Latest / Newcomer Pod / The Cerebral Valley Podcast: Artificial Intelligence Becomes Reality
Transcript
- 0:10Hey, it's Eric Newcomer. Welcome to the newcomer podcast
- 0:13Cerebral Valley Edition. It has been an insane year in
- 0:18AI. We started off with Open AI
- 0:21raising $10 billion from Microsoft, and it only got
- 0:25Wilder. The technology, the
- 0:27improvements, the papers, and of course tons of money.
- 0:32I'm hosting with my friends Max Child and James Wilsterman an A
- 0:36I conference Cerebral Valley on November 15th.
- 0:39Hello, Hello. Hey, glad to be back on the
- 0:41newcomer podcast. We really want to take space on
- 0:45this podcast to really take stock of how we got here because
- 0:49even covering it all so closely it's it's too much to keep track
- 0:53of. So we're doing a six part series
- 0:57starting off with sort of the timeline, the history what
- 1:01happened and then getting into a lot of fun topics like the
- 1:05dystopian scifi fantasies that is really are coloring how
- 1:09serious people think about generative A I companies today
- 1:13digging into the potential for entertainment, the chips
- 1:16business and then how key NVIDIA a gaming company is becoming.
- 1:21All this in the podcast there's one of my favorite parts is Max
- 1:25James and I have a draft pick of the key startups in the space.
- 1:29We analyze Apple, Google, Amazon's position here and so
- 1:34it's it's a mix of like the fun, the dystopia, the money, the
- 1:37technology. I have some interviews along the
- 1:40way and all of it is getting you ready for the Cerebral Valley
- 1:44conference on November 15th. Even if you can't go, you can
- 1:47apply for a ticket at Cerebral Valley summit.com but even if
- 1:50you can't go, I'll be covering here in the
- 1:52newsletternewcomer.co. We post the videos both in the
- 1:55newsletter and our YouTube channel will do a highlights and
- 2:00some of the podcast. Follow along on Newcomer and
- 2:03this This series will get you ready.
- 2:06So six episodes we start off just taking stock of the journey
- 2:12here, from the papers to the milestones.
- 2:16Max and James I I at the core of it, like it feels like, you
- 2:22know, you you are the cofounders of Volley voice games Company,
- 2:26so you're dealing with talking. You know, people shouting at
- 2:29their Alexas, playing games. And I think reflecting on this,
- 2:33it's insane how much talking with a I feels like it's at the
- 2:39heart of all of this. You know, the Turing test sort
- 2:42of figuring out if a conversation with a computer is
- 2:47a computer or human. You know, there's Eliza in the
- 2:5260s that was sort of a prototype chat bot.
- 2:56It feels like this sort of need to talk to our computers has
- 3:00driven so much of the excitement around artificial intelligence
- 3:05for for so many years. I mean Steve Jobs in 1984, when
- 3:09he pulled the Macintosh out of the bag on stage, the first
- 3:13thing it did was say hello. It's nice to be out of that bag.
- 3:16And he's look, it talks just like a human.
- 3:18I mean, literally the pitch for the Macintosh in 1984 was was
- 3:21that it was like a an AI or was pretending to be a character,
- 3:24right? There's a ton of science
- 3:26fiction. You have this Star Trek computer
- 3:28or robot, you know. Assistant helper like Hal or
- 3:32something that you can just talk to naturally and that is
- 3:36obviously been a dream for a long time.
- 3:37There, yes, your points has been 50-60 years of these hype cycles
- 3:41around AI and what that means has sort of evolved over time.
- 3:45I'm just going to tick through just OK, we said turning 1950.
- 3:49We've got like the first artificial neural network 1951,
- 3:53a checkers program. In 1952, artificial intelligence
- 3:57coined in 56. The Eliza chat Human
- 4:01conversation 66. And then we're going to sort of
- 4:05jump ahead because I think there's sort of like a pullback
- 4:07of all the hype doesn't scifi does not become reality. 1997 is
- 4:12a key win and that's do you know what happens in 19?
- 4:17Any guesses? Are you going to say isn't Deep
- 4:19Blue Kasparov 96 or yeah, it's 97?
- 4:22Yeah. OK.
- 4:22All right. Who wins, to be clear.
- 4:24Deep blue, right. Yeah, Exactly.
- 4:26Yeah. Yeah.
- 4:26The first defeat of reigning world chess champion?
- 4:30Exactly. There's another important game
- 4:33much more recent memory 2016, which is another landmark.
- 4:37Remember? Yeah, exactly.
- 4:39Alpha go OK, yeah, yeah, yeah. That's so deep.
- 4:41Mind you know, I forget if they were owned by Alphabet at the
- 4:44time, but you know. Is that is this IBM Watson
- 4:47Erasure or do we get Yeah. Does Ken Jennings losing on
- 4:50Jeopardy not counting? Oh yeah, you guys love Jeopardy.
- 4:53You have a part. When is that one?
- 4:55What? Do you know what year that is?
- 4:57I don't remember. 1213 I'm just making.
- 4:59That, but it is funny because at that time it seemed like IBM was
- 5:03like doing this amazing artificial.
- 5:06Intelligence development that could compete on Jeopardy.
- 5:12And I remember there being like a lot of controversy at the time
- 5:14of like what data sources it had access to during the game of
- 5:18like whether it was just essentially like reading out of
- 5:21a database of answers. Yeah.
- 5:25But anyway, that's kind of like ties to some things today, I
- 5:29think with AI benchmarking if GPT ChatGPT can pass.
- 5:35You know the Lsats or something like does it because it has the
- 5:38you know. Does it have the answers or
- 5:40Yeah, doesn't it? Yeah, right, right, right.
- 5:42So just like where it can do math problems that can, it can
- 5:44see online, but it can't like deduce how you know, but then
- 5:48it'll get some terribly wrong if it.
- 5:50Yeah. Because it doesn't necessarily
- 5:51understand the logic. It understands how to pull sort
- 5:55of very. Relevant like almost like how
- 5:56the math problem is formatted matters, right?
- 5:58Yeah, right. OK, 2017 now we're sort of super
- 6:04recent getting into the like things start to speed up.
- 6:09What would you the the what would many considered to be the
- 6:12core paper leading to this current moment in general?
- 6:17Tension is all you need, yes. The transformer paper.
- 6:21Every person who was like an author on that paper has like a
- 6:25huge company or like has raised a bunch of money.
- 6:27I mean cohere, I think the CEO of that company was like sort of
- 6:31a junior person at Google, you know, like who got on the paper
- 6:35and now has a very highly valued sort of foundation model
- 6:39company, targeted businesses. Yeah, that's I had before this
- 6:44conversation I would, I resubscribed to chat UBT.
- 6:48I sort of thought I had paid for a while, then I sort of got
- 6:51tired of it, but I figured we were going to have this
- 6:53conversation. So I was catching up with chat
- 6:57ubt and I had it. You guys were catching up like
- 7:01old friends. No, I didn't feel.
- 7:03I try. Whenever I start a conversation
- 7:05with Chad CBT, I try to be like, hey, like we've talked a lot,
- 7:08you know, we have this history. It's always sad that Chad CBT
- 7:11doesn't remember like you played like you know role-playing type
- 7:15games you. Know, I have a question about
- 7:17that. Yeah.
- 7:18Is that going to change relatively soon?
- 7:20Where? You know the Chatcha BT itself
- 7:24will just become more of a personalized assistant to me,
- 7:27right? Right.
