday are you
2:18 always excited as I am for Monday I had
2:21 a crappy day at work today not not like
2:23 nothing bad happened I just got there I
2:25 like stared into the middle distance for
2:26 most of the day just like
2:30 nothing it's just like
2:33 [ __ ] flatlined all day
2:36 83 oh thanks thanks for keeping me
2:39 posted on how many folks are in here uh
2:41 this is the AI learning lab I'll show
2:42 you these URLs in a second here
2:45 um my name is Kyle Shannon I'm an
2:47 entrepreneur
2:48 of many many years many decades actually
2:50 at this point I'm an old [ __ ]
2:52 and uh I swear occasionally so if you
2:55 don't like that I wouldn't stay uh I can
2:58 be a little smarmy I try to answer any
3:00 questions I can about uh about AI so pop
3:04 them in the questions below I am not an
3:06 engineer or a machine learning guy I'm
3:09 more like an entrepreneur that deals
3:11 with innovates with technology once it
3:14 gets ready for prime time and so I kind
3:16 of Mark
3:17 the launch of chat EPT is a new new uh
3:20 era a new inflection point it's it's
3:22 actually very similar I was uh I was an
3:25 entrepreneur back in the mid 90s when
3:27 the World Wide Web first came out and
3:29 it's there's a lot of parallels between
3:30 the early days of the web and the early
3:32 days of this generative AI stuff so
3:34 that's why I'm so excited about it
3:36 that's why I started the channel so if
3:38 you have questions about it pop them in
3:40 the in the comments below I'll do my
3:43 best to answer them oh let me let me
3:44 show you
3:45 so why do I have these on screen I'm not
3:48 affiliated with any of these
3:50 URLs or these companies in any way but
3:53 if you have not played with chat GPT go
3:56 to that top URL that's the official chat
3:58 GPT website chat.openai.com
4:01 go there and start playing if you if you
4:03 have zero idea what to do if you don't
4:05 know what it is it kind of looks like a
4:07 Google search bar once you sign up and
4:08 get an account
4:09 but it's so much it's it's so profoundly
4:12 deeper than a Google search
4:15 um so if you want to know how to talk to
4:17 it go to that URL right there
4:19 prompts.chat
4:22 that's a little primer that we'll talk
4:24 to you about what prompting is what
4:26 large language models like chat GPT are
4:29 um or the underlying model is is a large
4:31 language model chat GPT is the app
4:33 sitting on top of it
4:35 um
4:35 but go start playing uh and and you owe
4:38 it to yourself if you haven't played
4:40 with chat gbt is play with it long
4:42 enough until you have your Kevin
4:43 McAllister moment which is this moment
4:47 go see if you can find that moment for
4:49 yourself then bing.com is also chat GPT
4:52 if you click on the chat button there
4:54 Bing chat is gpt4 connected to the
4:57 internet you can also tell it to make
4:59 you images and it uses the Dali model to
5:01 make you images and then poe.com gives
5:04 you access to six different large
5:05 language models
5:07 um that I think are at least worth worth
5:09 playing with you can compare them there
5:11 which is interesting there's one from
5:13 quora from Poe there's two from open Ai
5:17 and there's three from anthropic the
5:18 quad models are there so that's an
5:20 interesting site and then futurepedia.io
5:23 that's a big directory of
5:26 um what are they called uh AI apps
5:30 so go there and do that
5:34 all right let's get to some questions
5:39 so what will we be learning about
5:42 wait what were we learning how about
5:44 ethics oh what will we be learning how
5:46 about ethics
5:48 um
5:49 well I I guess what do you want to know
5:51 about ethics like like
5:53 in what context
5:55 copyright
5:57 um alignment
5:59 um politics I'm not going to talk about
6:03 that ethics and politics doesn't exist
6:05 anymore so we're good there we're
6:07 covered it's non-existent
6:11 oh
6:13 does AI have ethics
6:15 or or do they need a
6:18 a comedy I I assume you mean the
6:20 community
6:21 um does AI have ethics
6:27 well
6:30 if you you should go listen to Sam
6:32 Altman talk about you know the dangers
6:34 of AI and the benefits of AI and things
6:37 like that he's a very thoughtful guy
6:40 um
6:41 the open AI model called human
6:44 reinforcement learning
6:46 um
6:47 is involves humans so there are humans
6:51 looking at things going that's good
6:52 that's bad that's how they train it
6:56 um there are flaws with that so even
6:59 though there may be ethics there's also
7:00 bias because humans are involved in the
7:03 data that these things are trained on
7:05 have biases in them so
7:07 so there's a lot of issues with that one
7:09 of the reasons I like anthropic which is
7:11 you can play with the anthropic models
7:13 at poe.com is that they use something
7:15 called the Constitutional learning model
7:17 where there they create a set of rules
7:20 like you know increased Prosperity don't
7:23 harm people things like that
7:26 and then the the uh training is done
7:29 against that Constitution so it's not
7:32 it's not a human Constitution I think
7:34 that approach is ultimately really
7:36 interesting when it comes to alignment
7:38 or aligning the goals of the robot with
7:41 the goals of the people
7:44 because we want those two goals aligned
7:47 and if they're not that's bad because
7:49 they could just decide they don't need
7:50 us anymore if they get smart enough
7:52 they're not there yet but holy crap is
7:54 it coming Microsoft just put out a paper
7:57 yesterday or the day before about a
7:59 billion a billion token context window
8:05 which is just insanity
8:08 insanity
8:10 uh so so
8:12 these things are going to get smart very
8:15 very fast like like way smarter than
8:17 they are now do you prefer a key
8:20 do you prefer a keyboard plus ai.com
8:23 domain
8:25 oh oh keyword Plus
8:28 ai.com
8:31 or keyword dot AI
8:34 I am incredibly frustrated right now
8:37 with the domains of every company in web
8:41 3 and AI right now because
8:45 the the web 3 crypto Bros started it
8:50 with the dot XYZ and the dot this and
8:52 the dot dot I mean we've had alternative
8:54 domains you know I started one of the
8:55 first agencies in the mid 90s there have
8:57 been alternative domains for a while but
8:59 they've never caught on
9:00 and all of a sudden in the past I don't
9:03 know three or four years like I can't I
9:06 can't ever remember any sites is it.com
9:08 is it dot AI is it dot IO is it dot XYZ
9:11 is it dot bubble butt like uh
9:17 it's just [ __ ] exhausting
9:19 so like like whatever anyone asks me you
9:23 know something I can always remember
9:24 remember the keyword the name of it I
9:27 can never remember the uh the domain so
9:31 I'm an old guy so you know I'm like get
9:34 off my lawn with you Daddy eyes you know
9:36 I got my company's called storybind I
9:39 did get storyvine dot AI because we're
9:42 doing some stuff in the space and I
9:44 thought it might be important we might
9:45 need to distinguish it from the.com
9:49 um
9:50 but I suppose I'm a traditionalist I
9:52 just can't remember all the alternative
9:53 domains I mean it's fine if you're
9:56 clicking on it right if you have it as a
9:58 hyperlink somewhere it doesn't really
9:59 [ __ ] matter what it's called
10:01 but
10:03 if you expect people to remember your
10:06 your url
10:08 dot com's the way to go
10:11 which company would you recommend
10:13 applying to if you want an entry level
10:16 if you want to do entry level AI
10:19 language model do you want to work on
10:22 the
10:23 on the engineering side or do you want
10:25 to work on the applying large language
10:28 models to a business side that's there's
10:31 a big distinction there because if I
10:33 think if it's the latter
10:36 um
10:38 I I'd say it's it's not so much the type
10:41 of company but more like the type of
10:43 attitude you want to find someone that's
10:45 embracing AI you don't want to go in and
10:47 be a salmon swimming up the river if
10:50 you're looking for a new place you might
10:51 as well go to some place committed to it
10:53 if you're on the other side I don't know
10:55 open AI seems like they're hiring
10:58 um inflection just raised 1.3 billion
11:00 dollars they've got that new pie
11:03 Pi agent
11:06 they look really interesting that's an X
11:08 Deep Mind
11:10 founder
11:12 um that created inflection and then
11:15 um
11:16 anthropic I think is really interesting
11:19 um stability AI I think is really
11:21 interesting if you want to work on the
11:23 more on the research and Science and you
11:26 know building the products rather than
11:27 using the products
11:57 oh do they need a committee for ethics
12:01 so oh the question about ethics do they
12:03 need a committee
12:04 um
12:06 yeah there's some interesting things in
12:07 here
12:08 um yes they they need an Ethics
12:10 Committee they need a Safety Committee
12:12 they need a human alignment committee
12:15 um
12:16 yes they do they need a regulatory
12:18 committee
12:20 so so
12:21 and I mean listen the good news the good
12:24 news is this the
12:27 companies that are big enough to have
12:29 like the big models open AI anthropic
12:33 inflection Google meta they're all big
12:38 enough to have
12:39 you know ethics boards and things like
12:41 that you might be cynical about whether
12:43 or not they're going to be ethical
12:44 that's a whole different conversation
12:47 what's what's a bit scarier right now is
12:50 because meta released their
12:52 foundational models out into the wild
12:55 there's now all these open source
12:57 projects that don't have any safety
12:59 guard rails at all
13:01 um so
13:05 the the good news right now is that
13:08 those models are the open source models
13:11 are not as powerful as gpt4 but they're
13:14 getting better all the time like you
13:16 know on a on a every three day basis
13:18 they're improving dramatically
13:20 so yeah pretty interesting stuff
13:23 um
13:26 Sarah Silverman is suing open Ai and
13:28 meta for copyright infringement like I
13:30 said the next 20 years are going to be a
13:33 never-ending stream of lawsuits at some
13:35 point someone's gonna make precedent one
13:37 way or the other
13:38 and you know
13:41 Sarah will win that lawsuit and after 38
13:43 appeals they'll be like okay you have to
13:45 shut down the AI
13:47 they won't at least there's already open
13:49 source like like at this point it's I I
13:52 think what these lawsuits ultimately
13:53 will do is sort of figure out where the
13:55 boundaries are and you know
14:00 similar to how streaming when streaming
14:02 services came out everyone freaked out
14:04 and said no one's going to get paid No
14:05 One's Gonna Get Paid there are all sorts
14:07 of lawsuits and then they came to some
14:10 relatively shitty compensation model for
14:14 musicians
14:15 um I think in the world of AI that
14:17 compensated compensation model gets even
14:20 shittier because we're entering a world
14:22 of infinite content creation right
14:26 and almost all of the
14:29 um
14:30 copyright monetization stuff is
14:33 happening on the usage side on the
14:35 output side
14:37 and if you're entering a world where
14:40 fan art may sound as good if not better
14:42 as the original artists work but it's
14:45 going to be derived from them
14:48 we might need to find compensation and
14:50 attribution models that happen on the
14:52 generation side not on the usage side
14:54 but we're not anywhere close to that
14:56 right now right so so we've we're gonna
14:59 have to go through four or five or ten
15:01 years of just everyone's suing everyone
15:04 and then at some point realizing well
15:08 we're not going to uninvent this [ __ ]
15:09 and uh
15:12 and and we've got to figure out a way to
15:14 to work forward so that artists can get
15:16 compensated and businesses can stay in
15:19 business so I don't have
15:21 any kind of a clue what uh
15:25 how it's gonna play out other than just
15:27 expect more and more lawsuits I'm a big
15:30 fan of Sarah Silverman and it'll be
15:32 interesting to see what happens I think
15:33 high profile lawsuits like that are
15:35 going to make a difference and they're
15:36 going to establish some sort of
15:38 precedent I think we'll see I'll we'll
15:40 see some on the copyright side as well
15:43 um at some point some famous person is
15:46 gonna
15:47 try to copyright something where AI was
15:51 used to generate a lot of it in the
15:52 copyright office is going to shut that
15:54 [ __ ] down and that's gonna I don't know
15:56 we're gonna see some interesting stuff
16:00 hello Albert how are you I wonder how
16:02 long it'll take for these lawsuits to
16:04 settle litigate like the the here's the
16:07 thing that I've learned about lawyers I
16:08 was I was uh I was part of a public
16:12 company my the agency I started in the
16:14 mid 90s we went public and when the.com
16:17 Bubble Burst and our stock price Tanked
16:20 I was part of four
16:23 class action lawsuits that named me
16:25 personally for
16:28 hundreds of millions of dollars
16:30 and
16:32 every single one of those
16:34 class action lawsuits settled for wait
16:38 for it
16:40 the exact amount of the lawyer bill
16:44 four out of four
16:46 class action lawsuits settled for the
16:49 amount of the lawyer bill
16:51 so lawyers don't give a [ __ ] who they
16:55 sue or what they do or if they win it's
16:57 nice if they win but even if they lose
16:59 they win right because those those
17:01 settle those they were frivolous
17:03 lawsuits
17:04 and the lawyers made a [ __ ] fortune
17:08 and that's that's what's the this is
17:11 just this is you know the lawyers are
17:15 out pre-ordering their Yachts right now
17:17 because of AI
17:19 so I think it's going to take a while
17:21 they have no incentive to make it to
17:23 make the lawsuits any less or any
17:25 shorter because the longer they are the
17:27 more more money they make
17:31 someone said.com is king.com gen
17:35 xers.com is the only way baby
17:40 main names if you can't spell.com get
17:44 off my goddamn lawn
17:50 how learn AI have to go to college no go
17:53 there right now
17:55 chat.openai.com go to prompts.chat start
17:58 learning so okay two things
18:02 there's an inflection point
18:05 before November 30th 2022 if you want to
18:08 get into machine learning and AI
18:10 probably good to go to college you need
18:12 to be a scientist a mathematician a
18:14 researcher something like that
18:16 post chat GPT
18:18 there's now I don't know I'm guessing
18:21 200 300 million people
18:23 who do not have engineering degrees that
18:27 are using these tools in their business
18:29 in their personal lives to start
18:31 businesses
18:33 you can learn all this stuff without
18:35 knowing anything about mathematics does
18:38 having some concept of technology help
18:40 absolutely because we're very early
18:41 these tools are very very sketchy and
18:45 janky and if you want to do any of the
18:47 advanced stuff you might have to do some
18:49 geeky stuff like fire up a Google collab
18:52 create a virtual machine and go make
18:55 yourself some Lang chain
18:57 or whatever
18:59 so it doesn't hurt you to have some
19:01 technical background but you don't need
19:03 it go to chat.openai.com go to
19:05 prompts.chat and that's why I have these
19:07 on screen for my entire lives
19:10 is go start playing right now start
19:13 getting AI literate start understanding
19:15 what these tools make possible because
19:17 they are insane
19:19 what does that mean billion context
19:21 window what it means is you could put
19:24 750 million words in a prompt
19:28 to put that in context
19:32 um
19:33 that's more words than an average human
19:37 will read in their lifetime
19:41 and it basically means you can put in
19:43 you could put in massive amounts of data
19:47 essentially into the
19:50 um
19:52 into your chat session right so right
19:54 now you've got a 3 000 word context
19:57 window so you know how when you're
19:58 chatting with chat CBT if you do it for
20:00 a long time sometimes it just forgets
20:02 what the hell you were talking about
20:03 you're like but I gave you my whole life
20:06 story up there and now you don't seem to
20:07 know who I am that's because the life
20:10 story slid out of that context window
20:12 it's only three thousand words
20:14 750 million words you can start you
20:18 could put in massive amounts of data and
20:21 still have massive amounts of data you
20:23 could you know do back and forth and it
20:24 would have sort of this incredibly long
20:27 memory
20:28 um
20:30 they're saying with the approach that
20:32 they came up with this is a bunch of
20:33 Microsoft researchers they could get it
20:35 into the trillions which basically means
20:38 like what what I saw a video on it today
20:40 and basically what it means is if they
20:42 get it into the trillions
20:44 you can put the entire internet
20:48 into essentially into active memory so
20:51 you could be chatting with the entire
20:53 internet
20:54 as you're sort of known knowledge
20:59 part of the memory it's it's insane like
21:01 so here's what it basically means it
21:03 don't worry about the numbers the
21:05 numbers don't mean anything what it
21:06 basically means is
21:09 if we're jumping from like right now at
21:13 poe.com you can play with the Claude
21:15 100K token model
21:17 if we're jumping from a hundred thousand
21:19 tokens to a billion tokens
21:24 the speed with which we get to AGI
21:27 starts to look more like 2024 than 2029.