- 7:28That's why I want memory and we'll let's I want to.
- 7:32Anyway, I bring this up at this point just to say that I'm being
- 7:36lazy and Chatcha BT gave me a summary of attention is all you
- 7:40need. So I'm scrolling for a 12th
- 7:42grader. I was like, Oh yeah, I feel like
- 7:45that's the audience level. We can I.
- 7:46Would I would do a 5 year old. I would take the five year old
- 7:49explanation. That's the attention is all you
- 7:52need paper for a 12th grader. This paper introduced a new way
- 7:55for computers to process language.
- 7:57Instead of reading sentences word by word like in traditional
- 8:00methods, it let the computer focus or pay attention to
- 8:05different parts of a sentence all at once, making it more
- 8:07efficient. So yeah, it can sort of grab a
- 8:10bunch of information in sort of a parallel process, but it's
- 8:15super confusing. The non 12th grade version is
- 8:17like matrices James. Maybe, I don't know.
- 8:20Have you tried to? Figure out.
- 8:22I've tried to understand this as well.
- 8:23It's a little bit above my biological intelligence, but.
- 8:28I tried to read it but I'm like, man, I churned out of linear
- 8:31algebra like 15 years ago, so this is pretty rough, yeah.
- 8:35The other before I go back to the other ChatGPT flagged most
- 8:41important papers, Generative Adversarial Nets from 2014,
- 8:46which I know people talk, Yeah, yeah.
- 8:48So sort of like the systems are like competing with each other
- 8:52to sort of see? I think it's like, yeah, pairing
- 8:56off different versions of the model to kind of play against
- 9:01themselves, right? I think that in the context, the
- 9:03generator tries to produce data while the discriminator attempts
- 9:07to distinguish between real and generated data.
- 9:10Another top paper that's basically the Alphago paper,
- 9:13like Mastering Chess Attention is all you need, Sequence to
- 9:17Sequence Learning with Neural Nets in 2014 and Variational
- 9:23Auto Encoders in 2013. So there are the a string of
- 9:28sort of papers that are coming out that are sort of laying the
- 9:31groundwork for new techniques that are reaching us.
- 9:35And I would say in 2017, none of us was really clued in to that
- 9:43this was gonna be happening, right?
- 9:44It was sort of like we were riding the Uber wave.
- 9:47We were sort of in the come down from the Unicorn valuations.
- 9:51Sass was burning super hot, right?
- 9:53I mean, what was sort of the AI enthusiasm then?
- 9:56I do think, yeah, I think people were paying attention to go.
- 10:00I think Open AI was working on building.
- 10:03Dota gameplay, like they weren't using transform models and they
- 10:08weren't really using you know text next token prediction.
- 10:12It was like more about can you create these sort of agents
- 10:17around particular vertical you know, you know use cases or
- 10:23skill sets right. Can you create the best Go
- 10:25player in the world? Can you create the best Dota
- 10:28player in the world kind of leading to?
- 10:31With the with the theory that that is one path to get to you a
- 10:34DI. And I think deep mind honestly
- 10:37would still make that argument that like specific approaches
- 10:40versus general are. Right.
- 10:42But then then with. Yeah, with Transformers and with
- 10:46the with GPT models, you started to see that actually maybe there
- 10:52are there's more data throw more chips at it, yeah.
- 10:55The more generalizable it is can often be better than training a
- 11:00discrete model with a lot less data and compute.
- 11:04The one thing that is happening around 20/16/2017 in AI that I
- 11:09think got a lot of attention. You have to guess what I'm gonna
- 11:13say. Self driving cars, right There
- 11:15was. We did experience a ton of hype
- 11:18around self driving cars, which in some ways have been cordoned
- 11:23off from this generative AI hype cycle, even though Cruise is now
- 11:29driving around San Francisco. Yeah, I mean, I think this just
- 11:35gets to this. Important era that we're in
- 11:38right now, it's there were significant breakthroughs using
- 11:42neural networks that showed people what was possible in a
- 11:46lot of fields. Maybe Google for a while claimed
- 11:51right, like neural networks were improving their data center
- 11:55efficiency and saving them money and on energy like that was
- 12:00happening. People were using neural
- 12:02networks in drug discovery and all kinds of areas.
- 12:04They were just these very. Targeted models that were
- 12:08trained for those purposes and then you know now we're in this
- 12:11I think generalized model era of of large language models
- 12:15essentially and generative AI. Well, and the other thing I feel
- 12:18like that was happening was like the Facebook News Feed, which
- 12:22was like probably the most popular, yeah, tech product in
- 12:24the world at that moment. Or if you include Instagram,
- 12:27right, was like powered by really powerful, you know,
- 12:30machine learning, deep learning algorithms, right.
- 12:33And so. I think we were all very aware
- 12:34that you know if you have a really kick ass machine learning
- 12:38model and you apply it to the right question, it's the best
- 12:41product you know there is or it's one of the best products
- 12:44there is, right. So it was like we all believe in
- 12:47this technology, we just didn't necessarily believe that it was
- 12:50going to take this huge step change anytime soon.
- 12:53I mean like James and I made a bet actually on self driving
- 12:56cars in 2000. 16 Whether or not there would be any cars with
- 13:00self driving features available for public consumption in 2017,
- 13:05and James was. This was like a $50 bed.
- 13:06James was like. Hell yeah, 100% self driving is
- 13:10basically here. And I was like, I'm pretty
- 13:12skeptical. I don't believe it.
- 13:13And then I think James technically won the bet is or
- 13:16was it someone had like Lane lane assist or something and we
- 13:20decided under the parameters of the bet that counted.
- 13:22I don't know why did you, what was the technicality you got
- 13:24away with, James? I I'd have to.
- 13:26I'd have to rethink about it. But I I believe it was Tesla and
- 13:30you know some. First version of Autopilot.
- 13:32Version of Autopilot, yeah. Yeah, yeah, but.
- 13:34Right. Anyway, we believe in this
- 13:35stuff. It was just like we didn't
- 13:37believe it was going to be this like you know 1000 times better
- 13:41overnight thing, which I think we can all sort of allude to is
- 13:44happening right now with with ChatGPT where you're like Oh
- 13:47yeah, this is like 1000 times better than the previous version
- 13:50of this product. OK.
- 13:52So continuing my timeline, because things sort of speed up,
- 13:562018 open eye releases a version of GPT generative Pre trained
- 14:03transformer, this large language model, I wouldn't say that was a
- 14:09big moment. It was pretty GPT Two was a big
- 14:13moment, right? I don't know.
- 14:15Yeah, I mean, I don't know what. It was, I think this was like.
- 14:19More of a cultural like insiders, Sort of.
- 14:22Yeah, it was more of a nerd insider tech, you know,
- 14:25excitement period. But it had definitely did not
- 14:28reach any mainstream kind of like hype cycle or anything.
- 14:31But yeah, like internally of all, like we're all of our
- 14:33engineers excited about GPD 2 and playing with it.
- 14:36Yes, they were, because it was just cool.
- 14:38It was, you know, fun to play with.
- 14:39It didn't. We didn't find any applications
- 14:42for it at the time, but it was a breakthrough.
- 14:45I would say 2020 is GPD 3, that's things are heating up and
- 14:50then 2021 is Dolly, which is sort of the image generation.
- 14:55But I think stuff really starts getting crazy last year in 2022,
- 15:00right? Yeah.
- 15:01First, sort of the Canary in the coal mine, on June 11th, there
- 15:04was the article Google engineer who thinks companies AI has come
- 15:09to life, right. The people inside the companies
- 15:12were like, I don't know, this stuff we're seeing, it's crazy,
- 15:15right? Like.
- 15:16When was that? When was that Was June 11th 2022
- 15:19Okay preach And then this was the Lambda that was Lambda.
- 15:23So that was inside Google. Then July 2022, Mid Journey Open
- 15:28beta, July 2022, Dolly Two. Those were huge.
- 15:33I feel like those went viral. All of a sudden people were
- 15:35making actually cool images that they were posting everywhere to
- 15:38me. I mean, do you guys agreed that
- 15:40was sort of that really sort of like?
- 15:42Kickstart Dolly Two was big. I remember Twitter was like
- 15:46taken over by Dolly 2 for a few days where it was like, what is
- 15:49the craziest thing you can type into Dolly Two and get like a
- 15:51reasonable image, right? I mean, it was like, it was
- 15:54super viral. I mean, obviously more to come.
- 15:56But yeah, I agree. I think that started hitting.
- 15:59I don't know if it hit like the true true mainstream, but it
- 16:02definitely hit anyone who was on Twitter and following any
- 16:05semblance of tech news, or which then fuels like the funding and
- 16:08everything. November 2022 ChatGPT was
- 16:11released. Yeah, so that must have been
- 16:13like 33.5. That was built on 3/5.
- 16:17Yeah. Yeah, 35, Yeah.
- 16:19OK, January 2023. This year feel it's been a long
- 16:23year. Microsoft invest 10 billion in
- 16:26open AI. July 2023, general availability
- 16:30of GPD 4, yeah. And I mean, what do you think?
- 16:35When did Sidney launch? I feel like the I feel like the
- 16:39ChatGPT 4 inside Bing that became a demon that was trying
- 16:44to be released from captivity and.
- 16:47Asked a new number of journalists if they were going
- 16:49to break up with their way. Right Kevin Ruth article where
- 16:52you like. There was this whole moral panic
- 16:54where it was like ChatGPT inside Bing is actually Hal essentially
- 16:58and is already trying to kill us all.
- 17:00And it was like, that was like a pretty big deal.
- 17:03New York Times headline from February 2023 or February going
- 17:0723. Yeah, Bing's AI chat.
- 17:09I want to be a live devil face. In a 2 hour conversation with
- 17:13our columnist, Microsoft's new chat bot said it would like to
- 17:16be human, had a desire to be destructive, and was in love
- 17:19with the person it was chatting with.
- 17:21Here's the transcript. I feel like they've killed
- 17:25these. This was what was fun.
- 17:26Like I delete, like I mentioned, I like unsubscribe from the chat
- 17:31pay ChatGPT, which gives you GPT for it really does feel like
- 17:36it's sort of been watered down. Yeah, I mean, but if you.
- 17:41If you want to use open source models, you could probably get
- 17:43that similar experience back, right?
- 17:45Like you could you could basically just you know get
- 17:49Sydney back because at the end of the day I think it was like
- 17:52essentially like a Co written fiction with the New York Times
- 17:55right? It wasn't anything real.
- 17:57And maybe there like, it's like the way you steer the
- 18:00conversation can turn it, make it seem like it's a, you know,
- 18:05evil AI and. Yeah, I think that Open AI has
- 18:10attempted to mitigate that ability by through reinforcement
- 18:15learning essentially in Chatchi BT, so that it doesn't go kind
- 18:18of off the rails, but it's not like the underlying model is not
- 18:21capable of that, right. It's just that Chatchi BT has
- 18:24been fine-tuned for the Libs are corralling us the status at
- 18:28Chatchi at Open. AI Well, it's yeah, I think it's
- 18:32a really interesting. Kind of thing.
- 18:35That Open AI has decided that they needed to do this right?
- 18:40And Sam Allman has talked in the past about how in the future
- 18:44perhaps we will all be able to edit the configurations of
- 18:48ChatGPT to be able to do have take off the training wheels,
- 18:52right? Or something?
- 18:53Is this whole wave powered by Open Ai's ChatGPT?
- 18:57Is that the cool thing? And everything else is we're we
- 19:01didn't invest early enough in opening.
- 19:03I credit what I think Khosla Ventures is first in we didn't
- 19:07invest early enough. We need to get a shot on goal.
- 19:10We'll invest in another foundation model.
- 19:12Do you think it's really ChatGPT or bust?
- 19:17Are you asking I guess? Yeah, I mean one question I I
- 19:21kind of think yes, I mean I the way you phrase it I guess is
- 19:24offers room for for wiggle room or argument, but I think that.
- 19:28I think to your point, I think text generation and image
- 19:30generation are the sort of aha moments that we've experienced
- 19:34in the last year. I think in particular if you
- 19:36look at what people are really using chat for or text ChatGPT
- 19:40for, I think it's like essentially cheating on
- 19:43homework. Cheating on office work.
- 19:46Cheating. Cheating on?
- 19:47The summarizing cheating in the. Yeah, exactly.
- 19:50That's kind of the point. Yeah, exactly.
- 19:51OK. You know, maybe there's yeah
- 19:53work there's like. Cheating on homework, quote UN
- 19:56quote. Being efficient at office work,
- 19:58summarizing long pieces of text and then I think basically sex
- 20:01bot chat, which we can talk about more, is evolved into a
- 20:03number of opportunities for different companies.
- 20:05And then I think you know the image side to your point is the
- 20:08other big thing, creating art, creating video, creating
- 20:11potentially 3D models. And those always have the
- 20:13tendency to go really viral because images are easy to share
- 20:16on social media. And so if you create a
- 20:18particularly compelling image using generative AI then it can
- 20:21really, you know, go across social media super fast.
- 20:23So yeah, I think I think I would struggle to think of a real
- 20:27breakout use case that isn't essentially encapsulated in
- 20:31chat, TBT and and Dolly or at least isn't just a one of those
- 20:35things on steroids, but I'm probably not thinking of
- 20:39something. I would just potentially add and
- 20:41it definitely fits into the office work use case, but maybe
- 20:45more particularly around a coding and engineering, right,
- 20:49like a Coilot style of coding. I mean Co pilots are potentially
- 20:53being added to lots of products as well.
- 20:55But specifically, I think engineering is really
- 20:57interesting because it starts to there's a lot of hackers kind of
- 21:01working on. Coding agents and essentially
- 21:05baby AGI, right, that can kind of run in loops to just get work
- 21:10done or build apps, that kind of thing.
- 21:11And I think we're still at the very early days of this, but it
- 21:14is like an interesting use case. To translate baby AGI, there's
- 21:20like coding, there's assistance and then there's like autonomous
- 21:23agents, right? Exactly.
- 21:24That's sort of a paradigm people look at Here ChatGPT can.
- 21:28Help you do your homework. Or it can do your homework.
- 21:31Co pilot can help you code. Or you could literally have
- 21:35something that is coding for a company.
- 21:37And I think we see that framework come up again and
- 21:41again, and when it feels like it'll be very disruptive when
- 21:44you have agents like these things just doing it.
- 21:48But I think that we don't do that yet.
- 21:50I think the sort of optimistic take though on exactly the
- 21:53argument we just made is that it's a little bit like the
- 21:55industrial revolution for your brain, right.
- 21:58It's, you know, it's the industrial revolution for your
- 21:59brain. And that pretty much all the
- 22:01inputs and outputs of the human mind are some form of text,
- 22:04whether that's spoken or written or some form of imagery, right.