21:31 AGI is artificial general intelligence
21:33 the point at which these machines
21:37 can reason become smarter than the
21:39 smartest human
21:41 starts to look like
21:43 a year year and a half
21:45 rather than
21:46 three to five
21:49 so that that's Insanity it's it's like
21:52 and then and then once that happens once
21:56 these machines get that smart then they
21:58 start to accelerate themselves and you
22:00 know and the the improvements keep
22:02 accelerating and this is why folks are
22:05 talking so much about alignment right
22:06 now and openai mentioned yesterday or
22:09 the day before I know it's the day
22:11 before
22:12 um that they're they're committing 20
22:13 percent of their compute resources to
22:16 Super alignment is what they're calling
22:18 it which basically just means let's make
22:20 sure these things have the same goals as
22:22 humans
22:24 so we're not at war with them because
22:26 they're going to get way smarter than us
22:27 I think a lot faster than any of us are
22:30 prepared for
22:31 yeah so crazy crazy so ethics covers a
22:36 broad spectrum of stuff exactly like I
22:38 don't know quite what you're talking
22:39 about there but for me it's about it's
22:41 about alignment the ethics thing I mean
22:44 ethics isn't you know some of the like
22:48 bias at least is in the eye of the
22:49 beholder different people have different
22:51 biases and they're gonna I think what
22:53 we'll have is we'll have different
22:54 models but the ethics and integrity and
22:56 Alignment thing is super important
22:59 depends on who's doing the AI training I
23:01 kind of think of it similar to dog
23:02 training that's probably not a bad
23:04 metaphor actually so I was on Discord
23:06 read the reason for Bing was the paint
23:08 wall issue
23:10 think they will bring it back yeah I do
23:12 think they'll bring it back so so here's
23:15 here's how I understand it
23:17 if you're even remotely geeky when when
23:21 you land on a a site like New York Times
23:24 or Fortune Magazine it was Fortune
23:27 Magazine was was the issue apparently
23:30 and it pops up that thing that says you
23:32 can't see this site there's kind of two
23:35 flavors of paywalls one of them if you
23:37 view the source code of the site
23:40 if they have a shitty paywall you can
23:42 actually go read the article you just go
23:44 look at the you just go look at the
23:45 source code of the site you know inspect
23:47 and then you can go read it and go copy
23:50 it
23:51 um then there's other paywalls that you
23:53 can't you've actually gotta have a a
23:55 token to to see the article
23:58 um I I think it was that the former kind
24:01 of paywall where you could just
24:02 essentially inspect
24:04 the uh the code of a site and that's
24:06 that's what it was pulling out
24:08 why it's been gone for it's been almost
24:11 a week now that that that feature hasn't
24:14 been there let's see if they brought it
24:15 back I kind of check it
24:17 once a day just to see if if they
24:19 brought it back
24:21 now wait let me reload let me reload a
24:26 chat GPT you got my money
24:33 nope it's not there I don't know why
24:36 it's taking so long I mean what's what's
24:38 weird is
24:39 um plug-ins still work and they're
24:41 connected to the internet and bing.com
24:44 is still connected to the internet like
24:47 that's still functional the chat there
24:48 is still functional so this was clearly
24:50 something about
24:52 open ai's implementation of browse with
24:55 Bing was doing something that obviously
24:58 broke something and it's more
24:59 complicated to fix than
25:02 I thought it would be but
25:04 I don't know
25:06 because you'd think
25:09 I I don't know maybe maybe they got to
25:12 build a whole white listing
25:14 kind of
25:15 application that if
25:17 if you know companies are on that list
25:20 then it it won't scrape it or something
25:22 like that I like I honestly have no idea
25:24 I don't know why it's taken so long
25:27 um.com is King yes applying large
25:30 language models side
25:32 oh applying large language model side
25:34 okay so then what I would do Special K
25:37 Special K 353 you're special
25:41 um
25:44 so what I would do is
25:47 um
25:48 I just start
25:50 I
25:52 I would get as AI literate as you can as
25:54 soon as you can so you know go to these
25:57 tools
25:58 um
25:59 understand the different image
26:01 generation models mid-journey stable
26:02 diffusion Adobe Firefly generative fill
26:06 within Photoshop you don't have to be
26:08 Masters at those but just at least
26:09 understand that world
26:12 understand you know
26:15 as deeply as you can what generative AI
26:19 tools like chat EBT make possible
26:22 um
26:23 you know understand that they can write
26:25 code and they can write copy and that
26:27 they can make tables and that they can
26:29 help you do research and that they
26:30 hallucinate and you know all that sort
26:32 of stuff so if you're if you're already
26:34 decently
26:35 AI literate like that great
26:38 the the next thing I would do is start
26:41 putting it out there right so you don't
26:43 need to start a tick tock Channel like I
26:45 did because that's [ __ ] insanity
26:49 but but you know post on LinkedIn hey I
26:52 just did this experiment with chat apt
26:55 um join meetups or join you know AI
26:58 communities put that on your LinkedIn
27:01 um if you can I work for this group and
27:03 we did this cool content generation tool
27:06 and the guy that runs the group said hey
27:09 do you mind if I give you the title of
27:10 Chief generative officer and I'm like
27:13 sure why not I don't care and he said
27:16 I'm pretty sure it's the first Chief
27:17 generative officer in the world I'm like
27:19 awesome I get to claim I had that first
27:22 um so whatever what does it mean it
27:24 doesn't [ __ ] mean anything but
27:27 when I put that there people started
27:29 reaching out to me so put yourself out
27:31 there
27:32 um talk about this stuff with your
27:34 friends and family even if they think
27:36 you're weird again just because we're
27:39 we're very quickly and it's I'm starting
27:41 to see it happening we're very quickly
27:43 moving into a mode where companies are
27:48 gonna say we gotta figure this out this
27:50 is urgent like someone's gonna get
27:52 religion in one of these companies and
27:54 it won't be religion like they
27:55 understand this stuff it's going to be
27:57 religion like holy [ __ ] we got to do
27:59 something who knows anything about this
28:01 and if they're surfing LinkedIn and they
28:03 see that you know about it they might
28:05 click on your link
28:06 um and then the other thing I would do
28:08 is as you start to
28:10 um get better and better at that you can
28:12 put yourself out there and do education
28:14 you could do workshops for companies
28:17 um and then from that maybe that turns
28:19 into a job
28:21 um
28:22 I would just be very very reticent of
28:24 any company that's like well we want you
28:27 to do AI but we're we're you know we're
28:29 taking a wait and see attitude with it
28:31 anyone who's not actively going
28:35 we're trying to disrupt ourselves here
28:37 this is this stuff is so profound it
28:40 puts everything we do at risk and we're
28:41 going to figure it out before we get you
28:43 know bowled over by it
28:46 um I would I would actively seek out
28:48 companies that that are on the more
28:50 aggressive side of dealing with this
28:51 stuff
28:52 will AI make humans obsolete if you go
28:55 listen to the uh
28:56 Lex Friedman interview with George hotz
28:59 George hot certainly thinks so he he
29:02 talks about the incremental value of
29:04 humans is is getting increasingly niche
29:07 as he put it
29:09 um
29:11 I don't know I I think that
29:14 um
29:19 where I'm hoping it lands is that
29:23 these machines can do a lot of the
29:25 low-level repetitive
29:29 um
29:30 non
29:31 like the you know the soul-crushing part
29:34 of your day like depending on your job
29:35 it might be like a lot of the day
29:38 but the part of your job that's just
29:39 like oh I gotta go write those 10 emails
29:43 or I gotta go deal with the insurance
29:46 company or you know I just write you
29:49 know
29:50 iterations of the same copy that someone
29:52 else came up with and I'm like just a
29:54 cog in a machine
29:56 I think that these AI tools do that [ __ ]
29:59 really well
30:00 so my hope is that all of the
30:04 complicated repetitive low-level
30:07 soul-crushing work of the world
30:10 will be done by these things and then
30:13 that frees us as humans to either
30:15 operate at a higher level
30:17 or work Less hours or do less for the
30:22 same amount of productivity output so so
30:25 we see it we see a gap where
30:27 what's required to do the work is lower
30:30 than than doing the work
30:33 what fills that Gap I don't [ __ ] know
30:37 AI is going to generate a bunch of new
30:39 jobs so there might be new jobs that
30:40 fill that Gap
30:42 I Now understand why Andrew Yang was
30:45 talking about Universal basic income
30:47 because that Gap could look like there's
30:48 a bunch of unemployed people and you
30:50 might need to fill that Gap with some
30:52 Universal basic income until the new
30:53 jobs come
30:55 but that Gap could also look like
30:59 now there's a bunch of people that were
31:01 doing kind of low level Cog in the
31:04 machine work that are now going to say
31:06 you know what I've always wanted to do
31:08 write the Great American novel start my
31:11 own company start a flower shop you know
31:13 I think I might like to be a plumber
31:15 that doesn't [ __ ] matter so I think
31:17 we're going to see a lot of
31:19 personal reinvention and a lot of
31:21 innovation
31:23 but I think it's also going to be
31:25 painful I think as as the AI
31:30 drives up productivity and capability
31:33 and we don't need as many people I think
31:35 that Gap is going to look like a lot of
31:38 displaced people for the next
31:40 two years three years maybe
31:42 there's a new show on Netflix called
31:44 unknown Killer Robots what happens when
31:47 machines make life or
31:51 the documentary will explore the dangers
31:53 of artificial intelligence in military
31:55 application yep
31:56 yeah
31:58 like like what Boston Dynamics has said
32:00 is you know if if you build robots just
32:03 rule number one don't put guns on them
32:11 you know the military's [ __ ] putting
32:12 guns on them you know that right you
32:14 know that right
32:16 so
32:20 this gets back to the alignment thing
32:22 right super alignment you know
32:25 are the goals of these things aligned
32:27 with human rights
32:30 and if they are good
32:33 but how do you do that
32:35 and when you have a world where you've
32:37 got open source large language models
32:38 which we do now and they don't have any
32:41 alignment and they don't have any
32:42 corporate overlords and they don't have
32:43 any regulation
32:45 what happens then
32:47 do you make any AI illegal unless it's
32:50 through one of the regulated companies I
32:52 don't think that's going to fly
32:54 because that's not going to happen
32:55 globally
32:57 but it could happen state by state it
32:59 could happen even Nation by Nation but
33:02 then your neighbors just gonna
33:04 do stuff and they're going to innovate I
33:07 don't know
33:08 it's wild
33:10 what's that new it's called unknown
33:11 Killer Robots see
33:13 that's the other thing right now is
33:17 the fear-mongering with this stuff is is
33:19 kind of stupid we've had 50 years of
33:21 Hollywood telling us the robots are
33:23 going to kill us so I think there's a
33:24 predisposition for people to um
33:29 fear them because change is hard anyway
33:32 and when you've got things that are this
33:34 profoundly powerful
33:38 that
33:39 the answer to if I'm a copywriter
33:42 like is my job the same as it was before
33:44 it's not and it never will be when you
33:47 have things that that produce kind of
33:49 instant existential questions for
33:52 people's livelihood
33:54 that sort of compounds the fear of
33:56 Hollywood telling us the robots are
33:58 going to kill us so I like I hate the
34:00 fact that they're calling it unknown
34:02 Killer Robots but whatever
34:06 that's why that's why I do this channel
34:07 because I think it's just too easy to
34:09 give into that fear it's really
34:10 sensationalistic I think it is it is as
34:13 likely that these things profoundly
34:15 improve life
34:17 as as much as they profoundly destroy
34:19 life but we'll see
34:22 it will certainly outlive URI well it
34:25 might so will our children and
34:27 grandchildren
34:29 um
34:33 it won't
34:35 if um
34:37 the internet and electricity doesn't
34:39 exist anymore if the infrastructure goes
34:41 bye-bye then AI doesn't survive it
34:45 although
34:48 to that point well no I mean you need
34:51 functional computers for this stuff to
34:53 work so so if they drop some uh
34:56 electromagnetic pulse weapons then then
34:59 you know that's a that's that's the off
35:01 button for this thing is computer seats
35:03 working but you know then then we're
35:05 then we're into the dystopian future
35:07 that we all saw in the movies
35:10 and I hope you're good at gardening and
35:12 defending yourself
35:13 [Laughter]
35:16 um I want we learn
35:19 we learned for my real estate company
35:24 you mean AI learning for your real
35:26 estate company what do you mean when
35:28 they say chat GPT hallucinates Okay so
35:32 I also like to describe it as
35:34 mansplaining as a service
35:37 so when you ask chat GPT
35:41 um
35:42 to
35:43 give you an answer it will give you an
35:45 answer
35:46 and then if you're like I'm pretty sure
35:48 that answer is wrong it's like oh I'm
35:50 sorry here's the right answer and then
35:53 that answer might be wrong and then
35:54 you're like no that's still wrong it's
35:56 like that's my apologies this is
35:58 embarrassing here's the real answer and
36:01 then that answer is wrong that's
36:02 hallucinating
36:05 the way the technology works
36:07 is you know when you're in Google and
36:10 you're typing a sentence and like the
36:12 the last three words of the sentence the
36:14 autocomplete shows up
36:17 that's all these tools are except it's
36:20 like that on steroids right so it writes
36:23 you A big one dissertation of you know
36:26 some research that you asked it to do
36:27 and then it provides citations of here's
36:30 the links to the articles that I used to
36:33 come to this opinion and then you click
36:35 on those links and they're all dead
36:36 links
36:38 when it's writing all of these words
36:40 it's doing them one word at a time just
36:42 as a probability engine it doesn't
36:44 actually go back and look at what it
36:46 wrote and go is that correct now there's
36:50 a lot of companies working on that piece
36:53 of it but it's not there right now so
36:56 just the nature of how it works it is
36:58 just
36:59 this it's
37:01 it's got the knowledge of what it does
37:04 as much as a calculator has knowledge of
37:06 what it does right we just trust that
37:08 calculators have the right formulas in
37:11 them when we put in two plus two it
37:13 seems to come out consistently of four
37:15 if calculators when you put in two plus
37:18 two came out with some random number we
37:21 wouldn't trust them as much right so
37:23 that's where these things are right now
37:24 is it's like we've built the fancy word
37:26 calculator that two plus two is most of
37:30 the times four but occasionally not
37:33 and it's up to you to deal with that now
37:39 there are times when that's completely
37:41 irrelevant right if I'm working on a
37:43 screenplay or some creative writing or
37:46 writing lyrics for a song or social
37:48 media copy whatever it is
37:51 I might not need that but if I'm doing a
37:54 blog post that makes a claim and that
37:57 claim has got to be backed up by some
37:59 research and that research is in there
38:00 well that's then on you to make sure
38:02 that that research is accurate there was
38:05 about two weeks ago there was a lawyer
38:07 that had chat EPT write a legal opinion
38:10 for him and he handed it to a judge
38:13 and all of this the the case law that
38:16 was cited in his opinion
38:18 didn't exist
38:22 and the judge called him out on it and
38:24 said what the [ __ ] is this and the
38:26 lawyer was like I wrote it in chat GPT
38:30 that's on the lawyer in my opinion you
38:33 know these things do this right now like
38:35 you know a year or two from now we won't
38:37 have to worry about this but right now
38:39 these things are they hallucinate
38:43 um
38:44 so if you're a lawyer
38:46 check your work
38:47 [Laughter]
38:49 all right is there an easy way to avoid
38:51 plagiarism of chat CPT writes an article
38:53 for me besides rewriting it all so it's
38:55 not plagiarism like I said the way it
38:58 works
38:59 is that
39:01 when you write your prompt it generates
39:04 essentially a mathematical probability
39:07 that it then looks at all of the words
39:10 it knows and all of the words it knows
39:12 is a lot of them it's so so the the core
39:15 large language model sitting underneath
39:18 um Chachi BT
39:20 it was trained on all of the public
39:21 internet
39:23 and so what it's not doing is it's not
39:26 copying and pasting from specific pieces
39:29 of work and pasting that into your
39:31 document
39:32 what it is doing is saying based on the
39:34 prompt that you wrote the mathematical
39:36 probability that the next word is the
39:40 is 99.999 so it'll type the word the and
39:45 then it takes what you wrote and the
39:46 word the and it creates the probability
39:48 for the next word and it just does that
39:50 over and over and over again so it's
39:51 looking at all of the stuff it knows and
39:54 whatever has been kind of filtered out
39:56 depending on your prompt right so that's
39:58 why if you tell chat EBT I want you to
40:01 act like a lawyer that went to Harvard
40:02 Law School you get better answers than
40:05 if you just say give me a legal opinion
40:07 because they give me a legal opinion
40:09 looks at all of the internet when you
40:12 say I want you to ask like a Harvard law
40:14 professor it's going to look at you know
40:16 the the words that are kind of in that
40:19 semantic
40:20 cloud
40:22 um so so it'll get it'll get better
40:25 um so it's not plagiarizing
40:28 what you might be talking about is
40:30 can it be detected that it was written
40:33 by AI
40:35 that's where you might want to have a
40:37 rewriter rewrite it if you're if you're
40:38 at a university where they're using AI
40:42 detectors you go find a Ai rewriter and
40:46 it will rewrite it in a way and
40:47 apparently
40:48 gpt4 uh doesn't get caught by the
40:52 detectors a lot
40:54 um I think this whole idea of detecting
40:56 whether something is a AI or not is
40:58 going to be very short-lived because
40:59 these tools are going to be incorporated
41:01 in into like Microsoft Word Google Docs
41:05 note taking apps it's going to be in
41:07 everything and and
41:08 the the way that when you start working
41:11 with this stuff on a regular basis it's
41:14 very rare that I'll take something out
41:16 of this machine and use that wholesale
41:18 what I'll often do is I'll say
41:20 um I kind of have an idea for a piece or
41:22 something I need to do and I'll use that
41:24 I'm like give me a first draft get me to
41:26 80 and then I kind of take it from there
41:28 and then if I get stuck in there I'll
41:30 say oh let me take this paragraph and
41:32 have it rewrite it or give me an example
41:33 and it'll do that and I'll put it back
41:35 in so almost all of what I write is some
41:37 sort of hybrid of
41:40 Ai and me and I think increasingly
41:43 that's how it's going to be so I don't
41:44 think so so a I don't think you need to
41:46 worry about the plagiarism piece if
41:48 you're talking about AI detectors that
41:51 that you'll need to do some research on
41:53 depending on the organization you work
41:55 for how does AI learn to use it for my
41:57 real estate comp you mean compensation
42:01 or computer
42:04 if it's compensation so I would so go to
42:07 prompts.chat go to that thing
42:09 read about prompting and how prompting
42:12 works and then look at so what that
42:14 document is is it's a whole list of act
42:16 like a so you could say act like a real
42:19 estate you know specialist that
42:22 understands real estate compensation or
42:24 I I don't know what comp means I assume
42:27 it means compensation but if you mean
42:29 computer basically tell it to act like a
42:31 real estate expert and then start asking
42:34 it questions for what you need