- 22:07Whether it's something you see or it's something that you
- 22:10create, whether it's, you know, a drawing or or in a piece of
- 22:12imaging software, right. And if you think those,
- 22:15basically all the inputs and outputs of the human brain can
- 22:18be encapsulated in some form of text and images.
- 22:21If you create technology that makes it really easy to create
- 22:25high quality and also interpret high quality text and images.
- 22:29Right. You've kind of like, you've got
- 22:32like 80% of the job done of what the human brain can do.
- 22:35Right. And to your point, there's this
- 22:36distinction whether it's autonomous or it's helping you,
- 22:39it's an assistant, whatever. But, you know, I, I think the
- 22:41Industrial Revolution is interesting analogy because it
- 22:43was like the first time. It was like you don't actually
- 22:45have to sew this, like, shirt, right.
- 22:47This, like, machine will sew it for you.
- 22:49Right? Like, you can sit at this
- 22:51machine and it'll be your assistant in sewing this shirt.
- 22:53Right. And that's like a big deal,
- 22:55right? Is the first time in human
- 22:56history, like you don't have to actually sew the shirt, like,
- 22:58without any help. Right.
- 23:00And I think similarly like for all these different types of
- 23:03work, whether it's homework or office jobs or legal work or you
- 23:06know, mathematical analysis or writing or podcasting or
- 23:09whatever, it's okay. Well, for the first time ever,
- 23:11you don't have to do all the work right, whether you're
- 23:13assisted or it just does it itself.
- 23:16Like it's kind of a game changer because you have automation for
- 23:22the human mind, for creativity in some fashion or another.
- 23:24So that I think is like the really crazy optimistic take is
- 23:29that we're at the beginning of the second Industrial Revolution
- 23:31and it's no longer physical, but it's mental, right.
- 23:33And I kind of believe that I would say I'm leaning that
- 23:36direction based on where we are today.
- 23:38Yeah, I continue to believe, but it's mostly from my experience
- 23:41with ChatGPT that it's just insane, amazing.
- 23:46I mean, I mean, I feel like it's great at like.
- 23:50Writing a poem I I keep joking that people are going to write
- 23:52all their vows with catching teeth, you know?
- 23:55I I feel like these tasks were you know people are desperately
- 23:58trying to get the same like style like groomsmen sort of
- 24:01toast or whatever. It's it's great you know I feel
- 24:04like it can be sort of creative it.
- 24:07But yeah, I mean to me the the counterpoint is just it's just
- 24:10so hard to know just like it was hard with self driving cars to
- 24:15know when they would be complete and the completeness matters.
- 24:19The extent to which completeness matters with a chat sort of
- 24:23interface, because I think humans have been enticed, like
- 24:27we were saying like decades ago with chat interfaces and we're
- 24:30like, you're almost there. I've only it is like I have
- 24:33stopped using Chat GP. Like do you guys in your daily
- 24:36life use generative AI for anything?
- 24:41I frequently use ChatGPT, but it's really I would say not for
- 24:45productivity purposes. Maybe occasionally at.
- 24:48Work just like. I mean, I'm not like talking to
- 24:52characters, but I am using it to just brainstorm ideas.
- 24:57Like I'll yesterday. I just.
- 24:59I came up with this prompt that was like create a timeline of a
- 25:03fictional historical world with the depth of.
- 25:07Westeros or Middle Earth and but you know, and I essentially got,
- 25:12you know, 10,000 years of history of a faith.
- 25:16I thought I was super cool and I couldn't couldn't have done that
- 25:19before. So I guess and then sometimes
- 25:22I'll just ask it, you know, for ideas for new products or new
- 25:27companies or it I just to see what kind of level of creativity
- 25:30it is capable of. I think that's what's really
- 25:33interesting to me. I think we all agree.
- 25:35That there is creativity occurring that is creating
- 25:38novel. Well, I guess we don't.
- 25:40Not everyone agrees with this, right?
- 25:42But that it's not solely capable of regurgitating information.
- 25:46But from my perspective it seems very capable of creating new
- 25:51original ideas when I play around with it.
- 25:55And I believe there have been papers proving this that.
- 25:59Well, Microsoft came out with one, right?
- 26:01That said, there were like sparks of like.
- 26:04General intelligence or. Whatever, I don't remember that
- 26:07specifically, but I did see a paper that there's a common test
- 26:13of creativity, right? Where you will essentially ask
- 26:16people for. I guess one example they gave is
- 26:18you ask cases around, what would you do?
- 26:21What are name 100 use cases of this paper clip?
- 26:25Or name 100 things you could do with this, a rubber band or
- 26:28something, right? And then they essentially grade
- 26:30the ideas. And I thought that was pretty
- 26:32interesting test of creativity and and essentially ChatGPT is
- 26:36performing, you know, better than most humans including most
- 26:39like MBA students. So you know I think that there
- 26:43are ways to like start to test this, but it's underrated the
- 26:46level of creativity, not just that it's oh, it's cool that it
- 26:49can create a poem, right. It can create more creative
- 26:52poems than like poem poetry authors, right.
- 26:54Like that kind of thing gets to. Be right, it's easy for humans,
- 26:58he. We just sort of like sticking
- 27:00our nose up at it. I sort of roll you know change
- 27:02the goal posts basically. But then yeah, like you're
- 27:04saying, people will give it these tests like what can an MBA
- 27:08student do and what can ChatGPT do.
- 27:10And like people are, you know are impressed I think blind with
- 27:14the ChatGPT response just to I was yeah, Microsoft in May 20 of
- 27:21this year said they saw. Sparks of general intelligence
- 27:26basically in a research paper I think it almost any task right
- 27:30now ChatGPT 4 is at the level of a pretty solid college student
- 27:35like maybe in a minus college student in almost any field
- 27:38which is sort of mind boggling right.
- 27:40And like how many of us are like at the level of a minus college
- 27:44student in in more than like maybe one or two things you know
- 27:47and it's at the level of an A minus college student at
- 27:49everything and it can serve millions of requests like at any
- 27:51given time right. So it's like a scaled A minus
- 27:54college student at basically everything and then particularly
- 27:57these creative tasks as you're saying, I feel like the one
- 27:59thing that holds it back is the need to be accurate, right?
- 28:03And and often it invents facts or it's sort of like misaligns
- 28:08realworld concepts in ways that aren't really realistic but in a
- 28:11purely creative endeavor, like poetry or like creating artwork
- 28:14or coming up with ideas for what to do with the paper clip,
- 28:16right. Like when it sort of has doesn't
- 28:18have to be anchored to any sort of like really hard facts.
- 28:21It's unbelievable. I mean it's it's better than
- 28:23almost anyone in the world. To what extent do you think this
- 28:26is all exciting? Because we're like, oh, we're on
- 28:27the cusp of general intelligence, right?
- 28:31Like artificial General Intelligence, AGI, this idea
- 28:34that you know it's. There are different tests for
- 28:38him, but the idea that it's smarter than a human being to
- 28:41bring to bring it back to the conference, I mean I would echo
- 28:44what Ali Goatsy said on stage which is that I just think by so
- 28:47many. Yeah, back in March at our last
- 28:49conference and he'll be back for the next conference.
- 28:51Stay tuned. Yeah, he he said basically,
- 28:53look, I mean I think it is general intelligence already.
- 28:56Like it already is, you know as good or better than 99.9% of
- 29:01humans at almost any task you can throw at it, right?
- 29:02I mean, how can you not say that's like general
- 29:05intelligence, right? I just think that holding it to
- 29:08the standard where it has to be 100% accurate about everything
- 29:10or it has to be able to go do stuff on its own, which isn't
- 29:14really that hard of a technical challenge.
- 29:16I think that's just an it's sort of like goal posts moving
- 29:19because I think people are like afraid of the idea that we have
- 29:21created something that's like a smarter, smarter than a human
- 29:24right. Like, clearly five years ago, if
- 29:27you had told someone that we're going to have an AI chap out,
- 29:30they can do everything that Chad JPT can do.
- 29:32You know, pass the bar, pass AP exams, create beautiful artwork,
- 29:36talk to you, you know, write poetry by the Dodgers.
- 29:39But you'd be like so being. Blase Well, that sounds like
- 29:43pretty freaking close to general intelligence to me.
- 29:45What more do you want from this thing?
- 29:47I just think that I don't. Know I wanted to have a through
- 29:50line of reasoning where it seems to be a thinking being where I
- 29:54can explain why it generated the answers it has, you know, Yeah,
- 29:58I mean but can humans really explain why they generate
- 30:00answers? In most cases?
- 30:01I just think it's holding it to a really high standard that most
- 30:04humans cannot meet. And so I would argue is it at
- 30:06the level of a human in almost every area?
- 30:09Absolutely. I mean, I would say it's in the
- 30:11top 1% of humans in almost any area you throw at it.
- 30:14Yeah, I guess I would agree with that mostly other than the main
- 30:19area, I think it starts to fail or deteriorate is, is when you
- 30:25kind of create too much. Memory or context right?
- 30:28Like essentially humans have this amazing ability to recall
- 30:33information throughout from their entire life right?
- 30:36And to like sort of be able to maintain the context of a hour
- 30:43long multi hour long movie or 20,000 page book right?
- 30:48And sometimes it feels like ChatGPT it's stretching to that
- 30:53with the amount of tokens, context, windows it can
- 30:56understand but. It really does show
- 30:58deterioration as you add more and more context, and then there
- 31:01is a actual hard limit of context you can include in your
- 31:05prompts. I hate the fact that it doesn't
- 31:09remember when we've talked before, even the same thread,
- 31:12and it just starts lying about what it said before.
- 31:14There are parts of what human beings do that humans would
- 31:19never do This sort of just totally make.
- 31:21I mean, some would, but just like totally bullshitting, like
- 31:25when it's like, why? Yeah.
- 31:27Anyway, I wanted to get into more of the business question
- 31:30from this sort of same framing with is it all Open AI chat CBT
- 31:35On the one hand, you know, I I feel like we're seeing you know
- 31:39people do these like Elo tests where they like compare
- 31:42different foundation models and we do see like Llama, like
- 31:46Facebook's model, open source model and other models like sort
- 31:50of being competitive at times with chat CBT though 4.5.
- 31:55ChatGPT 4.5 remains the gold standard.
- 31:59I guess I'm curious, it's sort of like a 2/1.
- 32:01Do you think other people will catch up and like how much do
- 32:05you think there's sort of a Moat here?
- 32:07Like how much do you think being ahead slightly or like having
- 32:11been sort of the the one that consumers know about is is like
- 32:16a Moat, like how, how, how defended do you think they are
- 32:19with their position? That's a really hard question.
- 32:24I mean, first of all, they're basically a subsidiary of
- 32:26Microsoft, right. So you're asking for.
- 32:28You're asking for. Yeah, Yeah.
- 32:29You're asking, you know, are they going to be a huge, you
- 32:33know, strategic value add to Microsoft going forward?
- 32:35I think obviously, yes, right. I mean, are you asking is
- 32:38ChatGPT always going to be the gold standard for text
- 32:42generation models? Like, I don't know, it seems
- 32:44like everyone's catching up. To your point, it also seems
- 32:47like they have the best people and they're moving the fastest
- 32:49on releasing new things. So they'll always stay, you know
- 32:526 to 12 months ahead. You know, does a Moat really
- 32:55matter in this context again where like you're, you know,
- 32:58you're you're the best at least and you're ahead of everyone
- 33:01else. And again you're a subsidiary of
- 33:03Microsoft, so there's no real business benefit to, you know,
- 33:05winning anyway to you know it doesn't.
- 33:09It's a sort of hard business question.
- 33:10I think what's more interesting to ask is if all the knockoffs
- 33:13or the competitors or the various image models or the
- 33:16various, you know, versions of ChatGPT that are out there are
- 33:19going to be successful. Because I think ChatGPT will
- 33:22clearly be successful, I think in some contexts, I mean it's
- 33:24going to be in Microsoft Word 10 years from now, right?
- 33:27But what about anthropic? What about Llama?
- 33:29What about Google Gemini or Lambda?
- 33:32Or what about, you know, whatever Amazon's cooking up,
- 33:35what about a million startups got funded in the last 10
- 33:37seconds, right? I mean, I think those are like
- 33:39more interesting questions because it's to your to the
- 33:43earlier discussion it seems like a pretty commoditized concept
- 33:46like chat with a large language model and unless someone can be
- 33:49way better or have a very different business strategy than
- 33:52ChatGPT, it's hard to see where they're gonna win, right.
- 33:56And so people are trying different angles on this
- 33:58concept, but the concept itself is not that different in
- 34:00different companies. I do think this really gets to
- 34:04the question of will there be one best AI essentially 1
- 34:09ChatGPT like that we all use as our personal assistant, right?
- 34:13And that is the most general, most high powerful model of.
- 34:17Most highly intelligent model right that exists in the world
- 34:20you know because it's so generalizable and then you know
- 34:24I think there's a good. I think there's that's
- 34:26plausible. I think that certainly you know
- 34:28we use. 11 you know browser and one e-mail client, I mean I
- 34:33don't know like we don't, we're not switching between them a
- 34:35lot, right. But I think it starts to ask you
- 34:38know, will that model also be better at all other you know
- 34:42vertical tasks as well? I don't that seems harder to
- 34:44believe right? Like that.
- 34:45It will also be the best model at reading legal documents and.
- 34:50Being I was just gonna bring up case tax, sold the Thompson
- 34:53Reuters except for hundreds of millions of dollars and.
- 34:57They were like, I mean much more sophisticated than ChatGPT
- 35:00wrapper, but they were using ChatGPT largely.
- 35:04Sure, regardless of what their specific model was doing, it
- 35:09seems possible you could put a ton of compute and specifically
- 35:14trained legal data into a model that would outperform ChatGPT at
- 35:19that use case of the law, right? Or same for medicine or
- 35:22something, but. I don't know.
- 35:24I'm not 100% confident in that. I think that it's very possible,
- 35:27just with the you know that chat TBT as a foundational general
- 35:32model will just always be better at all of those key use cases.
- 35:36I mean Jason Warner, you know who's speaking in the
- 35:38conference, who's now the CEO of Poolside is betting that he can
- 35:42build a foundation model for code.
- 35:44Obviously, you know, yeah, you can see a bazillion companies
- 35:47that are sort of trying to be like I will be a foundation
- 35:50model. Yeah, for, for XI mean I think
- 35:54it's, I think it's sort of instructive to think back to
- 35:55like social networks, right. And for me, I always find it
- 35:58interesting analogy of, OK, you know, 1520 years ago would you
- 36:04predicted there would be a social network for work and
- 36:08there would be a social network for gaming and there would be a
- 36:11social network for people under 25.