42:36 there's also if you pay for
42:39 chat GPT it's 20 bucks a month for the
42:43 plus version you get this new thing
42:45 called code interpreter
42:48 which holy crap
42:52 code interpreters insane you can upload
42:55 up to 100 megabyte file
42:58 to code interpreter and then start
43:00 interacting with it so you could put up
43:03 a hundred megabytes worth of real estate
43:05 data
43:06 upload it to this thing and start
43:08 interacting with it I don't know what
43:09 you're looking for but like find me oh
43:12 con I assume you maybe you mean comps
43:14 comparisons yeah so you could say
43:17 take data of here's all the houses on
43:19 the market find me any comps in
43:22 neighborhoods that are similar to this
43:23 one and it will go find them
43:26 um so check out code interpreter I can
43:28 show it to you here in a bit
43:30 remember VCRs no one was going to be
43:32 allowed to record television shows
43:34 lawyers yeah baby yeah exactly exactly
43:36 it's this is
43:38 the thing that's amazing about every
43:40 time these new technologies hit the
43:43 market it's
43:44 identical every time
43:48 the established professions are like
43:51 this is awful it's going to replace
43:53 everyone everyone's gonna die
43:56 and then
43:58 and then the lawyers are like we'll sue
44:00 them we'll block it we'll stop
44:02 innovation
44:03 well uninvent it
44:06 and it never happens it's the same every
44:09 single time now
44:11 the speed with which this is coming and
44:14 the depth of impact of this is is is
44:17 going to eclipse probably anything we've
44:21 ever seen shy of things like fire
44:25 printing press
44:27 um steam engine
44:30 it's big like that
44:32 um so we ain't stopping it
44:34 but it's going to be messy for a while
44:37 because just fear and disruption and
44:39 and things like that so why this channel
44:42 exists I want everyone here to get
44:43 curious about this stuff if you haven't
44:45 played with it yet if you're sitting on
44:47 the sidelines like well I heard it's not
44:48 all that accurate or yeah you know the
44:50 robots are gonna kill us I'll just give
44:52 it three years and see what happens
44:53 don't do that don't be that person
44:56 if you go learn this stuff and then
44:59 decide I hate it and I'm gonna disengage
45:01 from it more power to you
45:04 but don't just sit on the sidelines and
45:06 assume that you've got time because I
45:08 don't think I don't think we do we don't
45:10 have time it's coming very very fast
45:13 very very fast
45:15 um and I don't know that we're prepared
45:17 for it
45:20 remember microwaves I could have been
45:22 rich I know I've got I got a bunch of
45:24 those if you're on the planet long
45:25 enough you know you've got one of those
45:27 Oh I thought of that three years before
45:29 they did
45:30 yeah here we are
45:32 I mean that's the other scary thing
45:34 about this technology is
45:36 it's going to be trivial to
45:40 um replicate a piece of software trivial
45:43 so someone comes up with something novel
45:45 and interesting
45:47 anyone's gonna be able to go give me an
45:49 app that does that oh and by the way
45:51 make it look like that too
45:53 and come up with an intro
45:54 Ninja for it and change change
45:56 everything so the words are different
45:57 but it does exactly the same thing
46:00 so
46:03 that's going to be happening as well are
46:04 we going to get sued for writing books
46:06 with the help of AI I don't think you're
46:08 gonna get sued for it I think if you
46:10 want to copyright it you're going to
46:12 need to prove that your contribution to
46:14 it was significant enough that it's
46:15 copyrightable
46:17 um
46:19 but that's even up in the air
46:21 um
46:22 Sarah Silverman just brought a
46:25 multi-million dollar lawsuit to open AI
46:27 saying they trained
46:29 um chat gbt on
46:32 um her written material and right so if
46:35 she wins that then everyone who's ever
46:37 written anything is going to do the same
46:39 kind of lawsuit so I don't think it's
46:41 gonna
46:44 it this is gonna be this is gonna be not
46:47 dissimilar to what happened in the 90s
46:49 with sampling music and Hip-Hop
46:52 but the difference here is
46:54 in the 90s when you're sampling music
46:56 you were you were literally copying and
46:59 pasting
47:00 the Beats from someone else's song and
47:02 putting that in your song
47:04 this generative AI stuff is not doing
47:06 that it's it's
47:08 the the metaphor is closer to
47:12 this machine went to college and it
47:13 learned all the stuff
47:15 so if you went to college and you
47:17 learned all about
47:19 a particular kind of writer right
47:23 and then you went out into the world and
47:24 you started writing and you were so
47:25 inspired by that writer you kind of
47:27 wrote like that writer well that's still
47:29 your writing it's just you're taking
47:30 your knowledge of all the stuff you
47:32 learned
47:33 and the filter that happens to sound
47:36 right to you is writing that sounds like
47:38 they're writing that's not their writing
47:42 um
47:43 so so it's more like that
47:46 um but
47:48 copyright holders are still going to sue
47:50 they're going to say Hey you trained on
47:52 my stuff you're making money on that
47:56 but then I would go to what percentage
47:58 of the words on all of the public
48:00 internet are Sarah Silverman's it's
48:03 probably a very small percentage and
48:04 then what percentage of that ended up in
48:08 the thing that you wrote where you said
48:09 write me a joke about something
48:12 it's not copying and pasting her joke
48:14 and putting it in there if it were doing
48:16 that then I'd say there's a much bigger
48:17 issue
48:18 but she's just part of the contributing
48:22 Corpus of knowledge
48:24 that when you say write me a joke it's
48:26 probably pulling as much from Jim
48:28 Gaffigan as it is from Sarah Silverman
48:31 she would probably be offended by that
48:32 which is cool Taylor Swift is going to
48:35 have a lot of fodder for songs AI
48:37 plagiarism
48:38 yeah
48:40 well and there's so so there's really
48:43 interesting stuff there's uh Elon musk's
48:46 ex-wife or girlfriend
48:50 um
48:52 um she said she said sample my voice all
48:56 you want just give me 50 percent
48:58 so I think some people are gonna gonna
49:00 go the route of I'm gonna sue your ass
49:02 and other people are gonna go the route
49:04 of we're not stopping this thing so just
49:07 use it and compensate me fairly and I
49:10 think in the end that's where we'll end
49:12 up is people create models to say okay
49:15 here's the official Taylor Swift
49:18 model where if you want to create songs
49:20 that sound like Taylor Swift
49:22 you click on this link and then every
49:25 song that you generate Taylor gets a
49:27 nickel and if your song You Know starts
49:30 to make money on streaming services then
49:32 Taylor gets 50 or some [ __ ] like that so
49:35 those those systems will be created and
49:37 by the way one of the things I've been
49:39 doing on these lives is talking about
49:40 okay here's a business idea so there's a
49:42 business idea
49:43 create
49:45 create a thing that allows artists to
49:47 opt in to create their officially
49:49 licensed model to let other people
49:52 generate on that
49:54 and you know build that system make it
49:56 really easy for artists to do that I'm
49:58 sure there's competition for that but
50:00 that's that's going to be a massive
50:01 [ __ ] business
50:03 because because there's so so the
50:05 current copyright thing is after
50:08 something is published right depending
50:11 on the usage that's where we're going to
50:13 charge you money right after so
50:15 everyone's out here arguing about the
50:17 after you took my copywritten material
50:20 you trained your model and more shit's
50:23 getting out in the world and they're
50:24 they're all worried about usage and and
50:26 after the fact
50:28 I think there's money to be made on the
50:29 front end of this
50:31 let me take all those songs put them
50:33 into an official model so if Disney
50:35 let's Disney does this right
50:38 they're not doing it now but they they
50:39 could do this and and I would be
50:41 surprised if they didn't
50:44 you want something that looks like
50:45 Mickey Mouse in your
50:47 whatever the hell you want to do
50:50 wouldn't it be amazing to be able to
50:52 click on the officially licensed Disney
50:54 model
50:55 that says you can generate from this and
50:57 if you use it
50:59 here's here's what it costs you
51:02 so you can do original work that's
51:04 officially Disney licensed maybe they
51:06 have to approve it whatever
51:09 left you brain I don't know what that
51:11 means left you brain
51:13 you guys are weird with the English
51:15 tonight it's it's like I see words I
51:18 recognize but when I put them together
51:20 they make no sense
51:24 uh
51:25 uh welcome everybody if you're new it is
51:27 here urls
51:29 why are they here because if you haven't
51:31 played with chat gbt that first URL is
51:34 chat GPT go there go there right now
51:37 start playing while you hear me ramble
51:40 or just leave
51:41 uh and then bing.com is also chat gbt if
51:45 you click on the chat button at bing.com
51:47 that's chatgpt4 connected to the
51:49 internet and then poe.com's got six
51:52 different chat gpt-like things you can
51:53 play with plus a lot of bots post kind
51:56 of fun there's a lot to play with there
51:58 um
51:59 just go start playing go start
52:01 understanding what these things are all
52:03 about and then prompts.chat
52:05 if if you're intimidated by like the
52:07 just the blank box like what do I do
52:09 here prompts.chat will teach you about
52:11 prompting and interacting with these
52:13 things
52:14 um and getting some interesting results
52:16 out of them so
52:18 that's what this channel is all about
52:22 and if you've got
52:24 um questions about AI pop them in the
52:26 comments I do my best to answer all of
52:29 them
52:30 um why does Poe why does it saypo.com
52:33 when it's an app because it's also a
52:35 website you can go to poe.com but you
52:38 can also get the PO app but they're
52:39 they're basically the same thing so
52:42 if we go to poe.com
52:49 so there's poe.com
52:53 and so you've got you've got your
52:54 different models
52:56 you've got your different models here
52:58 and then you have a bunch of bots and
52:59 you can explore Bots you can create a
53:01 bot
53:04 come on focus focus focus
53:08 all right anyway
53:10 so that's why
53:22 do you listen to AI oh to the all-in
53:25 podcast with Jason kalkanus yes I do
53:29 I know Jason from the Silicon alley days
53:31 the early days of the internet when he
53:33 did Silicon Valley Reporter
53:38 I was featured in that a couple of times
53:39 Bing is terrible
53:41 um Bing is actually getting better but
53:44 it is not anywhere near as good
53:46 as uh
53:48 as GPT as as gpt4 at chat.openai.com
53:53 threads threads hit 100 million
53:56 hit 100 million users in five days so so
53:59 chat GPT is no longer the fastest
54:01 adoption of technology in history but
54:04 I will argue that it's cheating to start
54:06 with a billion person user base of
54:09 Instagram and just say
54:11 click over here to our new Twitter
54:13 competitor
54:14 um it's okay like like it's missing a
54:18 lot of the features like I want to be
54:19 able to
54:20 see things that that people put like I
54:23 want to be able to have followers and
54:24 all that sort of [ __ ] and I I don't know
54:26 it's just it seems a little basic right
54:28 now people seem to be enjoying it I'm
54:30 I'm exhausted like I you know I'm
54:34 I got a [ __ ] Tick Tock channel so
54:38 it's a lot it's a lot just to have a
54:41 whole nother competing
54:43 place to be clever
54:48 so I don't know I'll use threads when
54:50 when uh if if musk flushes Twitter down
54:54 the toilet and threads takes over I'll
54:56 move there I've got an account there I
54:58 did one post I put up a mid-journey
55:00 image
55:03 you can go see me I'm at Kyle Shannon
55:05 over at uh at the threads
55:09 Special K of school cereal
55:12 so why pay for po okay here's why pay
55:14 for Poe this is why I pay for Poe you
55:17 can choose to get that
55:19 right here so so
55:22 when you go to Poe there's six different
55:24 models that you can play with sage
55:26 Claude instant and chat EPT you can use
55:28 for free
55:29 and then there's these three that say
55:31 subscription access gpt4 Cloud plus and
55:34 Claude instant 100K gpt4 you know quad
55:38 plus is on anthropics version of gpt4
55:41 it's not as good but it's close
55:45 but this Claude instant 100K that means
55:47 it's a hundred thousand token context
55:50 window which basically means you can you
55:53 can have a 70 you've got a 75 000 word
55:56 context window you could put in a sixty
55:59 thousand word novel and and have it
56:02 write you an epilogue you can start
56:04 interacting with
56:06 um high level or like like high quantity
56:09 um kind of data where chat GPT is
56:13 limited to four or eight thousand tokens
56:16 which is about three quarters of that is
56:18 words
56:20 um so that's why I pay for it is for
56:22 that but now that chat GPT has this
56:29 code interpreter
56:32 I can now upload a hundred megabyte file
56:36 and start interacting with it so Poe all
56:39 of a sudden is a lot less interesting to
56:42 me because code interpreter is a big big
56:46 deal so if I were gonna pay 20 bucks a
56:49 month I'd pay it to open AI
56:56 ah
56:58 [Laughter]
57:05 best AI tool to assist with short form
57:08 content for tick tock
57:14 go learn how to prompt it now listen you
57:16 can go to futurepedia.io there are tools
57:19 like Jasper and copy Ai and any Myriad
57:23 of tools that are going to do tick tock
57:26 scripts for you easily
57:29 you can go pay them they're gonna
57:31 they're gonna charge you whatever
57:32 they're gonna charge you 20 30 40 bucks
57:33 a month
57:35 um
57:38 what what almost all of those services
57:41 are is someone has written The Prompt
57:44 for you
57:45 and I would say we're early enough and
57:47 these tools are raw enough and simple
57:49 enough that it it would it would serve
57:52 you to go learn how to do
57:55 you know have chat GPT write short form
57:58 content for you gpt4
58:01 is staggeringly powerful for [ __ ] like
58:04 that
58:05 like staggeringly powerful even 3.5 is
58:08 really good but four four is very very
58:10 good and so so go to prompts.chat
58:14 you know read the sort of primers in
58:16 there and some of the examples and
58:17 you'll start to get a sense of it and
58:19 then also just start to hang out like go
58:22 to go to the chat GPT subreddits
58:26 um start following people on Twitter
58:28 that
58:29 you know are talking about prompting and
58:31 prompt crafting there's another
58:34 um
58:35 Tick Tock maker here named Rachel Woods
58:37 she's got uh a newsletter called the AI
58:40 exchange she's really good at talking
58:42 about this stuff start following her
58:45 um and I would say just get good at it
58:47 because it
58:48 in a very short amount of time
58:52 you'll do a prompt and say make me a
58:54 tick tock script and you'll give it
58:55 enough context and it'll do something
58:57 and you're going to be you're going to
58:58 do one of these
59:01 I didn't I could do that
59:03 I made it to that yeah
59:06 so go go do that Nicholas
59:09 go teach yourself yeah open AI Chief
59:12 generative officer that's me man first
59:14 one
59:15 I think
59:16 not that it matters which certificate
59:18 would you recommend for generative AI
59:21 none
59:22 I you know the the certificate of Hard
59:25 Knocks like go learn this [ __ ] learn the
59:27 concepts below it
59:29 um I don't think certificates matter
59:30 right now
59:32 um if you want to if you wanna
59:34 play more on the
59:36 um on the backhand side then then
59:38 they're more important now
59:41 um Google just came out with one and was
59:44 it Facebook or LinkedIn
59:46 a couple of people just came out with
59:48 generative AI I don't know if they're
59:50 certificates but they're they're courses
59:53 um if you just want to get your head
59:54 around what these things are how they
59:56 work things like that you you can go do
59:58 those
59:59 um you can also go watch a bunch of the
1:00:00 videos on my channel like a lot of what
1:00:02 my early content is is here's what these
1:00:06 things are but there's a lot of content
1:00:08 out there but I think what's most
1:00:09 important right now is start using them
1:00:11 go there and start playing get curious
1:00:13 get exploratory think of it as an
1:00:16 adventure start learning it start
1:00:18 learning it start learning it
1:00:21 all right
1:00:24 um I identified another secret cyborg at
1:00:26 work today that's freaking awesome
1:00:28 so so what she's referring to there is
1:00:31 there's a Wharton Professor named Ethan
1:00:33 Malik and he's like my dog is whining
1:00:36 Hannah
1:00:38 what do you want
1:01:06 so so Ethan Malik is this Wharton
1:01:09 Professor that um
1:01:12 he wrote this wrote this piece a week or
1:01:15 two ago called detecting the secret
1:01:16 cyborgs and the secret cyborgs are
1:01:19 employees and organizations who are
1:01:21 getting AI literate but not telling the
1:01:23 organization about it because a lot of
1:01:26 the organizations some of them have you
1:01:28 you can't use chat GPT at all policies
1:01:31 some of them are like you know we'll
1:01:33 fire you if you use that some of them
1:01:35 are like wow we're taking a wait wait
1:01:37 and see attitude and and then also some
1:01:40 of the employees are like
1:01:42 I have dramatically increased my
1:01:44 personal productivity here and I don't
1:01:45 necessarily want my boss to know about
1:01:47 it because they're just gonna expect me
1:01:49 to do more so so you kind of have this
1:01:51 this this invisible class of AI literate
1:01:55 employees in companies that aren't
1:01:57 dealing with this
1:01:59 um and and he's talking about they need
1:02:00 to find a way to do that so it sounds
1:02:02 like cabruno found a secret cyborg
1:02:04 that's awesome
1:02:07 and that's you know that's again why I
1:02:09 think that I think that it's valuable
1:02:11 for you in an organization if you are a
1:02:14 secret cyborg if you are
1:02:16 AI literate talk about it
1:02:18 talk about it if you're working for a
1:02:20 place that's gonna fire you for being AI
1:02:22 literate and doesn't want to have
1:02:25 an interesting dialogue about it and say
1:02:27 tell us what you know and you know and
1:02:31 what you think about this stuff then
1:02:33 they're going to get blindsided at some
1:02:35 point you probably don't want to be
1:02:36 there anyway
1:02:37 which accredited institution would you
1:02:39 recommend or learn to get a
1:02:40 certification
1:02:43 um again right now for the generative
1:02:44 stuff like chat GPT I wouldn't
1:02:48 um if if you want to get into the
1:02:50 machine learning side of things and like
1:02:51 go work for open AI or anthropic or one
1:02:53 of these companies
1:02:55 then I don't know
1:02:57 I have a degree in acting so you're
1:02:59 asking the wrong guy Stanford MIT you
1:03:02 know start at the top and then work down
1:03:06 I mean not for nothing Stanford and MIT
1:03:09 they have free courses like all of the
1:03:11 universities have free courses to teach
1:03:13 you all of that [ __ ]
1:03:15 Google's got courses
1:03:17 um I'm sure meta's got them
1:03:20 Stanford MIT
1:03:24 it's all it's all out there for free so
1:03:26 just go do that or
1:03:29 if you don't have the discipline like I
1:03:31 don't like I I can't I couldn't
1:03:33 self-educate like that because I just
1:03:35 I've got add and I'm like I'd rather
1:03:37 just learn it myself
1:03:39 um
1:03:42 yeah I don't know I don't know
1:03:46 right now for for me it's more important
1:03:49 that this stuff is moving the reason
1:03:51 that I'm saying
1:03:53 any sort of certificate in chat GPT or
1:03:55 prompting or anything like that
1:03:58 well it's going to give you a leg up
1:04:00 right now
1:04:02 by definition anything that they're
1:04:03 teaching you is going to be obsolete by
1:04:06 the time they're teaching it because
1:04:08 this stuff is moving fast
1:04:10 like it's moving like like code
1:04:13 interpreter this thing that just came
1:04:15 out
1:04:16 it's a fundamentally different way of
1:04:18 interacting with chat gbt like I don't
1:04:20 have my head around it yet at all like I
1:04:22 know what it does
1:04:24 I know that you can upload a file and
1:04:27 then you ask it to do [ __ ] and it starts
1:04:29 writing python code and executing it to
1:04:32 do the thing that you wanted including