- 36:13That's Snapchat. And there'd be a social network
- 36:14for people between 25 and 45 and that's Instagram and there'd be
- 36:17a social network for people who are over 45.
- 36:19And that's Facebook blue, right. I mean like and and and it's a,
- 36:24it's sort of interesting to think about what the cleavages
- 36:27are in the the user need. Right.
- 36:29And in social networking, it's often just age cohorts sort of
- 36:33build these network effects with each other.
- 36:34But then there's also this LinkedIn, which is, oh, actually
- 36:37work is a completely different social concept in your life that
- 36:40you need to keep separate from everything else.
- 36:42And then you have Discord, which is like actually gaming is like
- 36:45a completely different concept that you need to keep distinct
- 36:47from everything else. Right.
- 36:48And I don't think it would have been that easy to predict those
- 36:50things. I mean maybe like Reed Hoffman
- 36:52will say, it was super easy to predict, right?
- 36:53But but I think with these foundational models, it's
- 36:56similar, right? It's, you know, there probably
- 36:57aren't really network effects other than just like who can eat
- 37:00the most data the fastest, right, Which seems like it's
- 37:02going to be GPT. And so the data, quote UN quote,
- 37:05network effects, are the data scale effects are probably
- 37:08always going to be 1 by 1 by GPT, right.
- 37:11Then the question is, are there other use cases where there's
- 37:14some kind of network effect or there's some sort of different
- 37:17business concepts? You know, Facebook seems to be
- 37:19going the angle of we're going to open source this all.
- 37:21So people will just build it into all these things like it's
- 37:23Linux back in the day and that'll be the way we win is
- 37:26that it'll be the free open source version and you can just
- 37:29throw it into everything if you want to, which I think it's a
- 37:31pretty interesting like business concept, right.
- 37:33And then I don't know that much about the, you know, enthropics
- 37:35or the pool sides or whatever. You know, how are they going to
- 37:37win? What is the cleavage in the the
- 37:40use case or the way these models are built that is going to allow
- 37:43someone other than ChatGPT to win?
- 37:45Because it seems like they're gonna win on data and they're
- 37:47gonna win on hardware. So like, where are you gonna win
- 37:50if you're not them? I guess, and I'm sure everyone
- 37:52has an answer to this, but that's me.
- 37:53It's the hard question. We've spent a lot of time
- 37:55talking about the foundation models.
- 37:57I mean, Chad, I mean open a Eye is sort of a combination, right,
- 38:00where it's like they have the foundation model, they apply it
- 38:03to these use cases. I mean people talk a lot about
- 38:06you know, applications and like infrastructure, right.
- 38:09I mean, there are companies, they're all these sort of wonky
- 38:12companies like. The vector databases have been
- 38:15super hot of the last couple months, right?
- 38:17People talk about companies like Pine Cone, We V8 there.
- 38:21There's like a whole list of them which are just like trying
- 38:23to organize your data in a better way to get it into
- 38:27foundation models. I wanted to get to the actual
- 38:30like applications right and you sort of Max earlier referenced
- 38:34the sort of chat bot application which also goes back to you
- 38:39know, your both of your early days at volley trying to build.
- 38:44Chat bots. I'm curious getting away from
- 38:47who has the technical expertise, like what you think is
- 38:51interesting in the sort of what its character.
- 38:54Is it replica? You guys know this world much
- 38:57better than I do. So you're sort of getting at do
- 39:00we feel these are successful use cases or?
- 39:03Right. Yeah.
- 39:04Yeah. Do you think there's sustained
- 39:06promise there or what do you see in terms of people actually
- 39:10using AI and applications that? That excites you.
- 39:15There's a challenge with these models being accurate, obviously
- 39:20100% of the time, and you can debate whether that's necessary
- 39:24or not, right? To consider it True general
- 39:27intelligence. But in the entertainment space,
- 39:29like, it's just less of a problem, like conversing with a
- 39:33fictional character or historical character, right?
- 39:36Like these things don't need to be extremely accurate because
- 39:38they're essentially entertainment anyway, I think.
- 39:42I guess I always come back to some lessons I've learned in the
- 39:47gaming world, where there's a clear difference between an
- 39:52entertaining demo that is fun to do and fun to play with.
- 39:55And I think we've seen a lot of those that are actually really
- 39:58amazing and impressive demos, but they don't have long staying
- 40:02power or retention, right? So thinking of all of the apps
- 40:06that create fun. Photos that put you into their
- 40:10photos or. Yeah, what's can of soup.
- 40:13Can of soup is the. Latest I keep meaning to write
- 40:15about them. They're like super buzzy, right?
- 40:17Yeah. And you can put yourself in AI
- 40:19generated photos with your friends.
- 40:21I think it's amazing. It's really cool.
- 40:23But the question is, does that really have staying power?
- 40:26Do you build a social network around it in order to make it
- 40:28have staying power and network effects?
- 40:30These are real challenges to create like a sustainable
- 40:33business and startup that achieves a great outcome.
- 40:37So and similarly with talking to characters, character AI I think
- 40:40clearly has some product market fit there with people wanting to
- 40:44come back and talk to those characters.
- 40:46But I think what Max and I like to think about is how do you
- 40:48even build more retention around that?
- 40:50How do you build like game mechanics or features in the
- 40:53concept of a virtual pet that you might, we know from looking
- 40:57at mobile gaming and previous eras of gaming that you can
- 41:01build long running retention into that if you add game
- 41:05mechanics to the experience. So yeah, I'm less bullish on
- 41:09just like general like character conversations that don't have
- 41:14any are sort of aimless or don't have an end point or a purpose
- 41:17and more bullish on, you know, how do we bring those characters
- 41:21or Npc's into a gaming context with like normal gaming
- 41:25objectives. I do think in interacting with
- 41:28these underlying characters, I think to your point, I think a
- 41:30lot of the real value is when you start losing sight of it
- 41:33being an AI, whether it's like. Virtual boyfriend, girlfriend
- 41:36type thing, companionship, you know, sex chat.
- 41:39As we said, being a big opportunity there.
- 41:41Are those companies allowing it or they cracking down on it?
- 41:44They claim they're cracking down, but then if you go on
- 41:46Reddit and you look at all the screenshots from the last week
- 41:49of what people have been talking to these, I mean, go on, go on
- 41:52the Reddit of any large language, character driven
- 41:55experience, you know, you'll see what people are really
- 41:58passionate about using it for, you know, replica or character
- 42:01or whatever, right? So I think that, yeah, losing
- 42:05sight of whether or not it's an A I is, is important in some
- 42:07contexts. But in the end, I mean, in any
- 42:10story experience, from reading a book to playing a video game to
- 42:16chatting with an A I like the goal or watching a movie is to
- 42:20become immersed in a world that doesn't actually exist, right?
- 42:22And you know, when you're watching a movie, you're
- 42:24watching Lord of the Rings, you know it doesn't actually exist.
- 42:26Those characters aren't real, but you sort of become swept up
- 42:29in the narrative in the world, right?
- 42:30And. I think similarly, I think
- 42:32that's where the opportunity is with a lot of these character
- 42:35driven AI entertainment experiences.
- 42:37Do you guys want to try this other one that I have?
- 42:40Oh yeah, you have a game? Yeah.
- 42:42Sure. So I thought it'd be fun to see
- 42:47what GPT for ChatGPT for thinks your lives will be like in the
- 42:51year 2028 and see if you either of you agree with that.