1:04:34 like making you charts and graphs of the
1:04:37 analysis of the data that's in the file
1:04:39 that you uploaded
1:04:41 okay
1:04:42 that's what it does but like what do you
1:04:44 actually use it for and how would you
1:04:46 actually use it in business
1:04:47 and how would I as someone who's
1:04:50 interested in marketing and copying this
1:04:52 and that like how can it be valuable to
1:04:54 me I don't know yet I'm still trying to
1:04:56 get my head around it so so that's the
1:04:58 the problem if you go learn prompting
1:05:00 you're going to be be essentially
1:05:01 learning three to six month old stuff
1:05:05 because it takes a while to get a course
1:05:07 together and like people have even asked
1:05:09 me on here to do a workshop and I'm a
1:05:11 little reticent because
1:05:13 I I don't know what I'm gonna teach that
1:05:16 isn't going to be obsolete so I mean
1:05:19 there's there's some Basics and Basics
1:05:20 are fine
1:05:22 but but I don't think it I don't think
1:05:24 it's
1:05:26 I ain't the amount of time it would take
1:05:27 you to find a course take the course you
1:05:29 could learn more on your own buy a lot
1:05:32 Chief disruption officer just applied
1:05:34 for the new position love it
1:05:38 I love using AI for emails yeah it's so
1:05:40 good for [ __ ] like that all that low
1:05:41 level crap what about Microsoft so so
1:05:43 listen well that's what bing.com is
1:05:45 Microsoft
1:05:47 um Microsoft's invested
1:05:51 um 10 billion dollars they've committed
1:05:53 10 billion dollars to open AI so they're
1:05:55 pretty invested
1:05:58 they've already created GitHub co-pilot
1:06:01 they're rolling co-pilot into all of
1:06:03 Office 365 they're rolling it into
1:06:05 windows they're rolling it into all of
1:06:07 their Azure Services they're rolling it
1:06:09 into Dynamics and it is the primary
1:06:14 um
1:06:15 Search tool at bing.com is now chat GPT
1:06:19 so they're all in and and I mean what's
1:06:22 I've been a Microsoft Basher for decades
1:06:25 I hated them
1:06:27 and
1:06:29 what they've done in the past six months
1:06:31 has been remarkable like the how they're
1:06:32 incorporating this stuff and how they
1:06:34 are
1:06:35 [ __ ] with Google
1:06:37 is is just entertaining
1:06:39 they're like gleefully [ __ ] with
1:06:42 Google because they were just they were
1:06:44 like gum on Google shoe for you know
1:06:47 what a decade and a half
1:06:49 in the search game they just couldn't
1:06:51 make inroads and like you know they've
1:06:53 got it built into the [ __ ] operating
1:06:55 system and they couldn't compete with
1:06:57 Google and Along Comes Lilo chat GPT
1:07:02 and it changed everything and so they're
1:07:05 all in so I'm I am I'm impressed with
1:07:09 what they've been doing
1:07:10 I'm impressed with Microsoft
1:07:13 um might they [ __ ] it up sure they're
1:07:15 big company
1:07:16 they can drive it off a cliff like
1:07:18 anyone but but I would say right now
1:07:20 they're looking they're looking in a
1:07:22 stronger position than Google currently
1:07:26 wouldn't count Google out but they're
1:07:28 definitely on their heels they so here's
1:07:30 the thing about Google they [ __ ]
1:07:32 invented the T in GPT is Transformer
1:07:36 Google invented it and then sat on it
1:07:39 for seven years
1:07:40 because I think they knew that if they
1:07:42 released something like Chach EBT it
1:07:44 would compete with their core product
1:07:46 and it is
1:07:48 and now they're trying to catch up
1:07:52 what comments is he reading I'm very far
1:07:55 behind new Eureka Queen okay awesome
1:07:58 name I'm the next New Yorker I'm living
1:08:00 in Denver now but I miss New York I miss
1:08:02 my New York my new eurekans
1:08:06 um I'm just way more far behind like
1:08:09 here's the deal this is Elder talk I'm
1:08:11 an old guy using Tick Tock so there will
1:08:13 be no twerking and um
1:08:16 I actually try to answer all the
1:08:18 comments so I am very far behind on
1:08:20 comments so that's why it seems like I'm
1:08:22 not actually answering answering them
1:08:23 but if you're still here new eurekin
1:08:27 know that I am
1:08:28 it sounds like AI is going to put the
1:08:30 marketplace through a brutal but
1:08:32 beautiful process that is a
1:08:35 beautiful way to put it brutal but
1:08:37 beautiful process yeah
1:08:41 the
1:08:48 every company and every person
1:08:51 every day
1:08:53 has the opportunity to reinvent what
1:08:56 they do
1:08:58 but they don't why
1:09:00 because change is hard and reinvention
1:09:03 sucks and it means you've got to give
1:09:05 something up and it means you've got to
1:09:06 flush some things down the toilet
1:09:09 and so I think that generative AI is
1:09:13 going to brutally
1:09:15 force reinvention of both people and
1:09:19 companies and it's going to do it
1:09:20 quickly
1:09:22 and
1:09:24 the tools are so powerful
1:09:27 that the ability to reinvent yourself is
1:09:30 going to be Amplified and accelerated
1:09:32 and so I think we're going to see
1:09:35 once we get through the brutal phase
1:09:37 which my guess is it's it's two to three
1:09:40 years right starting now like we're
1:09:43 starting to see it you're starting to
1:09:45 hear announcements of oh this company's
1:09:48 laying off 4 500 people and or they're
1:09:51 they're not gonna you know they're doing
1:09:52 a hiring freeze they were going to hire
1:09:54 7 500 but now they're not they're going
1:09:56 to focus on AI
1:09:59 um
1:10:00 so I think it's going to be a two to
1:10:02 three year brutal brutal patch but I
1:10:05 think the
1:10:06 the Renaissance of creativity and
1:10:09 Innovation that comes out the other side
1:10:11 of it is is going to be beautiful brutal
1:10:13 but beautiful process I love the way you
1:10:16 say that
1:10:19 Chief super alignment officer yeah oh by
1:10:22 the way open AI for the for the
1:10:25 alignment division for that super
1:10:26 alignment thing that they're talking
1:10:28 about they're hiring the base salary for
1:10:31 anyone in that group is four hundred
1:10:33 thousand dollars that's starting
1:10:35 starting salary
1:10:37 you give me money
1:10:41 so let me get this straight I just got
1:10:43 to get the robot not to kill the people
1:10:45 yep
1:10:47 and you give me 400 grand for that to
1:10:49 start
1:10:52 yep
1:10:54 go learn about alignment
1:10:56 start talking about it on LinkedIn have
1:10:58 a theory
1:11:00 go watch David Shapiro on YouTube
1:11:06 you want to feel intellectually
1:11:10 sketchy go watch David he's one smart
1:11:13 [ __ ] he's written five books on
1:11:15 this stuff
1:11:17 and he's got his last three or four
1:11:20 videos are all about alignment and about
1:11:22 open AI super alignment and what's good
1:11:25 about it what's bad about it he's
1:11:27 brilliant
1:11:29 yeah you want to get into this business
1:11:31 you want to get into this business save
1:11:33 Humanity it pays well
1:11:35 [Laughter]
1:11:38 also if you're geeky I'd get involved
1:11:41 with David Shapiro's uh open source
1:11:43 projects
1:11:45 I think I think that is probably time
1:11:47 well spent if for no other reason you're
1:11:49 going to learn a [ __ ] ton
1:11:52 um FRFR what and where are people
1:11:54 learning
1:11:55 um
1:11:57 well the what they're learning is what
1:12:01 these things are and what they can do
1:12:05 they're also learning about image
1:12:07 generation so mid-journey stable
1:12:09 diffusion Adobe Firefly Dolly those
1:12:12 tools are the picture equivalent of
1:12:15 these words
1:12:16 then if you go to futurepedia.io there's
1:12:18 a whole
1:12:19 other Suite of
1:12:22 um Technologies around video generation
1:12:25 audio generation text to audio
1:12:28 music synthesis music generation
1:12:34 things like that
1:12:38 um he's on some other stuff I'm on your
1:12:41 stuff now New York and queen I'm on you
1:12:43 I'm here for you ask a question I'll get
1:12:47 to it eventually
1:12:48 but then you gotta hang out maybe that's
1:12:51 maybe you know what I I just it just hit
1:12:53 me this is a it's a [ __ ] brilliant
1:12:55 strategy if you know that eventually
1:12:58 I'll get to your question you'll hang
1:12:59 out longer huh like a steel [ __ ] trap
1:13:04 all right
1:13:05 people are moving more motivated by fear
1:13:08 than desire of game yeah no I agree
1:13:10 people will do way more to avoid pain
1:13:12 than than to gain pleasure
1:13:16 um
1:13:17 and so what this channel is about is
1:13:22 well
1:13:23 the Pain's [ __ ] coming
1:13:26 so why don't you start now
1:13:30 and disrupt yourself before you get
1:13:32 disrupted that I guess that's that's
1:13:34 probably what where Where My Philosophy
1:13:36 is right now I want to learn as much as
1:13:39 I can about this one because it excites
1:13:41 me like the the the
1:13:42 potential of what's coming in the future
1:13:46 is really [ __ ] exciting for me it
1:13:48 looks really bright
1:13:50 but I also know
1:13:52 this shit's gonna [ __ ] up my like my
1:13:54 current companies an 11 year old
1:13:56 automated video storytelling platform
1:13:58 called storyvine this stuff's gonna [ __ ]
1:14:00 it up and I've got to reinvent it before
1:14:03 it gets reinvented for me so that's
1:14:06 that's where I want to be in these
1:14:07 conversations I want to be on the front
1:14:09 end of the reinvention not being forced
1:14:12 into it and that's that's where I think
1:14:14 the pain is going to be the highest is
1:14:15 people that are like I've got three
1:14:17 years I'll get to it
1:14:19 um
1:14:21 so yeah I totally agree
1:14:23 is one language any better than another
1:14:26 for creating prompts French more
1:14:27 descriptive I honestly don't know that's
1:14:29 a really good question I think that just
1:14:31 by the nature of the public internet you
1:14:35 know a lot of it being English
1:14:38 um I think it I think it probably
1:14:40 indexes more strongly toward English but
1:14:43 to your point certain languages are
1:14:46 going to be better at certain things I
1:14:49 do know that you can prompt in one
1:14:50 language and you know get results in
1:14:52 another like it's it's fantastic at
1:14:54 translation so but that that would be an
1:14:57 interesting experiment to to play with
1:14:59 if you've got that capacity
1:15:03 let's see not me having assuming it had
1:15:06 a meta awareness oh
1:15:09 what kind of tripod or phone stand do
1:15:12 you use for your lives it's what's
1:15:13 called a chicken foot monopod
1:15:16 so
1:15:17 so let's see wait can I
1:15:20 um
1:15:22 it doesn't matter it's basically a stick
1:15:25 with three feet on the bottom and then a
1:15:28 and then a camera tilting pan head on
1:15:31 the top
1:15:32 and just the phone thing and then this
1:15:34 little window this is just a tick tock
1:15:37 feature called dual
1:15:39 where this the back camera is on and the
1:15:42 front camera is on at the same time
1:15:46 um what will it do to the music industry
1:15:49 in Grammy's
1:15:50 will it make movies yes it will make
1:15:52 movies if you want to go make some
1:15:53 movies for yourself
1:15:55 they are laughably bad right now but so
1:15:57 were images a year ago and now they are
1:15:59 terrifyingly good so that's going to be
1:16:02 the same trajectory with video so go to
1:16:04 runwayml
1:16:06 I think it's runwayml.com
1:16:09 let me get that for you
1:16:15 yeah runwayml.com they've got two things
1:16:18 one called gen one one called gen 2. gen
1:16:21 1 you can upload a movie and then
1:16:24 describe a new style like claymation
1:16:26 animation and it'll take your clip and
1:16:29 turn it into a claymation animation
1:16:32 but you know with all the composition of
1:16:35 your thing and then Gen 2 is text to
1:16:38 video so you just describe a video and I
1:16:40 think right now it'll make a three or
1:16:41 four second clip
1:16:43 and
1:16:44 they're disturbingly bad but so were
1:16:47 images a year ago and so they're going
1:16:49 to get really really good in fact if you
1:16:52 listen to the CEO of stability AI he
1:16:54 said the the the models that they're
1:16:56 working with right now inside
1:16:59 stability AI stable diffusion
1:17:02 um
1:17:03 the the things that we get to play with
1:17:05 right now
1:17:06 to create an image it takes about I
1:17:09 don't know five seconds ten seconds
1:17:11 something like that depending on how
1:17:13 busy it is the models they're working
1:17:15 with internally are generating 60 images
1:17:18 a second
1:17:20 why is that number significant
1:17:26 that's that's the frame rate of video
1:17:30 games
1:17:31 24 is the frame rate of movies 30 frames
1:17:35 per second is the frame rate of video
1:17:38 60 is the frame rate of games they're
1:17:39 120 but 60 60 frames a second
1:17:43 so so we are very very quickly going to
1:17:47 be entering a world where you can say
1:17:50 you know write me these five scenes or
1:17:52 or tell chat gbt hey go write me you
1:17:56 know
1:17:56 16 different scenes of a video
1:18:00 send that to the stable diffusion API
1:18:03 and and then I want to watch that video
1:18:05 put this kind of music behind it put
1:18:07 this kind of dialogue behind it I want
1:18:10 Tom Cruise doing the voiceover I want
1:18:12 Morgan Freeman doing the voiceover
1:18:15 and go make me that movie and it'll just
1:18:18 start playing we're not that far from
1:18:20 that
1:18:25 um what will it do the music industry it
1:18:27 I it it transform it it transforms it in
1:18:31 ways we can't comprehend
1:18:33 so if you if you if you want to see kind
1:18:36 of the first
1:18:38 you know iterations of this just go to
1:18:40 Twitter and look up Drake the weekend AI
1:18:44 and you'll see a song there that someone
1:18:47 created that was like a fan generated
1:18:48 song where they took the weekend and
1:18:51 Drake's voice and put it in this AI
1:18:54 generated song and it got all these it
1:18:57 got like 20 million hits on Spotify and
1:18:58 then they took it down and that's when
1:19:01 Grimes Elon musk's ex said hey you can
1:19:04 sample me all you want go make as much
1:19:06 music as you want just give me 50 of the
1:19:08 proceeds
1:19:09 which I think is smart but but she's
1:19:12 going to be the exception most artists
1:19:14 are going to be like I'll sue your ass
1:19:17 but but we're quickly getting to a point
1:19:20 where fan generated music
1:19:22 is going to be as good or better
1:19:26 than the original
1:19:29 what do you do with that do you sue your
1:19:31 fans
1:19:33 I worked for a guy named Gordon
1:19:35 Patterson
1:19:37 who worked at new line Cinema and and
1:19:40 when the when the first Lord of the
1:19:42 Rings came out when it when it was you
1:19:44 know in production
1:19:47 um all the geeks on the internets
1:19:50 started making fan sites
1:19:53 and the lawyers at new line Cinema were
1:19:55 like we're gonna sue their ass
1:19:58 and Gordon Patterson was the guy that
1:20:01 said don't sue them invite them to
1:20:05 Comic-Con he was the guy that invented
1:20:08 social marketing for movies
1:20:12 I think it has to go in that direction
1:20:16 right
1:20:17 you're going to have fans creating
1:20:19 content that sounds like you
1:20:22 the songs might be better than yours
1:20:25 and they're leveraging your likeness and
1:20:28 technically they should be paying you
1:20:30 for that but they just did it out of
1:20:32 love
1:20:32 what do you do with them do you sue them
1:20:34 or do you embrace it
1:20:36 and and what does that do to the to the
1:20:38 if if a fan generated song
1:20:42 becomes a hit
1:20:43 what happens who gets paid I don't know
1:20:47 that shit's gonna have to be figured out
1:20:52 and and what's going to happen is the
1:20:55 recording industry and the movie
1:20:57 Industries it's happening right now with
1:20:58 the Writers Guild with movies
1:21:00 part of the reason they're on strike is
1:21:02 you can't use AI to write
1:21:05 screenplays without us
1:21:09 so you've got the most powerful writing
1:21:12 tool in the history of mankind
1:21:16 and the union for the writers is saying
1:21:20 don't use that
1:21:23 I've written seven screenplays seven
1:21:25 feature-length screenplays like I know
1:21:27 that world well
1:21:29 why the [ __ ] wouldn't you use
1:21:32 the most powerful writing tool Humanity
1:21:35 has ever seen
1:21:36 well because it's gonna [ __ ] change
1:21:39 everything
1:21:41 it's going to change the economics it's
1:21:43 going to change the power centers
1:21:45 it's gonna happen in music it's gonna
1:21:46 happen in film it's going to happen in
1:21:48 writing it's going to happen in the
1:21:49 agency world it's going to happen in the
1:21:52 Consulting world all knowledge work all
1:21:55 creative work
1:21:57 the models are going to be decimated
1:22:02 brutal but beautiful
1:22:04 was said just before
1:22:07 brutal but beautiful is coming what's
1:22:09 the beautiful I don't know we come out
1:22:12 the other side
1:22:13 I mean imagine a world where
1:22:17 we each get to have our own
1:22:20 radio station
1:22:23 that is music
1:22:25 we create that sounds like our favorite
1:22:29 music
1:22:31 whatever
1:22:33 how do you comprehend that I mean
1:22:36 there's a Spotify playlist sure
1:22:38 but what if that Spotify playlist is all
1:22:41 stuff that you created
1:22:43 they're singing about Taylor Swift is
1:22:44 singing about your personal issues
1:22:48 why is that a bad thing that's [ __ ]
1:22:50 staggeringly beautiful
1:22:52 what does the business side of that look
1:22:55 like I don't have a [ __ ] clue
1:22:59 but it's a pretty remarkable possibility
1:23:04 did the lawyer just straight up say that
1:23:06 I don't I don't know what I said but I'm
1:23:08 sure he said something stupid all right
1:23:10 is chat you PT hallucination drag legal
1:23:13 or illegal I want to try it
1:23:15 yeah that's good it's just oh drug legal
1:23:20 yeah
1:23:21 hallucination drug
1:23:23 it's you know it basically Hallucination
1:23:26 is a fancy word for it it's just it just
1:23:28 it's just ignorant of what it said
1:23:35 I'll explain I'll explain how the
1:23:37 hallucination works it's actually
1:23:39 fascinating
1:23:40 so Andre carpathy who's the I guess he's
1:23:42 the the CTO of open AI
1:23:45 he was talking to the Microsoft
1:23:46 developer conference
1:23:48 and he was talking about
1:23:50 um
1:23:51 why these large language models
1:23:52 hallucinate like what's going on
1:23:55 so the the metaphor I like to think is
1:23:58 so imagine the Grand Canyon is full of
1:24:00 all the words that were on the public
1:24:02 internet so everything that open ai's
1:24:04 been trained on right the large language
1:24:05 models
1:24:07 and if you're in a helicopter and you're
1:24:09 and you're high enough you can see down
1:24:10 all the Canyons in the Grand Canyon you
1:24:12 can see all the words so if you say to
1:24:15 chat GPT solve this mathematical
1:24:18 equation
1:24:19 and you're really high in the helicopter
1:24:22 it's gonna see
1:24:24 all of the places where that that
1:24:26 mathematical equation was solved and and
1:24:28 in that data set you're going to have
1:24:30 the Stanford mathematics professor
1:24:33 that got it elegantly right
1:24:35 and you're going to have all of the
1:24:37 students from all of the universities
1:24:39 who got it devastatingly wrong it's
1:24:42 gonna all of that's going to be in the
1:24:43 data set So when you say solve this
1:24:45 mathematical equation your helicopter is
1:24:47 really high up it sort of takes all
1:24:49 those answers amalgamates them and gives
1:24:51 you an answer it says here's your answer
1:24:53 and you're like that answer is comically
1:24:56 bad and it's like I'm sorry sorry I got
1:24:58 that wrong here's the answer it's like
1:25:00 well that's better but you didn't add
1:25:02 these two numbers right oh I'm sorry I
1:25:04 got that wrong because it's it's sort of
1:25:06 amalgamating right and wrong answers now
1:25:08 if you say to chat gbt I want you to act
1:25:10 like a Stanford mathematics professor
1:25:12 that specializes in this branch of
1:25:14 mathematics that really understands this
1:25:17 particular corner of the universe of
1:25:20 this kind of equation and I want you to
1:25:21 solve that equation what you've done is
1:25:24 you've kind of flown your helicopter
1:25:25 down in one of the little side Canyons
1:25:28 where the information in that side
1:25:30 Canyon is only for professors of
1:25:32 mathematics and so it will hallucinate
1:25:34 less because it's got less information
1:25:36 to draw from
1:25:38 which is just wild and and again all all
1:25:41 it's doing all it's doing it's just
1:25:43 [ __ ] remarkable but but what it's
1:25:45 doing is predicting one word after the
1:25:47 other to give you the answer so your
1:25:50 trick in prompting these tools is to
1:25:53 give it enough context to limit the
1:25:56 semantic universe
1:25:58 so that you have a higher likelihood of
1:26:00 getting what you want that's the art
1:26:03 is lower your helicopter into one of the
1:26:05 side Canyons
1:26:07 when will it start thinking for itself
1:26:10 probably within three years
1:26:17 and that paper from Microsoft where
1:26:20 they've got a billion
1:26:24 token context window
1:26:26 750 million words
1:26:29 um
1:26:31 if that [ __ ] starts showing up it maybe
1:26:33 happens sooner than that
1:26:38 sometimes it really pauses for a long
1:26:40 time when I ask it complex questions
1:26:42 generally Angela That's not that you ask
1:26:45 that a complex question generally it's
1:26:47 that the servers are just slammed
1:26:50 because hundreds of millions of people
1:26:51 are using it
1:26:54 um
1:26:54 that that's generally just a a
1:26:58 server load issue
1:27:05 yeah
1:27:07 good evening Emilio's wife hope you're
1:27:10 well
1:27:11 I think comp equals comparables I got
1:27:13 that eventually in real estate
1:27:15 terminology
1:27:16 it took me a while it was like com
1:27:18 compensation and then oh yeah real
1:27:20 estate comps but it can do comps really
1:27:22 well
1:27:23 our prompts copyrightable there's a
1:27:25 there's a site called prompt
1:27:28 what the hell is it called
1:27:34 [Music]
1:27:44 prompt base if you go to promptbase.com
1:27:48 you can sell prompts
1:27:51 why you would buy a prompt there I mean
1:27:54 so here's the deal about prompting right
1:27:56 now prompting is really important
1:28:03 it isn't going to be for long if if you
1:28:05 want to build tools you're going to need
1:28:06 to be good at prompting if you just want
1:28:08 to use these tools
1:28:10 a lot of this prompt crafting [ __ ] is
1:28:13 going to drift to the background because
1:28:14 the tools are going to start prompting
1:28:15 themselves if you go look at what what
1:28:18 um
1:28:20 what
1:28:23 um code interpreter does
1:28:26 you you upload a file so let's say you
1:28:29 you upload an Excel file and then you
1:28:31 say tell me what's in this
1:28:34 it writes itself its own set of
1:28:36 instructions it then
1:28:38 writes itself python applications to
1:28:41 execute those instructions executes that
1:28:44 python application and if it gets it
1:28:46 wrong it says you know what
1:28:48 I didn't I didn't do that right there's
1:28:50 another way we could attempt this let me
1:28:52 try that and it'll write another python
1:28:54 script and execute it so these things
1:28:56 are increasingly going to be
1:29:00 self-executing
1:29:03 um
1:29:04 very sophisticated prompts from very
1:29:07 minimal input they're not there right
1:29:09 now so that's why learning The Prompt
1:29:11 crafting right now is valuable
1:29:13 but I don't know that you would need to
1:29:16 um
1:29:18 copyright a prompt because they're so
1:29:20 easy to to
1:29:22 you know just copy
1:29:24 but go to prompt base and maybe short
1:29:27 term you can sell some
1:29:29 if you're really good at it
1:29:31 and you got something slick that no one
1:29:33 else has figured out
1:29:37 why not
1:29:39 I doubt it Captain lit
1:29:42 is one language any better than the
1:29:43 other for creating prompts oh we talked
1:29:45 about that one you keep talking about
1:29:47 that code interpreter I just started
1:29:49 installing it and using plugins today
1:29:52 um all right let me go show you we'll go
1:29:55 do a code interpreter thing here
1:30:01 if you pay for chat GPT plus you get
1:30:05 code interpreter and plugins
1:30:08 um I'll show you how to turn them on in
1:30:09 a second here
1:30:11 there's a fourth mode that used to be
1:30:14 there that's no longer there which is
1:30:15 browse with Bing it's it's gpt4
1:30:18 connected to the internet
1:30:19 they they [ __ ] up
1:30:21 and so they had to kill it so they'll
1:30:23 bring it back when they fix it
1:30:25 but I'll show you code interpreter if
1:30:27 you don't see code interpreter if you do
1:30:29 pay for Chad gbd plus and you don't see
1:30:31 code interpreter you have to turn it on
1:30:33 so you go to your settings
1:30:37 and
1:30:39 you click on beta features and then you
1:30:42 have to turn it on so you have to
1:30:44 explicitly turn on plugins and code
1:30:46 interpreter
1:30:48 um
1:30:49 and then when you put code interpreter
1:30:51 here you can see there's there's a plus
1:30:53 button here
1:30:56 and so
1:30:58 that plus button is you can upload a
1:31:01 file
1:31:03 so I am going to go grab
1:31:13 let's grab this one
1:31:15 so it's IMDb
1:31:18 tv.zip so this is a ZIP file
1:31:21 so I'm going to say what is in this
1:31:25 question mark
1:31:31 so it says let's unzip the file and
1:31:33 inspect its contents okay cool that's
1:31:36 cool well how's it doing that all right
1:31:39 let me turn down my brightness so you
1:31:40 can see a little bit better
1:31:42 now you can barely see it okay
1:31:45 um if you open these things and show the
1:31:48 work
1:31:49 in order for it to open the zip file it
1:31:51 actually wrote a python application it
1:31:53 it imported libraries and wrote python
1:31:57 to unzip the file
1:31:58 so it says it seems there was an error
1:32:00 the unzipped folder's name might not be
1:32:02 IMDb
1:32:04 and then but then it did it so so it
1:32:07 unzipped it
1:32:10 and then it says okay what's in here
1:32:13 are the following CSV files so mystery
1:32:16 sport biography so this is IMDb
1:32:19 categories data about IMDb TV shows
1:32:22 these files are likely data sets related
1:32:24 to TV series in their respective genres
1:32:26 would you like to inspect
1:32:29 the content of any specific file
1:32:32 can you
1:32:34 um
1:32:35 analyze
1:32:37 the
1:32:39 data across these files and
1:32:44 visualize
1:32:46 the trends
1:32:48 you think are interesting in them
1:32:55 question mark
1:33:00 sure to start the analysis so now it's
1:33:02 writing instructions load the CSV files
1:33:06 into Data frames inspect the structure
1:33:08 of the data to understand its columns
1:33:10 and their types identify interesting
1:33:12 columns for analysis so let's start by
1:33:15 loading the first file
1:33:16 so the csvs contain title ID release
1:33:19 year genre cast synopsis rating runtime
1:33:23 certificate the age certificate of the
1:33:26 series oh that's interesting number of
1:33:28 votes and then gross revenue
1:33:31 given this structure we could
1:33:32 potentially analyze and visualize Trends
1:33:34 related to ratings genres release years
1:33:37 certificates and number of votes across
1:33:40 all the files let's proceed by loading
1:33:42 all the CSV files and combining them
1:33:45 into a single data frame
1:33:47 so so then it writes code to do that
1:33:52 and look here's
1:33:54 this is a it made a little thing here
1:33:56 certificate PG-13 number of votes
1:34:01 Revenue not available
1:34:03 genre mystery
1:34:06 so we're going to clean the release year
1:34:09 to only include the start year of the
1:34:11 series
1:34:15 uh distribution okay there you go look
1:34:18 so
1:34:19 so now it said let's start by looking at
1:34:21 the distribution of ratings across the
1:34:23 different genres
1:34:25 we can do this by creating the box plot
1:34:27 for each genre so it figured out what
1:34:30 chart it wanted to build it wrote The
1:34:32 Python code to take all the data
1:34:35 it colored all this it created the Box
1:34:38 chart it labeled it all so this is
1:34:39 Mystery
1:34:42 sport biography so here's all the
1:34:44 different genres down here
1:34:49 so here's some sort of scatter chart
1:34:50 that didn't doesn't look like it had
1:34:52 good data so that was a bad piece of
1:34:54 data it looks like this is
1:34:57 number of series released per year per
1:35:00 genre
1:35:03 and you know 1920 there was not much TV
1:35:07 released
1:35:09 and then 60s
1:35:12 80s 2000 so here's so right after the
1:35:15 80s is when cable comes in and you start
1:35:18 to see it increasing and then now we're
1:35:21 in streaming and you know
1:35:23 all this stuff and and this thing's
1:35:26 going to continue to go exponential
1:35:27 because
1:35:29 um of all the content that's going to be
1:35:31 created by AI but anyway like what what
1:35:33 was our input here our input here was
1:35:36 two or three requests and it just went
1:35:39 off and is doing full-on data analysis
1:35:43 across all of the CSV files within this
1:35:47 one zip file
1:35:50 so
1:35:58 that's what code interpreter is now what
1:36:00 can you use that for I don't know yet
1:36:04 you can upload images and pull out
1:36:07 color palettes out of it you can swap
1:36:09 images you can turn a static image into
1:36:12 an animation with a zoom and a pen
1:36:16 um
1:36:18 you could upload
1:36:21 tax information you could upload
1:36:23 insurance information start interacting
1:36:26 with it
1:36:31 I think something interesting I might do
1:36:33 is
1:36:35 throw all the screenplays
1:36:37 I've written
1:36:39 plus all of the fragments of screenplays
1:36:41 I've started which there's probably 20
1:36:43 or 30 of those
1:36:46 put them all in a zip file and upload
1:36:49 that and then be able to interact with
1:36:51 be able to ask it about anything that
1:36:53 I've written
1:36:55 and have it write me new [ __ ] that
1:36:57 follows my style
1:37:01 what would that do I don't know
1:37:08 I mean I could have it compare plot
1:37:10 lines
1:37:11 across screenplays and say
1:37:15 what are things that that my screenplay
1:37:19 structures have in common
1:37:22 would that be valuable I don't know
1:37:23 maybe
1:37:24 it might
1:37:26 it might inspire me to tell a new story
1:37:29 I might be able to take a fragment of a
1:37:30 screenplay and say hey
1:37:32 you've got seven complete ones here
1:37:34 you've got four fragments
1:37:36 which of the fragments seems most
1:37:40 um
1:37:41 thank you very much I appreciate that
1:37:44 which of the fragments seems the most
1:37:46 interesting to you and let's go ahead
1:37:48 and complete that
1:37:51 so give me an outline complete the
1:37:53 outline so that it matches what I like
1:37:56 to do apparently
1:37:59 and then
1:38:00 you've got two scenes there let's let's
1:38:02 write the third scene I don't know
1:38:05 it's like I
1:38:08 I was just starting to get my head
1:38:10 around chat GPT and feeling like I had
1:38:12 it together
1:38:13 and then code interpreter came out and
1:38:15 now I'm like right back to Ground Zero
1:38:17 feeling like an idiot what do I do with
1:38:18 this I don't know I'm gonna figure it
1:38:21 out
1:38:22 I want a grand Adventure
1:38:26 you should join the adventure go there
1:38:29 go there now if you have not played with
1:38:31 chat EBT go there until you have this
1:38:34 moment
1:38:36 and then start your adventure
1:38:39 you forgot the light bulb I don't know
1:38:41 which light bulb you're talking about
1:38:46 the ticket for the train car first seat
1:38:48 is a dollar but there's another train
1:38:50 coming for two uh
1:38:53 um
1:38:54 and another train but the ticket is ten
1:38:58 I think you're building to a point there
1:39:00 or maybe that was something I was
1:39:01 talking about and you're adding to it it
1:39:04 seems right
1:39:06 get on the train now
1:39:08 okay wait so code interpreter can take
1:39:11 in 100 megabytes of excel software data
1:39:14 I want to give it a word doc give it a
1:39:16 word doc
1:39:18 give it a PDF you can give it a PDF
1:39:23 you know how there's PDFs that have data
1:39:25 in them and there's PDFs that are just
1:39:27 scans that are useless
1:39:29 you can give it a scanned PDF and it
1:39:32 will OCR the PDF pull the data out of it
1:39:35 and throw it into a spreadsheet for you
1:39:39 it's [ __ ] insane
1:39:42 [Applause]
1:39:47 and by the way code interpreter is the
1:39:49 worst possible [ __ ] name for it we
1:39:51 talked about it last night we're going
1:39:52 to come up with a better name jumped
1:39:54 into AI just for fun but eventually
1:39:55 found myself using it daily for work and
1:39:58 personal life Yes Albert that is the
1:40:00 journey
1:40:01 I jumped into it just for fun and then I
1:40:03 realized holy [ __ ] [ __ ] did you know
1:40:06 what this thing can do
1:40:09 that's the perfect Evolution Albert it's
1:40:12 the perfect evolution
1:40:14 now I'm using it daily and and I think
1:40:17 the evolution after that is
1:40:21 I got an idea for something
1:40:24 I'm going to change how I do something
1:40:26 I'm going to reinvent something I'm
1:40:27 going to reinvent myself I'm going to
1:40:28 reinvent my job
1:40:31 I think this I think this is just going
1:40:33 to be this rolling
1:40:35 everyone who who starts playing with
1:40:37 this stuff early
1:40:39 it's going to be a rolling set of
1:40:41 Revelations about what it makes possible
1:40:47 and it's going to be both invigorating
1:40:49 and [ __ ] terrifying
1:40:53 and I think that's the right like
1:40:55 I keep going back to this idea of we're
1:40:58 on an adventure whether we want to be on
1:41:00 the adventure or not it's like
1:41:03 at night there seems to be a dragon
1:41:05 coming into the village and killing
1:41:07 people
1:41:10 and they're like well who's gonna go
1:41:12 [ __ ] deal with the dragon and
1:41:14 everyone's like I don't want to deal
1:41:15 with the dragon
1:41:18 and then there's like
1:41:19 five people that like are having drinks
1:41:21 one night they're like you know what I I
1:41:23 think it's just gonna keep killing
1:41:24 people maybe we should you want to go
1:41:26 find this dragon
1:41:28 well that's really scary yeah but you're
1:41:30 strong and you're good with a bow and
1:41:32 arrow and you can do those magic spells
1:41:33 maybe we should you wanna all right
1:41:39 you know and then little cousin Lucy's
1:41:41 eaten by the dragon the next night and
1:41:43 they're like [ __ ] it we're going
1:41:46 I feel like that's where we are with
1:41:47 this stuff
1:41:50 so so those kind of Adventures are
1:41:52 terrifying and they're inspired
1:41:57 and you can't do them alone and
1:42:04 you figure out
1:42:07 who's the secret hero and you figure out
1:42:09 who's not
1:42:14 that's where we are
1:42:18 so yes Albert that feels very right to
1:42:21 me
1:42:22 how are they making money from it
1:42:24 you got my money
1:42:26 not from those [ __ ]
1:42:29 um
1:42:30 so here's the deal about Ai and these
1:42:32 tools is they're just tools now
1:42:36 they happen to be the most powerful
1:42:38 tools in the history of humanity so
1:42:41 you know I'm not diminishing the value
1:42:44 of them but um they're just tools so how
1:42:47 you make money with tools is you solve
1:42:49 problems and so
1:42:52 what I say a lot is so how people are
1:42:54 making money with them right now is
1:42:56 they're learning these tools well enough
1:42:57 that they can teach other people how to
1:42:59 use them so they're they're doing
1:43:00 consulting they're doing workshops
1:43:02 they're
1:43:03 um
1:43:03 building automations for people they're
1:43:06 saying hey I can I can make you've got a
1:43:08 social media agency I can build you a
1:43:10 tool that will automatically generate
1:43:11 tweets and posts and this and that all
1:43:14 you have to do is put in the topic and
1:43:16 answer a question or two and it'll
1:43:17 generate all that for you you want me to
1:43:19 make one of those for you uh sure okay
1:43:23 what do you want me to deliver that by
1:43:24 Friday you can you can do that yeah
1:43:28 here you go here's your tool that's
1:43:29 [ __ ] remarkable yeah I know do you
1:43:32 want to learn how to do that yeah okay
1:43:33 pay me this money to you know so that's
1:43:35 that's how people are making money right
1:43:37 now
1:43:39 um
1:43:42 and then and then it it comes down to
1:43:47 what are you good at
1:43:50 so no name
1:43:52 coming up with clever names is
1:43:54 Apparently one of them
1:43:57 um but you know so you've got writing
1:43:58 skills or you've got project management
1:44:00 skills or you're good at coding or
1:44:03 you're
1:44:04 good at illustrating doesn't matter like
1:44:06 what is the thing you're good at and
1:44:08 then I would say take what you're good
1:44:09 at what you're passionate about
1:44:11 ideally those things are decently
1:44:12 aligned if they're like this well
1:44:14 that sucks sometimes sometimes you got
1:44:17 to pay the bills right
1:44:19 but take the things where you've got
1:44:20 competence and ideally where you've got
1:44:23 some passion and then start applying
1:44:24 these tools to that stuff
1:44:28 start learn
1:44:30 how AI can make what you do in project
1:44:33 management better faster stronger
1:44:35 cheaper
1:44:36 and then start putting it out in the
1:44:38 world and say hey
1:44:40 I figured out a better way to do project
1:44:41 management you want to pay me for that
1:44:43 and that could look like a software
1:44:45 application that could look like an
1:44:47 agency that could look like Consulting
1:44:50 so you solve problems and then these are
1:44:53 just the tools
1:44:54 so the trick is you got to get competent
1:44:57 enough with the tools to understand what
1:44:58 the [ __ ] they are in the first place and
1:45:00 that's almost impossible because it
1:45:02 changes every three months
1:45:04 but start now like if you're feeling if
1:45:07 you're feeling like holy [ __ ] am I
1:45:09 behind that's do you you don't think I
1:45:11 feel like
1:45:13 one thousand percent inadequate to deal
1:45:16 with something as powerful as code
1:45:18 interpreter
1:45:19 I don't know anything about data
1:45:21 analysis I have a degree in [ __ ]
1:45:23 acting and even though I've been kind of
1:45:24 tech you know in the tech periphery I've
1:45:27 innovated on the on the sort of
1:45:29 boundaries of of technology for years
1:45:33 that's significantly beyond my current
1:45:35 capabilities but
1:45:38 it's going to change the [ __ ] world
1:45:40 so why would I not play with that so I'm
1:45:42 gonna go through the pain
1:45:44 on the adventure
1:45:46 of walking through the swamp with all
1:45:48 the snakes and the leeches
1:45:52 thinking that I might drown or die or go
1:45:54 in quicksand
1:45:57 to get on the other side of it because I
1:45:59 know on the other side if I can get my
1:46:01 head around what code interpreter is
1:46:03 I can do some [ __ ] remarkable [ __ ] we
1:46:06 all can we all will
1:46:08 I just want to be there first
1:46:11 well no not first I want to be there I
1:46:14 want to be in the party that slays the
1:46:15 Dragon
1:46:17 and I think that that you know
1:46:20 that party is is
1:46:22 entering the Forest right now
1:46:25 I think this [ __ ] gets boring like five
1:46:27 years from now when all this shit's
1:46:29 figured out
1:46:31 and we're back to like grinding out the
1:46:35 social media post for whatever the [ __ ]
1:46:37 SEO looks like five years from now it
1:46:39 ain't gonna look like what it looks like
1:46:40 today I'll tell you that
1:46:47 I'd like to see AI do mashups it can do
1:46:49 all sorts of mashups figure out which
1:46:51 ones you want them to do and go find the
1:46:53 tools and start doing it
1:46:55 how do they control bias and AI
1:46:58 um they do it through a combination of
1:47:02 um
1:47:03 improving the training data that they
1:47:06 put in the base model
1:47:07 they improve it doing the fine tuning of
1:47:10 how they get it from that base model to
1:47:12 something usable
1:47:14 um
1:47:15 right now with open AI it's done with
1:47:17 humans human reinforcement learning
1:47:20 I think increasingly it'll move to
1:47:23 machine mediated so if you're interested
1:47:27 in this stuff go look up a company
1:47:29 called anthropic
1:47:32 um that's got a model called Claude
1:47:34 there are a bunch of EX open AI people
1:47:35 and they got this thing called the
1:47:37 Constitutional learning model where they
1:47:38 basically write a set of rules like
1:47:40 don't hurt people increase Prosperity
1:47:43 make happy whatever the heck their rules
1:47:46 are and then when their system generates
1:47:49 things it Compares it against those
1:47:51 rules I think that's a really
1:47:52 interesting model
1:47:54 what's up
1:47:56 hang on
1:48:17 love you
1:48:18 all right
1:48:19 I love that
1:48:24 um I have a beer truck side business and
1:48:26 I've been using AI program some things
1:48:28 on our website oh that's super cool yeah
1:48:31 and I mean you can also use uh AI Albert
1:48:34 to do uh you know have it come up with
1:48:36 marketing ideas
1:48:38 um tell it you want to do you know
1:48:40 community events and you want you want
1:48:42 ideas for that