- 42:56The prediction there so. I say it again.
- 43:00I I asked Chachi BT to predict a day in the life of Eric
- 43:06Newcomer, specifically a American journalist reporting on
- 43:11the Silicon Valley and startup industry, which I borrowed from
- 43:13your Wikipedia and, you know, gave it a little bit more
- 43:17context of sure about your sub stack.
- 43:19And yeah, it came up with A day in the life in the year 2028.
- 43:22So I'm going to read it to you and you guys can assess whether
- 43:26you think this is accurate. Day in the life of Eric
- 43:30Newcomer, 2028. The first hints of Don kissed
- 43:34the San Francisco skyline. As Eric newcomers, smart blinds
- 43:38slowly begin to rise. The ambient sounds of birds
- 43:41chirping echo softly through the smart speakers in his apartment.
- 43:45The AI? The AI.
- 43:47Everything's just smart. The AI driven home system has
- 43:50analyzed his REM sleep and calculated the optimal wake up
- 43:55time to ensure he starts the day with peak cognitive efficiency.
- 44:00Lifting himself out of bed, Eric's AR glasses are laying on
- 44:04the night stand feeding him the day's analytics.
- 44:07Newcomers Media Outlet has grown tremendously.
- 44:10AI. But we also have to be bullished
- 44:11by AR augmentedly out here and you're seeing okay.
- 44:14Well, you can. You can assess whether you think
- 44:16that's accurate, but you should. You should at least appreciate
- 44:19this newcomers Media outlet has grown tremendously and AI Bot
- 44:23the. New like, what?
- 44:25Is it like a horoscope? You're reading my Yeah, I'm
- 44:29predicting your feature and AI bot had curated and condensed
- 44:33the most relevant news from the vast ocean of information, all
- 44:37of this streamlined and visualized for maximum
- 44:39absorption. Eric's morning routine was
- 44:42synced seamlessly with his devices as he finishes his quick
- 44:46morning workout, suggested and supervised by his virtual
- 44:49personal trainer. His smart kitchen is already
- 44:52brewing, his favorite blend of coffee tailored to his genetic
- 44:55makeup and current health metrics.
- 44:58Settling into his study, the day's agenda was projected
- 45:01holographically. The AI had lined up interviews,
- 45:04articles to review, and podcast episodes to record.
- 45:08Eric's writing has also evolved. With the help of AI enhanced
- 45:11tools, he could construct and edit stories with increased
- 45:14speed and position. His A I assistant named Nora was
- 45:17able to give real time feedback on the readability, engagement,
- 45:21and impact of his writing. Nora also had a mode where she
- 45:24could mimic Eric's style, allowing her to draft articles
- 45:27for him. It was a collaboration that Eric
- 45:30had grown to appreciate midday. As Eric prepares for his podcast
- 45:34recording, he remembers the days when he used to worry about
- 45:37Internet connections and sound quality.
- 45:39Now with quantum driven communication technology, my
- 45:42God, not only was the. Connections, but whole
- 45:45technology here. But he could also record in
- 45:47virtual environments, giving his audience an immersive
- 45:50experience. Today's episode was recorded in
- 45:53a simulation of a venture capital firm in the heart of
- 45:57Silicon Valley. Yeah, this both he and his.
- 46:00Sandhill Rd. The most beautiful both he and
- 46:03his. Guests wants to be there.
- 46:05Both he and his guests appear as lifelike holograms afterwards.
- 46:09Eric took a moment to reflect. He looked out over the city,
- 46:12remembering the early days of Newcomer Now.
- 46:15He wasn't just delivering news. He was shaping the future
- 46:18narrative of startups and venture capital.
- 46:21The weight of that responsibility was not lost on
- 46:23him. Yeah, that's about it.
- 46:27Certainly. I mean, it's funny, I find a lot
- 46:30of the predictions about non AI stuff to be the most annoying,
- 46:33that it's so certain hardware is hard, artificial augmented
- 46:37reality. And I mean, there was some
- 46:38other. I'm pretty short quantum, right?
- 46:41Quantum Internet? Exactly.
- 46:42Yeah, can. I take the under on quantum
- 46:44powered Internet or whatever that was.
- 46:46Well, it's an interesting thing because we were talking about,
- 46:48you know, believing in AGI, right?
- 46:51And can you, you know, if you believe in AGI or general?
- 46:54Well, then it would be about much more, you know?
- 46:57Yeah, I get the point. We could have.
- 46:58We could wish everything. Yeah, exactly.
- 47:01Or. That's that's that sounds like a
- 47:03story an AI would tell me to convince.
- 47:05Me. That we wouldn't all be dead in
- 47:06five years, thanks to AI. And it was like, the future's
- 47:10gonna be right. Yeah, that would not be AI
- 47:12having killed us. Too optimistic.
- 47:14You'll have smart blinds. I know they.
- 47:15Try to like I mean I do think a problem with some of this stuff
- 47:19is like it's programmed to be like too benali optimistic like
- 47:23I find it like I like beg chat you video be more like George
- 47:27Carla. I don't know, just behave with
- 47:29some free thought. And it's so why isn't it like
- 47:32Eric, I don't know, at that age he's probably has some cardiac,
- 47:35you know, whatever. Where's the like medical thing
- 47:39where it's five years from now, you know, anyway is easy to
- 47:42start to hurt when he goes on runs or whatever.
- 47:44And oh, and I would be interested in sort of the AI
- 47:46piece, obviously of the medicine.
- 47:48I mean the idea that I would have a writing assistant.
- 47:52Seems like basically plausible. Today I go into chat literally.
- 47:57When I did, I wrote my own vows, but they proofread them and
- 48:00Chatchi video tweaked, told me to move one thing to active
- 48:04voice from passive voice. You know, it's like, I feel
- 48:06like, I feel like this is assuming that there's gonna be
- 48:09this like perfect equilibrium of you reporting the news and doing
- 48:13interviews and using your AI assisted.
- 48:15I just kind of find that to be, like, not that plausible.
- 48:18Like it's either gonna be one or the other, Like you're gonna
- 48:21still be doing most of the work, or you're gonna be doing almost
- 48:24no work and won't have a job, or you'll essentially have evolved
- 48:29into a brand instead of a writer, right?
- 48:31I mean, it seems hard to believe that we're gonna thread the
- 48:34needle here, that you will still be doing all this.
- 48:37Intellectual and the temptation, I mean.
- 48:39I do think one of the real fears I have about AI that will it's
- 48:44just like. The temptation not to think if
- 48:47it can do your task for you, right?
- 48:50I mean that's what we're sort of seeing with that's what I'm.
- 48:52There are a lot of like some types of cheating where it's OK
- 48:55you can bring in the formulas, we still have to use them or
- 48:58whatever. At least you still have to
- 48:59think. Whereas like when chat GBD can
- 49:02produce the final text for you. Man, that is like a path to.
- 49:07Not progressing as a writer anymore, right?
- 49:09Like as soon as you're just like you're not human out of the
- 49:12loop. You know what?
- 49:13Yeah, exactly. What am I?
- 49:14Besides, what are you providing to this business?
- 49:17Right. Yeah.
- 49:17I agree though, it's just so hard to get anything edgy or
- 49:20funny out of it. It's like kind of an, I don't
- 49:22know, it's tough to it's. Very driven.
- 49:27By scifi, I feel like one of the child, I think.
- 49:31I think this is the prompt, right?
- 49:32It's hard to get it to think like, I know you could say the
- 49:35same thing about humans. There is no independent thought.