1:48:44 um you know describe your community
1:48:45 describe your business you know give it
1:48:48 give it as much context as you can and
1:48:51 just yeah keep kind of expanding that
1:48:53 out that's that's really awesome I mean
1:48:55 and Albert it's so I love the comments
1:48:58 that you're making so far because it's
1:49:01 I think I think it's very easy to over
1:49:04 complicate what you do with these tools
1:49:07 like I gotta learn about machine
1:49:09 learning and data science and this and
1:49:11 no no no no
1:49:13 thank you
1:49:14 um
1:49:15 you already know what you know
1:49:19 you just need to learn enough of these
1:49:21 things to understand oh those could help
1:49:24 me do the [ __ ]
1:49:26 I already do better
1:49:29 like these are amplifiers of humanity of
1:49:32 what you want to do so you're still
1:49:36 you still get to drive the train this is
1:49:39 just your tool like it's it the best way
1:49:41 I can describe what what Chachi BT is is
1:49:44 it's like an idiot savant meaning
1:49:48 it's really clueless as as a human as a
1:49:51 you know you got to tell it everything
1:49:53 and it will lie and it won't you won't
1:49:57 even know that it lied it will
1:49:58 confidently lie it will mansplain to you
1:50:01 but it's also a savant in that it knows
1:50:04 everything it knows all the stuff
1:50:07 so if you say Okay idiot savant I want
1:50:12 you to act like a master marketer that
1:50:14 specializes in food trucks and I've got
1:50:16 a beer truck and I got a this and I
1:50:18 gotta that and this is my community
1:50:22 can you come up with some ideas for me
1:50:24 like what what are some ideas that
1:50:26 aren't going to cost me a lot but are
1:50:28 going to drive lots of Engagement and
1:50:29 drive lots of awareness oh sure
1:50:33 here you go
1:50:35 okay I like two and three but the rest
1:50:38 are horrible can you give me more like
1:50:39 two and three oh sure sorry about that
1:50:45 so so
1:50:49 while they're remarkably powerful
1:50:51 they're at your service
1:50:53 you're the pilot that's the co-pilot
1:50:57 so yeah love that
1:51:01 love it love it love it love it love it
1:51:04 love it love it love it now Tick Tock
1:51:06 just jumped me to the bottom and I lost
1:51:09 my place in the questions and I'm a dum
1:51:11 dum dum dum dum dum
1:51:14 what'd you do tonight watch some old guy
1:51:16 scroll through tick tock
1:51:18 that sounds awful yeah pretty much
1:51:21 so as if I got a haircut yes I did thank
1:51:24 you Mike looks good thank you oh wow
1:51:27 it's I hadn't even really noticed it it
1:51:29 was up in the letters above that's uh
1:51:32 that that's almost like I tried it
1:51:34 says
1:51:38 thank you hair growth geek
1:51:42 oh good lord good Lord the hair Saga at
1:51:46 the AI learning lab it's it's always
1:51:48 something
1:51:55 Chief fear officer that was a good one
1:51:58 sorry I'm still trying to find where he
1:52:00 was oh well it didn't look it's not let
1:52:02 me scroll all the way back
1:52:03 Google should have bought chat GPT
1:52:06 yeah Google should have used the
1:52:08 technology they invented I said this the
1:52:11 other night
1:52:12 this happened at Kodak hey bosses I just
1:52:16 invented the digital camera I think this
1:52:18 is going to change everything the bosses
1:52:21 how's that going to sell us more film uh
1:52:24 it doesn't use it's not it doesn't use
1:52:27 it's a new way of making pictures yeah
1:52:29 that's not for us we make film let me
1:52:32 let me get let me give you a little some
1:52:34 lessons here little kitty this is about
1:52:36 the silver halide business not about
1:52:39 pixels
1:52:42 that's what Google did when they
1:52:44 invented the Transformer apparently
1:52:47 so yeah
1:52:49 they could have bought open AI at some
1:52:52 point I suppose
1:52:54 but they were hiding from the technology
1:52:55 they invented because it threatened
1:52:57 their business model so now Microsoft is
1:52:59 like [ __ ] it we're in here's 10 billion
1:53:01 dollars you want to go do some damage
1:53:03 together Sam
1:53:05 all right let's go
1:53:08 and off they go I use code interpreter
1:53:10 to create a vegetarian menu
1:53:12 and ask it to make Excel sheet with
1:53:14 ingredients that's awesome and then with
1:53:17 plugins you can take that the
1:53:18 ingredients and just have it order it
1:53:20 although you're in a restaurant business
1:53:21 so you'll get them from your wholesalers
1:53:23 that's awesome
1:53:24 all right okay so I renewed my chat CBT
1:53:27 and installed new IOS app but I can't
1:53:29 find how to upload code generator Okay
1:53:31 so
1:53:33 the official
1:53:35 open AI chat GPT app
1:53:39 doesn't have all the features so what I
1:53:41 do on my phone is just go to Safari or
1:53:45 Chrome if depending on the device you're
1:53:47 on and go to chat.openai.com
1:53:51 and log in there now in your settings
1:53:52 you're gonna have to turn on
1:53:54 the beta features to make sure it's
1:53:56 available so either on a on a laptop go
1:53:59 to that site or on your phone go to that
1:54:01 site and then you'll be able to see code
1:54:03 code interpreter
1:54:05 code generator is so new that Google
1:54:07 doesn't know what what I'm talking about
1:54:09 when I ask for help I know it I know I
1:54:12 know this is we're on the we're on the
1:54:14 that's that's where we are right now
1:54:17 I think how steady and consistent wait I
1:54:20 think how steady and consistent you but
1:54:23 I feel bad because I don't know
1:54:26 you would get to all the comments oh
1:54:30 oh yeah well
1:54:32 you don't need to feel bad I just that's
1:54:34 just the nature of how I do this is I
1:54:37 just read all the things you don't need
1:54:38 to feel bad at all I I wish I could read
1:54:40 faster and talk less but I just this is
1:54:43 how I talk
1:54:43 [Laughter]
1:54:52 I'm like Chachi pite you can make money
1:54:54 much of a tie
1:54:58 just like life beautiful and brutal
1:55:01 the brutal and beautiful
1:55:03 the the yeah brutal but beautiful that's
1:55:08 that's what we're at the beginning of
1:55:10 and what's what's fascinating about it
1:55:13 is
1:55:16 like there's a couple of things going on
1:55:18 that I find really fascinating I think
1:55:20 because
1:55:21 blockchain and nfts
1:55:23 were over hyped and then crashed and
1:55:26 burned dramatically
1:55:30 I think there's
1:55:33 a reticence to oh this is just another
1:55:35 hype train so okay so let me get this
1:55:39 straight last year it was nfts we're all
1:55:41 gonna get rich on nfts and and this year
1:55:42 it's AI is it that's how it's going all
1:55:45 right yeah that's good
1:55:46 so so you got that
1:55:48 and then you've got 50 years of
1:55:50 Hollywood telling us the robots are
1:55:51 going to kill us
1:55:53 and then you've got
1:55:55 a tool that
1:55:57 got to 100 million users in six weeks
1:55:59 that is profoundly powerful where enough
1:56:02 people
1:56:04 are using it and discovering its power
1:56:06 like Albert I just started playing with
1:56:08 it and all of a sudden now I'm using it
1:56:09 for real [ __ ]
1:56:13 that we're in this kind of weird
1:56:15 confusing place right now
1:56:18 um
1:56:19 but man is it coming
1:56:23 brutal but beautiful
1:56:26 any recommendations for books not on
1:56:28 this [ __ ] well there's probably lots of
1:56:31 books if you want to understand machine
1:56:33 learning and neural networks and
1:56:35 Transformers and weights and embeddings
1:56:39 and Vector databases and all that sort
1:56:41 of [ __ ] if you want to be on the science
1:56:43 side of this
1:56:45 there's probably lots of books I don't
1:56:46 know that world at all I'm I'm on the
1:56:48 other side of the inflection point let's
1:56:51 just call it post chat gbt where it's
1:56:53 about applying these remarkable tools to
1:56:56 what we do day to day to day
1:56:58 um
1:57:00 if there are books on that [ __ ] right
1:57:02 now I wouldn't buy them
1:57:04 because because they're they're by the
1:57:07 by the time you could even do a
1:57:08 print-on-demand book right now for this
1:57:10 generative AI [ __ ] it's going to be
1:57:12 obsolete
1:57:15 I was I was pertin near an expert
1:57:18 at
1:57:20 um content generation engines using
1:57:22 zapier and GPT 3.5
1:57:26 and now they're deprecating 3.5
1:57:29 you've got to go to four
1:57:32 um or 3.5 turbo which acts different and
1:57:35 you've got to prompt it different and
1:57:36 then Along Comes code interpreter which
1:57:39 is a completely different Paradigm and
1:57:42 is probably closer to what this thing
1:57:45 looks like in the future anyway
1:57:47 um and I don't even think there's API
1:57:49 access to it yet so it's like just
1:57:51 everything's changing everything's
1:57:53 changing all the time and and like like
1:57:55 we're on it's funny in in when you
1:57:59 listen to VC's talk they're looking for
1:58:01 what they call the hockey stick where
1:58:03 revenue for a company is like zero or
1:58:05 going along like this and then all of a
1:58:07 sudden it goes like that or or you know
1:58:09 adoption goes like that
1:58:11 what's happening with these tools right
1:58:13 now is that's happening with Innovation
1:58:14 where the Innovations have have hit this
1:58:17 exponential curve where these things are
1:58:20 getting more and more and more powerful
1:58:21 exponentially and and we as humans don't
1:58:25 have a good way of understanding
1:58:26 exponential growth
1:58:28 so
1:58:30 it
1:58:33 it took me four months four months
1:58:38 so I've got a company that I've been
1:58:40 building for 11 years I know that thing
1:58:42 [ __ ] cold I know what it's good at I
1:58:44 know what it's bad at I know how to
1:58:46 position it I know generally how to sell
1:58:49 it although I wish I were better at that
1:58:51 but I know it called it took me four
1:58:54 months as I started learning this AI
1:58:56 stuff to even get them like this like
1:59:00 not even like this like it took me four
1:59:02 months to just have it not be a
1:59:04 completely foreign concept
1:59:06 and then over those four months as I
1:59:08 learned it more and more I'm like oh
1:59:09 okay so wait this piece of what that
1:59:11 does that applies to this piece of what
1:59:14 we do and then
1:59:15 and then like I'm now at the point where
1:59:18 I can kind of see the future of my
1:59:20 company with this AI stuff
1:59:22 based on
1:59:25 what I learned in February
1:59:30 now I take something like code
1:59:32 interpreter and it's like it's like
1:59:35 right over here so now I've got this
1:59:36 vision of my company that's based on
1:59:38 February technology
1:59:41 and here's code interpreter I'm like I
1:59:43 don't even [ __ ] understand what it is
1:59:44 much less how it relates to my business
1:59:46 or to what I do
1:59:49 so
1:59:50 there's no books
1:59:53 I mean don't get me wrong there's
1:59:55 probably lots of books right because
1:59:57 this thing can write books really quick
2:00:00 in fact I talked to a guy at an event
2:00:02 he's like hey I published 12 books in
2:00:05 the past three days and I'm like oh good
2:00:07 God
2:00:08 really
2:00:10 he was so proud of it
2:00:12 they had covers and titles and
2:00:14 self-publishing you had them printed out
2:00:16 and you could look at them and hold them
2:00:18 he's like yeah you just tell it to write
2:00:20 a book and it writes a book I'm just
2:00:22 publishing them
2:00:25 I'm [ __ ] exhausted
2:00:31 so so yeah there's lots of books there's
2:00:33 no books I would recommend
2:00:35 400 000 we're starting starting at
2:00:39 open AI
2:00:41 so if you go to the openai.com blog
2:00:45 look at their article for super
2:00:48 alignment
2:00:50 that's a new group that Sam Altman is
2:00:53 dedicating 20 of their compute resources
2:00:55 to making sure that their AI models are
2:00:59 aligned with human
2:01:01 goals or or rights
2:01:05 and that group the starting salary if
2:01:08 you get a job in that group is four
2:01:09 hundred thousand dollars
2:01:11 this 56 year old guy had fun blowing the
2:01:14 mind of a 20 year old with introducing
2:01:17 Chachi BT to the yes
2:01:20 [ __ ] gen xers I think I'm telling you
2:01:22 man this is this AI [ __ ] is Revenge of
2:01:25 Gen X we're gonna come out of the
2:01:27 Shadows because we know how to use
2:01:28 computers like we know how to [ __ ]
2:01:31 use computers we started some [ __ ] and
2:01:34 we're gonna get good at this and with 30
2:01:36 years of life experience plus these
2:01:39 tools we're gonna be able to [ __ ] some
2:01:40 [ __ ] up
2:01:42 and blow some 20 year olds Minds I like
2:01:44 it I like it that's solid that's great
2:01:47 oh wait wait you're going so fast go
2:01:51 where for image video music oh sorry the
2:01:54 very bottom URL is called futurepedia.io
2:01:58 that's a directory of like 4 000 AI
2:02:01 generated tools so
2:02:07 here's here's where I am philosophically
2:02:09 with all these tools the only things I'm
2:02:11 really paying attention to right now are
2:02:14 chat GPT and variations of it and and
2:02:17 some of the image generating tools
2:02:19 because I like doing the images so mid
2:02:21 Journey stable diffusion Adobe Firefly
2:02:26 um
2:02:28 all of the other [ __ ] text to video the
2:02:31 music stuff
2:02:33 I'm doing a little bit of voice
2:02:34 synthesis stuff I'm doing a little bit
2:02:35 of video stuff for my business but just
2:02:38 kind of on the periphery
2:02:40 um there's so many tools out there
2:02:42 there's tools that you describe an app
2:02:44 and it will design an app for you and
2:02:46 put the files in figma there's ones
2:02:48 that'll do that for websites there's
2:02:50 ones that'll take a website and extract
2:02:52 the design out of it and build you a
2:02:54 whole figma design Library there's
2:02:56 things that write songs there's things
2:02:58 that like I can't keep up with the tools
2:03:00 so that's futurepedia.io
2:03:04 is like 4 000 tools most of which are
2:03:07 [ __ ] right but there's probably 400
2:03:09 in there
2:03:10 that are like earth-shatteringly
2:03:13 powerful
2:03:15 um so
2:03:17 that's why I have that there because I
2:03:18 can't because I don't I don't [ __ ]
2:03:20 know
2:03:23 I can't keep my head around chat GPT
2:03:31 see fear officer
2:03:36 does an acorn grow after being cracked
2:03:39 into an oak tree yes it does
2:03:43 that's why I'm here every night just
2:03:45 waiting for Mr Shannon to answer my
2:03:48 question see I answered your question
2:03:50 emilios that's right it's it's all a
2:03:53 master it's a dastardly strategy to get
2:03:55 you all to hang out here longer
2:03:58 I like it all right what are the best
2:04:00 image programs mid-journey I've tried
2:04:02 repeatedly to generate specific images
2:04:05 but it never works okay Angela so
2:04:08 um
2:04:08 mid journey is the one in my opinion
2:04:11 that is the most sophisticated and best
2:04:14 stable diffusion is the one that is the
2:04:17 most flexible and has the most
2:04:19 developers developing things for it but
2:04:22 you need to be geeky to use it
2:04:25 there's a tool called leonardo.ai
2:04:29 that's built on stop and on top of
2:04:31 stable diffusion oh this is a good this
2:04:33 is a good note I got for my tick tock
2:04:38 um
2:04:40 leonardo.ai you have to get on a wait
2:04:42 list but I heard last night from someone
2:04:45 in my comments
2:04:47 if you get on Leonardo's wait list and
2:04:50 go join their Facebook group they'll
2:04:52 give you instant access to Leonardo so
2:04:54 Leonardo's an incredibly well
2:04:58 done website where you that sits on top
2:05:02 of stable diffusion so you don't need to
2:05:04 do all the geeky stuff so go do that
2:05:06 I would also play around with Adobe
2:05:09 Firefly
2:05:10 their generative tools if you've got
2:05:12 access to the Adobe suite play with
2:05:14 Adobe Firefly and then play with
2:05:16 generative fill within Photoshop
2:05:19 because that's really good as well so
2:05:21 those three are the three that I would
2:05:23 play with
2:05:25 you can go to bing.com and just tell it
2:05:28 to create an image but that's built on
2:05:30 top of Dolly which is the weakest of all
2:05:32 the image models so if you want to see
2:05:34 something
2:05:35 generated poorly go there if you want to
2:05:38 see something good mid-journey
2:05:41 or Leonardo
2:05:43 and then it's worth
2:05:46 understanding the Adobe products because
2:05:48 they're commercially safe
2:05:51 John Henry drilled a 14-foot hole into a
2:05:54 rock but the steam engine was only able
2:05:57 to drill nine
2:06:00 that seems like something I should look
2:06:02 up up on chat GPT
2:06:05 you too can learn the history of the
2:06:08 steam engine which had gbt
2:06:10 [Laughter]
2:06:13 that's awesome
2:06:18 is Chachi PT knowledge standardized
2:06:20 across languages
2:06:23 that means other cultures can influence
2:06:26 the West
2:06:27 yes
2:06:28 it knows all the languages it knows all
2:06:30 the computer languages it knows all the
2:06:32 language languages
2:06:33 I translated I did something from
2:06:36 English
2:06:37 what did I do English to Dutch to
2:06:39 Japanese to Portuguese
2:06:41 back to English and I could still read
2:06:43 it
2:06:45 um it it knows languages and and yeah I
2:06:47 agree and that's that's the other thing
2:06:49 Google just announced two weeks ago or
2:06:52 something like that one of their new
2:06:54 real-time translation models
2:06:57 um
2:06:57 is is is gonna do Eleven Hundred
2:07:01 languages I didn't know there were 1100
2:07:03 languages to translate
2:07:05 um
2:07:06 again we're probably within a year maybe
2:07:09 two
2:07:10 of being able to speak in English with
2:07:13 someone else who's a native French
2:07:15 speaker
2:07:16 I'll speak in English they'll hear me in
2:07:19 French in fluent French sync to my lips
2:07:24 they'll respond in French I'll hear them
2:07:26 in English that's coming
2:07:30 it'll start out for like the top eight
2:07:32 languages and then it'll quickly go to a
2:07:35 hundred and then it'll quickly go to a
2:07:37 thousand and then it'll quickly go to
2:07:39 you can speak with people in made-up
2:07:41 languages
2:07:42 I don't know
2:07:44 so yes I think and and I think
2:07:47 your point Angela that
2:07:52 what what the what this tool actually
2:07:54 kind of does it's kind of a mirror that
2:07:56 reflects Humanity
2:07:59 so I think this is going to build
2:08:01 Bridges I think it's going to build
2:08:02 empathetic Bridges and I think it's it's
2:08:05 going to build cultural Bridges
2:08:07 where we learn we're all kind of the
2:08:09 same and this will become the Babel Fish
2:08:13 kind of cultural language
2:08:16 Humanity translator and amplifier
2:08:20 um I think it's got It's it's very much
2:08:22 got that possibility in its in its uh in
2:08:27 its future
2:08:28 you just got up to the first comment
2:08:30 from like an hour ago I know it made no
2:08:32 sense without context yeah I know most
2:08:35 of my words make no sense and have no
2:08:38 context so that's part of the joy of the
2:08:40 of the channel
2:08:44 wait I sat here for two hours for that
2:08:47 uh-huh
2:08:48 hahaha
2:08:51 you did
2:08:54 all right do you think you'll be able to
2:08:57 take a screenplay and produce a film all
2:08:59 on your own at some point yeah and I
2:09:01 don't even think you'll need the
2:09:02 screenplay quite frankly
2:09:04 I did an experiment
2:09:07 what was it with it was just with was it
2:09:09 chat gbt
2:09:11 yeah I did an experiment
2:09:15 where I said
2:09:18 I said take the
2:09:21 take the
2:09:23 structural plot elements of Star Wars
2:09:28 and write me a screenplay outline of
2:09:32 an animated movie
2:09:34 of
2:09:36 animals on the Galapagos Island at the
2:09:39 time of Charles Darwin
2:09:42 um and basically follow the plot of Star
2:09:44 Wars and it was [ __ ] brilliant so
2:09:48 so I think you're not even going to need
2:09:50 to to
2:09:52 start with the screenplay I think you
2:09:54 can say
2:09:56 I want to see something that has the
2:09:59 dramatic Arc of you know Indiana Jones
2:10:02 and the Temple of Doom
2:10:04 the comedic stylings of Caddyshack
2:10:08 starring The Simpsons
2:10:10 with a soundtrack by David Bowie
2:10:14 and
2:10:15 it's going to generate that now I you
2:10:18 know that's that's not going to be for
2:10:20 some time but maybe within five to ten
2:10:22 years
2:10:23 so I think the Stepping Stones to that
2:10:25 are going to be
2:10:27 um
2:10:29 you know take this movie and change the
2:10:31 ending or take take these two movies and
2:10:35 and combine you know I want these actors
2:10:37 in that movie
2:10:39 and I think I think that kind of stuff
2:10:40 is going to be coming soon I'm scared so
2:10:43 scared is good as long as it goes with
2:10:46 curiosity and and uh a sense of
2:10:50 adventure
2:10:52 um don't let the don't let the fear just
2:10:54 have you sit back and not engage
2:10:57 because it's really easy to dismiss this
2:11:00 stuff as either evil and bad and the
2:11:02 robots are gonna kill us or as well it's
2:11:05 not perfect it makes mistakes it didn't
2:11:08 I asked it to give me something and it
2:11:10 sucked
2:11:13 don't dismiss it because of that
2:11:15 this stuff is as bad as it's ever going
2:11:18 to be
2:11:19 and it's [ __ ] remarkable but it's as
2:11:21 bad as it's ever going to be it only
2:11:23 gets dramatically better from here in
2:11:25 terms of its capability
2:11:28 um so so being scared's okay we're we're
2:11:30 we're we're entering a brutal but
2:11:33 beautiful transition
2:11:36 um and and being scared I feel like is
2:11:37 an appropriate response to this looks
2:11:39 like we have a bunch of new people here
2:11:40 my name is Kyle Shannon this the AI
2:11:42 learning lab welcome everyone uh I
2:11:44 ramble for hours and uh that's what this
2:11:48 channel is about apparently
2:11:50 um the URLs on screen right now these
2:11:52 are
2:11:54 the top one is the official chat GPT
2:11:56 website if you haven't gone there please
2:11:57 go there just go there
2:11:59 um and if you don't know what to do
2:12:00 there prompts.chat this thing down here
2:12:04 um is a little primer that will tell you
2:12:06 about what prompting is and how these
2:12:08 things actually work and it'll give you
2:12:09 a whole bunch of ideas of things that
2:12:12 you can tell that to do like act like a
2:12:15 screenwriter act like an accountant act
2:12:17 like a therapist act like a coach
2:12:19 and it will just go off and do that and
2:12:22 and and
2:12:23 you owe it to yourself to use that
2:12:27 long enough for you to have this moment
2:12:31 and once you have that moment then you
2:12:33 should be hooked and and like Albert in
2:12:36 here you can start out and say oh it's
2:12:38 just [ __ ] around with it and now I'm
2:12:39 using it every day to run my business
2:12:42 but yeah
2:12:44 now I'm going to reinvent my business
2:12:45 now I'm going to reinvent myself
2:12:53 all right Deluxe random AI only creates
2:12:57 from the past I
2:13:00 disagree
2:13:04 I mean you could argue that humans
2:13:07 only create from the past
2:13:10 we create from what we have learned
2:13:13 and they can say well we've got
2:13:15 imagination we create original new stuff
2:13:18 if you use these things with
2:13:22 you know any level of depth these things
2:13:25 create
2:13:28 now it might be assimilating what it
2:13:31 does from different places
2:13:33 but by me prompting it to give me
2:13:37 something new I want you to combine this
2:13:39 concept with that concept I want it to
2:13:42 be structured like this
2:13:44 go out it comes that's a new construct
2:13:48 that's never existed before
2:13:53 so did it create it did I create it I
2:13:56 prompted it it's my intention to combine
2:13:59 this with that with that it just
2:14:01 happened to do it like that
2:14:03 and it happened to have the knowledge of
2:14:05 all the past to be able to do that but
2:14:07 it actually created something new so I
2:14:09 actually disagree with that
2:14:11 I think all of what Humanity does to
2:14:13 some degree is based on the past right
2:14:15 it's all language that we created in the
2:14:19 past
2:14:20 all meaning
2:14:21 all of what we do all words
2:14:26 were invented by someone else that's
2:14:29 dead now
2:14:32 says this
2:14:35 and then fans are going to want artists
2:14:37 to sing songs of them that they didn't
2:14:40 make and the fans are going to be able
2:14:43 to make that song that's that that was
2:14:45 my point on that thing that's right
2:14:47 fans are going to want those songs and
2:14:49 fans are going to make those songs and
2:14:51 some of those songs are going to be
2:14:52 better
2:14:53 than the artist's songs
2:14:56 and I guarantee you there's going to be
2:14:58 a hit song created by a fan that sounds
2:15:01 like an artist
2:15:03 that's going to be bigger than the
2:15:05 artist's biggest hit
2:15:07 mark my words I think we'll see that
2:15:09 within two years
2:15:11 and then what do you do do you sue the
2:15:12 fan
2:15:13 because the fan has a bigger hit
2:15:16 that sounds like you than you do
2:15:19 or do you invite the fan out on tour
2:15:22 and give them backstage passes
2:15:24 and encourage other fans to make more I
2:15:27 don't know
2:15:29 brutal but beautiful
2:15:35 brutal but beautiful
2:15:37 that's I love that
2:15:40 hey Kyle what's the first policy a
2:15:42 company should have about chat CPT
2:15:44 um that's a good one
2:15:47 I mean it's a good question I don't have
2:15:49 a good answer well
2:16:01 it's a really good question
2:16:06 I think the first policy is
2:16:16 they should aggressively encourage
2:16:22 everyone in the company to understand
2:16:24 what this stuff makes possible
2:16:28 combined with
2:16:32 understanding the risks
2:16:37 that it
2:16:39 the the risks of using it right so
2:16:43 okay we want you to use chat gbt but
2:16:47 while we're all learning it
2:16:49 you can't
2:16:50 put any client data in it and you can't
2:16:53 put any of our trade secrets in it but
2:16:55 you can put recipes in it you can put
2:16:56 little project management stuff in it
2:16:58 but
2:16:59 right so I think you need to find some
2:17:01 boundaries like until we know until we
2:17:04 have a security and privacy policy
2:17:05 around this stuff and until maybe we've
2:17:08 created our own Standalone server that's
2:17:12 got its own firewall so that we're
2:17:14 confident that our data is safe you
2:17:16 can't do anything but in the meantime
2:17:18 let's learn about this let's learn about
2:17:20 it together
2:17:21 let's have
2:17:23 weekly lunch and learns where someone
2:17:26 who learns something new we did this at
2:17:28 agency.com in the early days we had
2:17:29 these things called geek and tells like
2:17:32 show and tells and what a geek and tell
2:17:34 was was someone would geek out on
2:17:36 something they'd learn something new
2:17:37 with the early you know web technology
2:17:40 and then they go oh I'm having a geek
2:17:42 and tell and people would gather around
2:17:43 their computer and they'd show them how
2:17:45 they did it and then everyone would go
2:17:47 oh that's great they'd run back to their
2:17:48 desks and they'd go do that and they'd
2:17:50 go play with it
2:17:52 um
2:17:54 so so I think I think having both
2:17:57 management and staff in an organization
2:18:01 get AI literate
2:18:03 um aggressively
2:18:06 um and and do it in a way that
2:18:08 encourages individuals to share ideas
2:18:15 and have
2:18:16 management and and the executive team be
2:18:19 open to those ideas because some of
2:18:21 those ideas might look like
2:18:24 a challenge to the current business
2:18:26 model
2:18:27 and if an individual says oh I've got an
2:18:30 idea and they put something out there
2:18:31 that's a challenge to the business and
2:18:33 one of the executives chops their head
2:18:35 off for it
2:18:36 no one will ever
2:18:38 speak up ever again right so that's a
2:18:41 cultural thing
2:18:42 so I think getting aggressive about
2:18:44 learning
2:18:45 but then also
2:18:48 learn enough to understand data and
2:18:50 privacy risks and and make sure to
2:18:52 mitigate those so
2:18:55 I don't know if that's a good answer or
2:18:56 not that's that's where I'd be
2:18:59 Sarah Silverman's suing open AI she is
2:19:02 that's correct social marketing is
2:19:05 evolving thanks to AI a new monitor wait
2:19:08 social marketing is evolving for sure a
2:19:10 new modernization strategy is needed to
2:19:12 stay ahead absolutely
2:19:15 brutal but beautiful
2:19:18 yes imagine an unlimited supply of music
2:19:21 from a deceased artist oh that's coming
2:19:23 that's absolutely coming in fact
2:19:25 here's a business I keep giving business
2:19:28 ideas away take them run with them
2:19:30 start a label that is nothing but
2:19:34 um
2:19:35 new music from artists who've passed or
2:19:38 no longer making music at different eras
2:19:41 of their life young Elvis
2:19:44 Vegas Elvis
2:19:46 bloated Elvis
2:19:50 yeah exactly
2:19:53 the lawsuit claims the three authors did
2:19:55 not consent to use the use of their
2:19:57 copyrighted books yep
2:19:59 again lawsuits are gonna be everywhere I
2:20:03 don't whether or not these it's going to
2:20:05 be really interesting to see where the
2:20:06 sort of fair use and all of the laws
2:20:10 that that emerged out of the hip-hop
2:20:13 fair use stuff
2:20:15 it's going to be interesting to see how
2:20:16 they get leveraged and and how they
2:20:18 evolve
2:20:19 but none of this stuff's getting
2:20:21 uninvented and the the open source
2:20:24 llms are gonna
2:20:26 anything that happens to open AI it's
2:20:28 just going to squeeze through the cracks
2:20:29 and and migrate over to the open source
2:20:32 llm so none of this stuff's stopping
2:20:36 GPT isn't good enough to write scripts
2:20:38 yet
2:20:39 um you can I've gotten it I've gotten it
2:20:42 I've gotten it to absolutely
2:20:46 um
2:20:47 give me good outlines I've gotten it to
2:20:49 give me
2:20:51 dialogue that's probably 60 there but
2:20:56 it's at least a first draft
2:20:58 um non-screens screenplay stuff I can
2:21:01 usually get it to about 80 percent
2:21:03 of finished work
2:21:06 so 60 Cent ain't bad I mean you have to
2:21:09 do a lot of back and forth with it a lot
2:21:11 of collaborating a lot of you can give
2:21:12 it examples here's an example of a scene
2:21:14 you know that I wrote so follow that
2:21:16 tone follow these characters
2:21:20 I reckon business will shift to a more
2:21:22 intellectual property based than content
2:21:26 um possibly yeah I think
2:21:28 I think businesses shift to
2:21:31 what becomes
2:21:33 more and more critical is critical
2:21:36 thinking strategy
2:21:39 um
2:21:40 you know the why why are we doing this
2:21:42 what are we trying to accomplish and
2:21:44 then the tools will do all the Tactical
2:21:46 execution
2:21:49 mixed Tiesto with Elvis songs that would
2:21:51 be cool
2:21:53 first Industrial Revolution destroyed
2:21:55 physical labor now this destroys
2:21:57 creative labor and knowledge work yes
2:21:59 correct but it also will generate entire
2:22:04 new categories of jobs we can't even
2:22:06 imagine today
2:22:07 so again the brutal beautiful part
2:22:10 the brutal is the next two the
2:22:12 disruption in the next two to three
2:22:13 years is going to be brutal
2:22:16 the re-emergence of whatever the new
2:22:19 jobs are and whatever the new Prosperity
2:22:22 is when
2:22:24 our ability to produce accelerates
2:22:27 dramatically without an increased need
2:22:30 and labor
2:22:33 it's brutal initially because that looks
2:22:35 like people that are displaced but
2:22:37 ultimate ultimately this starts to look
2:22:40 like increased prosperity
2:22:42 and and new ways of doing work and new
2:22:45 ways of doing all sorts of things
2:22:52 all right my butt's starting to get numb
2:22:54 that means it's been at least two hours
2:22:56 [Laughter]
2:22:59 so if it's 10 30 here in Denver
2:23:03 10 24. Do You Know Where Your Children
2:23:05 Are listen listen you and me we gotta
2:23:08 talk
2:23:11 what do you do on these Tick-Tock lives
2:23:13 I don't know I just I just talked to my
2:23:14 phone
2:23:16 why do you do that I know this there's
2:23:18 people there do they talk to you sorta
2:23:22 why why do they watch you I don't know
2:23:26 it seems dumb yeah oh it's dumb it's
2:23:28 dumb dumb it's a dum dum
2:23:31 all right carry on all right
2:23:34 let's see
2:23:39 wait
2:23:41 Alibaba Cloud unveils tanji wanzing an
2:23:45 AI image generation model oh cool
2:23:48 that's interesting yeah it's I mean it's
2:23:50 interesting we haven't seen kind of a
2:23:52 major open AI type player come out of
2:23:56 China yet so maybe we'll see like a
2:23:58 mid-journey kind of thing
2:24:00 Alibaba Cloud okay I'll look that up
2:24:02 that's interesting haircut yes thank you
2:24:07 damn we gone
2:24:09 can I write a plug-in and deploy it
2:24:11 internally for private use
2:24:15 well
2:24:16 sort of so
2:24:20 so I don't know about the plugins
2:24:21 themselves I think those have to go into
2:24:23 the plug-in store but
2:24:25 the the GPT 3.5 turbo API now has
2:24:30 function calling
2:24:31 so basically you can describe a set of
2:24:34 functions within your code you can
2:24:35 describe 20 functions
2:24:37 and then the chat bot will dynamically
2:24:40 know which of the functions to call
2:24:42 based on your prompt which is
2:24:44 essentially plugins so yeah look up look
2:24:47 look up function calling and the answer
2:24:49 is yes
2:24:50 Elton
2:24:52 Elton and Harlem
2:24:54 that's a cool name
2:24:57 does AI learn and retain my prompts to
2:25:00 assist other people's future prompts uh
2:25:03 sort of you can tell it not to but you
2:25:05 lose your prompt history so make sure
2:25:06 you download your data
2:25:08 um
2:25:11 I don't think there I I don't think it's
2:25:12 so much the prompts it's the combination
2:25:14 of the prompt and response and like if
2:25:16 you give it thumbs up they're using what
2:25:18 you're doing to train the model
2:25:20 but it's not like they're taking your
2:25:22 prompt and saying hey someone else you
2:25:23 were talking about this want to prompt
2:25:25 someone else wrote it's not like that
2:25:27 it's just more like they're they're
2:25:29 improving the response accuracy based on
2:25:32 the work everyone's doing
2:25:35 is there a way to use an AI tool to find
2:25:36 or remove malware
2:25:39 probably but I don't know I would go to
2:25:41 futurepedia.io and look up security
2:25:43 security and things like that
2:25:46 um
2:25:47 there will be ways to use AI to do that
2:25:49 a lot of the malware detection tools are
2:25:51 probably already using AI don't forget
2:25:53 ai's been around for two or three
2:25:55 decades
2:25:56 um and it's been you know powerfully
2:25:58 integrated into almost everything we do
2:26:00 for the past decade so
2:26:03 um so this stuff is just kind of a a a
2:26:06 layer sitting on top of a lot of those
2:26:09 core technologies that are already there
2:26:13 what plug-in is this I think I was
2:26:15 showing code interpreter which is the
2:26:18 new mode it is a plug-in but it's it's
2:26:20 way more than a plug-in it's a so what
2:26:22 what code interpreter is actually doing
2:26:24 this is freaking wild and I I I'm
2:26:27 wondering if if you're if your chat GPT
2:26:30 chat session is already this but I just
2:26:32 didn't get it before
2:26:34 what code interpreter is doing is when
2:26:36 you launch a code interpreter chat
2:26:38 session it's basically spinning up a
2:26:40 virtual computer
2:26:42 and then when you ask it to do stuff it
2:26:46 writes code and then executes the code
2:26:49 within your virtual private server
2:26:51 executes the code
2:26:54 generate stuff saves it to the virtual
2:26:58 memory or hard drive
2:27:00 and gives you results and so you're
2:27:03 basically interacting with a with a
2:27:05 dynamically generated computer that
2:27:09 dynamically writes and executes code on
2:27:11 your behalf
2:27:13 that's [ __ ] nuts
2:27:16 what in zip file that's crazy so fast
2:27:19 too yeah I think you can have anything
2:27:21 in the zip file
2:27:23 you can have hundreds of five like
2:27:24 that's what I'm going to do I'm going to
2:27:26 put
2:27:27 I'm going to put PDFs of my screenplays
2:27:29 into a zip file
2:27:32 plus I might like I've probably got like
2:27:35 word outlines I probably got all sorts
2:27:38 of things
2:27:39 I'm going to put in a zip file and then
2:27:41 just have it go read what's in the zip
2:27:43 file and then start interacting with it
2:27:45 it's I don't know
2:27:47 so many subscriptions to pay for I've
2:27:49 limited mine too right now I have openai
2:27:52 po mid Journey
2:27:54 Google collab zapier
2:27:58 and I'm probably going to get rid of Poe
2:28:00 because code interpreter I can now do
2:28:02 long form
2:28:04 long form stuff with code interpreter so
2:28:06 I'm going to probably get rid of Poe
2:28:08 those are those are my biggies right now
2:28:12 I don't I don't think you need more than
2:28:14 that
2:28:15 what a what I a I am I using that I just
2:28:18 mentioned them
2:28:20 I have it checked for plot holes that's
2:28:22 a great idea or or plot cliches or
2:28:25 dialogue cliches
2:28:27 I worked with a screenwriter once and
2:28:30 whenever he didn't know what to write he
2:28:32 he used the word look
2:28:34 so it'd be like two people to be talking
2:28:36 and there'd be an awkward pause and then
2:28:37 the one character would go look I want
2:28:40 to tell you this and whenever you use
2:28:42 the word look I knew it was shitty
2:28:44 dialogue because that was his like
2:28:45 crutch
2:28:46 and ever since working with him I now
2:28:50 notice in TV shows and movies all the
2:28:52 time
2:28:53 you can see a character go look and I'm
2:28:56 like ah
2:28:58 that writer didn't know what to write
2:28:59 right there
2:29:01 use code interpreter to do part of the
2:29:03 number crunching that I don't like
2:29:05 leaving more than
2:29:06 more time for what I do yeah yeah
2:29:09 exactly you're talking about screenplays
2:29:11 and code interpreter what do you think
2:29:13 about
2:29:14 personal.ai
2:29:17 I don't know personal.ai do I
2:29:22 um
2:29:31 everyday messaging
2:29:35 supercharge your memory and
2:29:37 relationships with your own AI messenger
2:29:40 I'll check it out remember everything
2:29:42 and connect every oh okay so this is
2:29:44 something that's gonna gonna basically
2:29:47 add memory
2:29:50 of who you are and what you talk about I
2:29:52 think
2:29:53 there's gonna be yeah so
2:29:55 so I think um Sam Walton has said that
2:29:59 that that's going to be something that's
2:30:01 added to open AI is is personal profiles
2:30:05 and things like that I think that's
2:30:07 that's absolutely where we're going I'll
2:30:08 check that out I haven't played with it
2:30:10 but but that's absolutely where we're
2:30:12 going
2:30:14 the ability to
2:30:17 have it have memory of our interactions
2:30:19 and have it trained on all of our data
2:30:23 is
2:30:24 the a big place
2:30:26 how big of a CPU do you need you don't
2:30:29 need one if you're using
2:30:31 chat.openai.com they're using Microsoft
2:30:33 servers like like it costs them a lot of
2:30:36 money every time you do a search it's
2:30:38 like 75 cents it's some ridiculous
2:30:40 amount of money
2:30:42 um if you want to run things locally
2:30:44 then you need you know
2:30:46 a big GPU or you need at least an M1 Mac
2:30:49 ideally an M2 Mac
2:30:52 um
2:30:54 but if you're not if
2:30:57 like wait a year and probably even
2:31:00 before a year
2:31:01 someone is going to create an open
2:31:03 source large language model that's the
2:31:05 equivalent of gpt4 that you can just
2:31:08 install on your laptop or maybe even on
2:31:10 your phone
2:31:11 so
2:31:12 if so right now like to use
2:31:16 chat.openai.com you just need you can do
2:31:18 it on your phone it's all the processing
2:31:21 is being done in the cloud so you don't
2:31:22 need your own if you want to train your
2:31:24 own models that's a whole different
2:31:25 thing
2:31:27 then you've got to rent like a Google
2:31:29 collab with virtual servers and virtual
2:31:32 gpus or you got to have your own iron or
2:31:35 you got to buy one of George hotz's new
2:31:37 tiny boxes which has six gpus in it and
2:31:41 costs 15 grand
2:31:43 so if you want to spend money you can
2:31:46 Sam Altman said he thinks it's going to
2:31:49 take a hundred billion dollars to get to
2:31:51 AGI for openai
2:31:53 so
2:31:55 how can it make money you can make money
2:31:59 commenting on the time delays like
2:32:00 learning to structure chat GPT prompts
2:32:03 character limit in context wow mind
2:32:05 blown honestly
2:32:07 is code generator another subscription
2:32:10 code interpreter it's not it's just part
2:32:12 of GPT Plus
2:32:15 how do you upload you click the plus
2:32:17 button
2:32:20 um is chat CPT
2:32:23 a good name and what would you name it
2:32:25 if you were marketing and branding it
2:32:28 well at this point Chachi BT because it
2:32:32 got to 100 million users so quickly
2:32:36 um
2:32:37 I think it effectively established its
2:32:40 own brand like I think it's
2:32:43 I think it's there there's real problems
2:32:46 with it it's not really a chat bot chat
2:32:49 Bots have a really bad Legacy
2:32:51 of just being bad and then GPT everyone
2:32:55 calls it GTP GPT PPT
2:32:59 so it's horrible in that sense
2:33:03 um
2:33:06 what would I call it
2:33:09 I think the the inflection one pi
2:33:12 pi
2:33:14 is is really interesting something like
2:33:16 mac right like the reason Steve Jobs
2:33:18 called the Macintosh
2:33:20 um Macintosh was because the nickname
2:33:22 was Mac it made it informal it made it
2:33:24 memorable it was three letters
2:33:26 so I think something like that
2:33:29 um something like mac something like pie
2:33:32 something that implies it's your
2:33:34 assistant it's your friend
2:33:37 best friend Eva Eva something like that
2:33:42 um
2:33:44 but
2:33:46 I I
2:33:54 see I don't think chat GPT
2:33:58 becomes like the World Wide Web where
2:34:00 it's just this sort of like Xerox kind
2:34:03 of thing I I think what's going to end
2:34:05 up happening is
2:34:09 the underlying technology that right now
2:34:11 we have to go to chat GPT that all goes
2:34:14 away what it gets replaced with is
2:34:16 Agents like like pie
2:34:20 or Apple's gonna have one you know they
2:34:22 call them memojis right now which is
2:34:24 [ __ ] horrible but but imagine
2:34:28 the equivalent of a memoji that's just
2:34:32 like your agent and it looks like you it
2:34:34 sounds like you it understands you it it
2:34:37 it's it knows your entire history and
2:34:40 then you can just tell it to go do stuff
2:34:43 um I think the underlying technology
2:34:45 people aren't going to talk about it for
2:34:46 long I think what they're going to talk
2:34:48 about are the agents and what the agents
2:34:50 are doing
2:34:51 and then those agents are going to start
2:34:53 virtual
2:34:54 and then you know with Vision Pro you'll
2:34:56 start to have your agent in the world
2:34:58 and then with physical robots you'll
2:35:00 start to have those agents manifested in
2:35:03 some sort of physical form so I think
2:35:05 the underlying technology will pretty
2:35:07 quickly drift to the background
2:35:09 um so chat gbt is okay not great code
2:35:13 interpreters [ __ ] atrocious
2:35:16 foreign
2:35:19 Magnum CI yeah exactly we were calling
2:35:21 it Magnum CI last night I didn't come up
2:35:23 with that but it's a really good one
2:35:25 because it you know it plays into the
2:35:26 Gen X thing we're leaning heavy into Gen
2:35:28 X
2:35:30 all right a few more few more questions
2:35:32 then I gotta go go go go go go go go go
2:35:34 go go go go go go go go go go go go go
2:35:36 whoever finds the dragon may discover
2:35:38 how to tame the dragon and use it for
2:35:40 good exactly
2:35:42 right I just imagine that there's all
2:35:44 the flame was coming out it goes and the
2:35:46 little smoke goes
2:35:49 tamed him or we've killed him yeah
2:35:51 yeah and then back to the Village he can
2:35:53 be our friend he can help us Farm
2:35:55 [Laughter]
2:35:58 all right you won't be replaced by AI
2:36:00 you'll be replaced by someone who uses
2:36:02 AI exactly that's from uh from
2:36:05 Paul rateser of the marketing AI
2:36:07 Institute said that that is correct so
2:36:10 be one of the people that learns ai go
2:36:13 there play learn that [ __ ]
2:36:15 today we celebrate the birthday of
2:36:17 Nikola Tesla I forgot that just imagine
2:36:19 if he had AI holy crap you know what
2:36:21 would be really cool oh
2:36:28 this is something to do
2:36:30 with code interpreter take everything
2:36:33 you can
2:36:34 from Tesla patents any notes you've got
2:36:37 any books that right
2:36:40 throw that all into
2:36:42 a zip file and upload it to code
2:36:45 interpreter
2:36:46 and
2:36:48 let's go discover some of the [ __ ] in
2:36:50 Tesla's patents that haven't been
2:36:53 discovered yet
2:36:56 that would be a fun one
2:36:58 that would be a fun one
2:37:02 because one of the things that these GPT
2:37:04 models are seem to be astounding at like
2:37:06 the speed with which it can parse
2:37:10 a massive spreadsheet of data
2:37:15 is just mind-numbing
2:37:19 so go look at all his data
2:37:24 that would be fascinating
2:37:28 hmm
2:37:32 I wonder if you could even because there
2:37:34 were a bunch of patents of his weren't
2:37:36 there a bunch of patents in his
2:37:37 apartment that the government came in
2:37:39 and took and scurried away
2:37:42 I Wonder If You Could reconstruct some
2:37:45 of his work
2:37:46 through notes and
2:37:49 you know application documents and [ __ ]
2:37:52 like that I I bet there's something
2:37:54 there
2:37:56 do you think we can figure out how to
2:37:58 pick strawberries faster have you seen
2:38:00 the new [ __ ]
2:38:06 laser tractor
2:38:09 this seems [ __ ] insane they've got
2:38:11 this tractor now this this this big sort
2:38:15 of I don't know 15 or 20 foot wide box
2:38:18 that sits underneath the tractor
2:38:20 and it rolls along at like three miles
2:38:22 an hour and lasers it's got cameras that
2:38:26 are identifying the crop from the weeds
2:38:29 and lasers are [ __ ] zapping the weeds
2:38:32 as it rolls along
2:38:34 so yes and then I saw
2:38:37 I saw yesterday
2:38:41 um someone invented this thing it's it's
2:38:44 like a it's like a rack a rolling rack
2:38:49 um that goes in between like down in
2:38:51 apple orchard and then it's got drones
2:38:54 that each of the drones has like a long
2:38:56 pipe with a suction cup on the end and
2:38:59 so the drones sort of fly off this
2:39:01 platform identify the ripe apples fly up
2:39:04 to the Apple suction cup it pick it and
2:39:07 drop it into the bin in the in the cart
2:39:10 so there were like 10 drones picking
2:39:12 apples and dropping the apples into the
2:39:15 cart all coordinated so yes I think
2:39:17 they're going to be picking strawberries
2:39:18 faster
2:39:21 it's it's just gonna I'm telling you man
2:39:24 we
2:39:26 it's this is going to start accelerating
2:39:28 like this kind of remarkable
2:39:30 mind-blowing stuff that that in any
2:39:34 other year of any other decade in any
2:39:37 other century
2:39:39 you would you would [ __ ] stop the
2:39:41 presses and go holy [ __ ] everyone's got
2:39:44 to know about this and now it's
2:39:45 happening like every three days and it's
2:39:48 gonna start happening like every day and
2:39:49 then it's gonna happen like three times
2:39:51 a day
2:39:52 so yes I think Matthew
2:39:56 we're picking more strawberries faster
2:40:00 oh it's flipping amazing it's just
2:40:03 amazing what is a prompt Library so
2:40:05 prompt libraries just if you've got a
2:40:07 prompt base or or if you go to that
2:40:09 document prompts.chat right there that's
2:40:11 a prompt Library
2:40:13 um it's just it's just a collection so
2:40:14 here's the deal
2:40:16 unlike Google what what Google wants is
2:40:19 for you to be as concise as possible
2:40:22 find me the competitors of this company
2:40:24 right they want you to just put in one
2:40:27 short thing that doesn't kind of pollute
2:40:29 the search and then they're going to
2:40:30 give you as relevant results as possible
2:40:33 this generative AI stuff works kind of
2:40:35 the opposite it needs a lot of context
2:40:37 right so you say I want you to act as a
2:40:40 magician and you're a magician that
2:40:43 specializes in stage magic and your job
2:40:47 today is going to be to teach me you
2:40:49 know how to saw a lady in half in a way
2:40:51 that seems relevant to Modern audiences
2:40:54 do you understand yes I understand so
2:40:56 all of the crafting that you do to sort
2:40:59 of wind
2:41:01 chat GPT up to deliver good results
2:41:05 it takes a lot of troubleshooting like
2:41:07 you'll you'll get it you'll think it's
2:41:09 great and then you'll ask it something
2:41:11 and it's like oh no that's not great and
2:41:12 you're like okay I need to know what I
2:41:14 need to do and you'll modify your prompt
2:41:15 and and and the prompts can get quite
2:41:18 complicated and Byzantine and right and
2:41:21 so a prompt library is just to say okay
2:41:22 for the for the you know saw a woman and
2:41:26 half magician dude here's the prompt for
2:41:28 that you throw that in a library and
2:41:30 then someone else can
2:41:31 if they want to do magic stuff they
2:41:33 would go search your library if it were
2:41:35 public and go see it so that's all it is
2:41:43 you're great thank you Mike appreciate
2:41:45 that I use chat EBT to create a quiz to
2:41:49 tell me a per wait tell me a person's
2:41:51 invest and risk tolerance yep oh that's
2:41:55 great
2:41:56 so it created a quiz so so the next step
2:41:59 for that Anthony
2:42:01 take that quiz make that a Google form
2:42:07 then
2:42:09 do a zapier automation where when
2:42:11 someone fills out that form
2:42:14 you send their answers to chatgpt with a
2:42:17 prompt that then responds to them with
2:42:19 you know here's your risk tolerance
2:42:21 report and then you could end it with by
2:42:24 the way if you want to talk more about
2:42:25 this reach out to me and we can talk I
2:42:28 can help you with this
2:42:30 or you know here's a here's a here's the
2:42:33 kind of stuff you might want to invest
2:42:34 in
2:42:37 that would be a fun one to build
2:42:42 you talking about Hunter maybe I don't
2:42:45 know damn RT I'm not sure what you're
2:42:47 talking about
2:42:49 um why do all the AI Channel
2:42:51 why do all the AI channels talk like
2:42:53 their profits
2:42:55 because this
2:42:58 is the most
2:43:00 profound technology in the history of
2:43:03 humanity and so the minute you get your
2:43:07 head around what it's actually doing and
2:43:09 what it makes possible
2:43:12 it's really hard not to go to a place of
2:43:16 this is going to be brutal and beautiful
2:43:19 which was talked about earlier in this
2:43:21 live
2:43:22 um so I don't know that there's a way to
2:43:24 talk about this stuff without talking
2:43:25 about the future
2:43:27 because we don't know what it's going to
2:43:29 be and it could be bad it could be good
2:43:31 it's definitely going to be profound
2:43:33 it's definitely going to be disruptive
2:43:36 so probably because they're paying
2:43:39 attention
2:43:42 I mean what I can promise you is I don't
2:43:44 have any [ __ ] answers I have a lot of
2:43:46 thoughts
2:43:49 based on my experience of having gone
2:43:51 through this with the World Wide Web
2:43:53 going from the early days of that
2:43:54 through the.com Bubble Burst
2:43:57 um
2:43:59 like I've seen these markers before
2:44:04 and what I can tell you is with the
2:44:05 early days of the web
2:44:07 I could look at what that did
2:44:11 and I had a really clear sense of what
2:44:13 the future looked like 20 years out and
2:44:15 it pretty much looked like that
2:44:17 this I don't have a [ __ ] clue
2:44:20 because this is so
2:44:23 staggeringly more powerful than what the
2:44:25 World Wide Web was all the World Wide
2:44:27 Web really did
2:44:29 was allow you to connect this piece of
2:44:31 information to this piece of information
2:44:32 instantly
2:44:34 that's pretty straightforward
2:44:37 this
2:44:39 takes
2:44:41 the collective output of humanity or at
2:44:44 least what was on the world wide web and
2:44:46 and you know ultimately it gets much
2:44:47 broader
2:44:49 and it makes it available as this sort
2:44:52 of
2:44:53 simmering
2:44:56 pot of knowledge you can do anything
2:44:58 with
2:45:00 what does that mean for the future I
2:45:03 don't [ __ ] know but I know it's going
2:45:05 to be big and profound so if
2:45:09 this Channel and other channels like
2:45:11 this sound prophetic probably because
2:45:12 they're paying attention
2:45:16 don't let Lucy die in vain my brother's
2:45:18 writing a book on high Crown sombreros
2:45:21 it helped him research using Chachi BT
2:45:23 Yeehaw that's awesome AI these nuts nice
2:45:27 that that's a that's a solid solid
2:45:30 retort there Dust in the Wind love it
2:45:33 if you're gonna troll here be creative
2:45:35 with it you know ideally your troll is
2:45:37 somewhere between a compliment and a put
2:45:39 down but AIDS nuts is is it's it's it's
2:45:43 class it's a classic it's a classic you
2:45:47 can bring it up almost anywhere it still
2:45:49 works still good solid
2:45:54 I gotta go to bed people what what do
2:45:57 you want with me
2:45:58 I saw that Salesforce is getting AI yeah
2:46:00 they have like 10 tools or something
2:46:02 like that it will build workflows which
2:46:04 sounds awesome
2:46:06 Salesforce needs an AI front end so
2:46:10 [ __ ] bad it is the most painful
2:46:12 torturous piece of [ __ ] that you can't
2:46:14 get rid of it's horrible
2:46:17 and if you could just go
2:46:19 met this guy
2:46:22 stick it in there
2:46:24 and it just automatically listens to
2:46:26 your conversation with him writes the
2:46:28 notes summarizes them pops it in there
2:46:30 does all whatever crap sales force
2:46:32 requires you to do because it you
2:46:34 customized it for your stupid [ __ ]
2:46:37 sales workflow like we have
2:46:40 God damn does it need AI like I don't
2:46:42 want to have to deal with Salesforce is
2:46:45 such the perfect example of where these
2:46:48 tools need to disrupt and destroy the
2:46:51 complexity of the world
2:46:53 it's the perfect example
2:46:56 ugh Salesforce
2:46:59 it needs AI so bad like it can't come
2:47:01 soon enough hey lightning the the new
2:47:04 interface lightning
2:47:06 it's the slowest most horrible piece of
2:47:09 [ __ ]
2:47:11 oh my God does it need AI hey I just put
2:47:14 LegalZoom out of business yeah and a lot
2:47:16 of lawyers AI is an amazing tool I use
2:47:18 for brainstorming projects creating
2:47:20 outline saves hours it does hours days
2:47:24 I'm autistic and honestly getting chat
2:47:26 gbt was so exciting it fact dumps and it
2:47:29 doesn't mind pointed questions it
2:47:31 doesn't Angela and it's it's um
2:47:36 to the communication thing the thing one
2:47:38 of the things that really surprises me
2:47:41 about it but it shouldn't but it's just
2:47:43 because of the nature of our society
2:47:45 right now
2:47:47 because we're so polarized there there
2:47:49 are certain words you literally can't
2:47:51 say right like right now you can't say
2:47:52 trans right because trans just means
2:47:55 right so there's certain words you can't
2:47:58 say because we're so polarized but you
2:48:00 can with chat GPT you can tell you can
2:48:02 say anything and you can ask it anything
2:48:04 and it will respond with Nuance it'll
2:48:07 say well
2:48:08 here's part of the situation here's part
2:48:12 of the situation here's why people might
2:48:14 act like that here's how white people
2:48:16 might feel that way oh okay I can
2:48:18 actually have a conversation with that
2:48:21 to understand like sometimes you just
2:48:23 want to understand
2:48:25 an issue right but we're so polarized
2:48:28 right now you can't even bring up the
2:48:30 issue if you wanted to understand it and
2:48:31 if you Google it all you find is like
2:48:33 the SEO injected misinformation [ __ ]
2:48:36 like you know whoever spent the most is
2:48:38 is you you get their point of view but
2:48:41 with this thing you can just say could
2:48:43 you educate me about this completely
2:48:45 polarizing issue it's like sure
2:48:48 here's a piece of it here's a piece of
2:48:50 it would you like to know more yeah tell
2:48:52 me about this ah well here's the history
2:48:54 of that oh great it's amazing for that
2:48:57 amazing
2:49:01 all right let's see I'm gonna do one
2:49:03 more and I'm gonna go
2:49:08 Bard Bart sucks
2:49:10 that's why Bard's not on this list yet
2:49:12 Bard will get on this list when it stops
2:49:15 sucking ass
2:49:21 all right how can we slow down
2:49:23 globalists from taking over with AI okay
2:49:26 we'll do this as the last one
2:49:29 laughs
2:49:31 let's see
2:49:33 um
2:49:35 I think meta
2:49:37 I think meta has actually you know it's
2:49:40 so funny it's like
2:49:41 Zuckerberg was just like the like the
2:49:44 evil billionaire forever and all of a
2:49:46 sudden in the past
2:49:47 two or three months like he's gonna end
2:49:50 up being the voice of reason
2:49:53 um
2:49:54 I think the the fact that meta released
2:49:57 their llama models to the developer
2:49:59 community
2:50:00 and the fact that there's a vibrant open
2:50:03 source large language model Community
2:50:05 right now
2:50:07 um
2:50:10 means that
2:50:12 the globalists or just you know huge
2:50:16 centers of power whatever you want to
2:50:17 call them
2:50:20 huge centers of power aren't going to be
2:50:23 able to monopolize this thing
2:50:25 um as easily
2:50:27 like if
2:50:29 if chat gbt hadn't been popular if if
2:50:32 Microsoft
2:50:33 said okay we're going to make a billion
2:50:35 dollar investment in open Ai and they
2:50:36 launched chat GPT and no one really gave
2:50:38 a [ __ ] about it because not for nothing
2:50:41 gpt3 was out for like a year and a half
2:50:43 before
2:50:44 you could you could ostensibly do with
2:50:47 gpt3 what Chad gbt was right so it had
2:50:50 been out for a year and no one really
2:50:51 gave a [ __ ] so I mean you know people
2:50:53 that knew about this stuff did
2:50:55 but Chachi BT
2:50:57 getting to 100 million users forced
2:51:01 Google
2:51:03 to reveal what they were doing with AI
2:51:05 forced meta to do it they launched
2:51:09 released llama Into the Wild
2:51:12 um
2:51:13 AWS is in the game you you've got all
2:51:17 all of the major players with the
2:51:19 exception of Apple who's just they're
2:51:21 keeping a little low-key right now
2:51:22 because I think they're going to come
2:51:23 out with something big
2:51:25 um like they do
2:51:28 um
2:51:29 if it had just been kind of open Ai and
2:51:32 Microsoft and then Google and meta and
2:51:35 they all still had their
2:51:37 their AI platforms kind of behind the
2:51:40 firewall and they're like well we're
2:51:41 going to control all these then I think
2:51:44 it would have been much easier for the
2:51:46 the centers of power to monopolize that
2:51:48 and and control it I think the the fact
2:51:52 that meta released their their weights
2:51:55 their large language model the secret
2:51:57 sauce
2:51:58 actually changes the game and and the
2:52:00 fact that chat GPT was so popular so
2:52:03 quickly it means that massive amounts of
2:52:06 money
2:52:07 and massive amounts of engineering
2:52:10 Talent are all aimed at
2:52:13 all this AI stuff right now
2:52:15 so I think the speed of innovation
2:52:20 and the speed of innovation outside of
2:52:22 those controllable companies is is
2:52:24 actually beyond the ability to control
2:52:26 it at least for the short term maybe
2:52:29 long term over 20 years so so the early
2:52:31 days of the internet was very much like
2:52:33 this where there were some big players
2:52:35 but the promise of it was this is open
2:52:38 and it's Democratic and everybody gets
2:52:40 to do anything they want over time it
2:52:42 kind of normalized toward commercial
2:52:44 entities right Facebook Google right
2:52:50 um I think with this AI stuff it's going
2:52:52 to be similar and and maybe over time
2:52:54 you know it centralizes toward the
2:52:58 bigger companies if indeed those bigger
2:53:00 companies have an advantage
2:53:05 by needing massive amount of resources
2:53:09 but what the open source Community is
2:53:11 kind of showing us is that we might not
2:53:13 need massive amounts of resources now
2:53:15 that we've got the core models that we
2:53:17 can play with so I don't know
2:53:20 it's going to be fascinating I got no
2:53:22 flipping clue I don't think we can slow
2:53:24 it down at all though but I don't think
2:53:26 it's the globalists that are pushing the
2:53:28 agenda I think this is it's it's a
2:53:30 remarkable technology that time that's
2:53:32 time has come and chat GPT was the thing
2:53:35 that broke the dam open so with that
2:53:38 that's probably something good to end it
2:53:39 on I really appreciate y'all hanging out
2:53:41 with me this is remarkable that you're
2:53:43 all here and we'll end a list of this
2:53:45 old fart
2:53:47 talk prophetically about the brutal
2:53:51 beautiful
2:53:54 destruction that is coming
2:53:57 transformation that is coming it's not
2:53:59 just destruction
2:54:01 old things die new things are born
2:54:03 that's what's coming
2:54:05 but um I really appreciate all your time
2:54:07 peace out
2:54:09 I got him to ten thousand likes awesome
2:54:11 man thank you until next time see you
2:54:13 Silver Fox I appreciate everybody when
2:54:14 are you back tomorrow night I've been
2:54:16 doing these nightly I do this instead of
2:54:17 sim racing which is which is my my love
2:54:20 but I have uh
2:54:21 uh Sim racing heel right now so
2:54:25 and this is hopefully going to make a
2:54:27 difference so peace everyone