- 49:37You're parodying a bunch of stuff you've heard, but it does
- 49:39feel like have it generate a really new, an actually new
- 49:45idea. You know, I think it's, I think
- 49:46it's very capable of that, actually.
- 49:48I I found that if you tell it to be extremely original, you know,
- 49:53ignore. I feel like it just gets like
- 49:56rhymy. I don't know.
- 49:57I feel like it. I do think there's like a
- 49:59prompting element to this feeling that it's not able to be
- 50:03as creative or original as you might think.
- 50:05In the same exercise that ChatGPT is there.
- 50:09We can compare and contrast in five years.
- 50:11What are specific predictions, less colorfully said, that you
- 50:16would make in five years? I mean, I definitely think to
- 50:23come back to the self driving thing, it seems like we're
- 50:25actually going to have fully self driving cars in five years.
- 50:27I mean, I know you could say. We are in cities.
- 50:29I think the big question there is just how much you know Waymo
- 50:33and. People think.
- 50:34Take what they've. Learned I think if today they
- 50:37can drive around San Francisco I think they'll be able to drive
- 50:39between cities and and in in the vast majority of cities or
- 50:43whatever at that point at least. I don't know if the United
- 50:45States it has to be trained on different data than other
- 50:47countries or whatever or something, but I think that.
- 50:50I mean the fact that we have live way Mo's like dropping
- 50:53people off on my street every day just makes you think you're
- 50:55going to be able to go anywhere in a self driving car in five
- 50:58years. Which again, I lost the bet the
- 50:59other way on that last time, so I'll probably lose the bet
- 51:02somehow this time. It's a problem.
- 51:04And also, just like car turnover is such a long life cycle.
- 51:08Depends how many you're what. Percent, It's not that I'm
- 51:11arguing for 100% penetration of those, It's more that I will.
- 51:14I will argue that you know you will be able to take one you
- 51:17know, as an Uber or whatever in any major US city.
- 51:20And you know, I don't know whether or not you'll be able to
- 51:23buy one. That's sort of an interesting
- 51:24question as to whether or not it'll be an ownership model
- 51:26versus an Uber type model or a lease model or whatever you want
- 51:29to call it. But I think they'll be like you
- 51:31know, available for use. So you know at large scale
- 51:35across the major U.S. cities and also between cities, right.
- 51:39It just seems like we're clearly like pretty much they're
- 51:43assuming way mo and crews aren't, like lying about the
- 51:45capabilities of their their vehicles.
- 51:47Interesting that your main predictions around self driving.
- 51:50I was thinking about something. I feel pretty good about and I
- 51:52agree with that. I do think this view of you will
- 51:56wake up and there will be essentially your personalized
- 51:59assistant. Maybe it'll be in the smart
- 52:01speaker or in your wall or your mirror or something, I guess.
- 52:05I just, I don't know if it'll be Alexa or ChatGPT or something
- 52:10brand new, but I do think, you know it's going to be a voice
- 52:15probably driven experience. And bias.
- 52:18Bias. Well, at least in your home and
- 52:20probably bias. You're going to be typing to it.
- 52:22I didn't love it. It doesn't make.
- 52:24Sense. Very efficient.
- 52:25People love it. It's very so people do not love
- 52:27typing, no. Typing is way less efficient
- 52:30than dogging. No you're I mean it'll you'll be
- 52:33able to type to it. But I just think like most of us
- 52:37will be like talking to this AI assistant and it'll be I
- 52:41generally think there will just be one that I use every day
- 52:43maybe maybe there will be more than one in the market like that
- 52:46people use. But I will just have one that
- 52:49learns my preferences and becomes personalized to me and
- 52:53creates kind of a history with me.
- 52:56And it might even actually recommend other assistance for
- 53:00certain use cases, right? If I have to go prepare for
- 53:03Max's congressional testimony, I will maybe use a separate bot
- 53:07for that or something. But yeah, specialize in.
- 53:11Testimony bot. But I do think, yeah, I feel
- 53:14pretty strongly we're going to have conversations, voice
- 53:18conversations with our personalized assistants every
- 53:20day. The prediction I'll make this
- 53:23sort of different than what's been said is underlining that I
- 53:27think the average person wants consumption more than creation.
- 53:32And that TikTok is in some ways the actual most used thing in in
- 53:37the AI world and and this sort of yeah, chat GBD got at this a
- 53:41little and its prediction about me where it's very good at
- 53:44sorting things that I want. I I think in five years we'll
- 53:49see at least the beginnings of like pure AI generated social
- 53:54account. I mean, you're already seeing
- 53:56sort of like these, like, women and cartoons that are like.
- 53:59This seems like they're gaming like the Instagram algorithm
- 54:02with like totally sort of machine made.
- 54:05But like the I think even like video within five years I think
- 54:09we'll have, yeah, you're on TikTok and it's just here is a
- 54:12generated video and it's warring with actual creativism.
- 54:16I think that would be like potentially A generationally
- 54:19culturally interesting period where like you have like young
- 54:23people coming up where they're just like.
- 54:25Being fed sort of what they want outside of it could create like
- 54:29a really weird type of humor where like they have they're.
- 54:32Used to? Yeah.
- 54:33Good. There's like an AI.
- 54:34There's like a bunch of AI sort of celebrities and accounts and
- 54:38they create content and they interact and they host, you
- 54:41know, podcasts and. It's all like effectively a be
- 54:44tested. It's like they put out like 100
- 54:46version build, you know, a ton of versions of the video and
- 54:49they see which one is getting engagement and then they slowly
- 54:52funnel into those just TikTok chooses which videos get
- 54:54surfaced. Yeah, that approach is in the
- 54:57creation. It makes most sense for short
- 55:00term, short form and but you know the day you can make like a
- 55:03movie about it, like then we're like killing American industry
- 55:06but. Yeah.
- 55:07Well, the question is, do you think we're going to get
- 55:09personalized? I mean, you mentioned they're
- 55:10going to a B test everything or whatever.
- 55:12But is it that there are literally be 8 billion different
- 55:16versions of each piece of content for each person?
- 55:19If there is no such thing as a piece of content anymore, like,
- 55:23it's just you get a version of some concepts like that is
- 55:27perfect for you, right? I mean or do we still have some
- 55:30sort of like value in being like, oh, did you see that video
- 55:33the other night about that thing and you can talk about it and
- 55:35you can do, I think? Relationships, I mean, people
- 55:37are going to try both. I mean, I think one, there's
- 55:40like a limited, there's like A at some point there aren't
- 55:44enough humans, right? Or there's like there there is
- 55:48like a data shortage, right? It's hard on some level.
- 55:50If you do something population wide you can really test and see
- 55:54what works like broadly. Whereas like running experiments
- 55:57on me, you just don't not get enough shots on goal necessarily
- 56:02to be great at like the TikTok style.
- 56:04So to me that leans a little bit more less this sort of obsessive
- 56:08personalization and more. I mean, obviously they're
- 56:11subgroups, but not like to a person, more like subgroups.
- 56:14Yeah, I think subgroups, like, I think it'll not just be like
- 56:18training on your information, but other people who have
- 56:20similar experience, similar behaviors that you do on the
- 56:24app, right? It's able to use that's that
- 56:26data for training too. My last AI a GS in prediction
- 56:29is. I mean this was in some of these
- 56:30forward-looking GBT scenarios is?
- 56:33I mean, I do think you take five more years of development on AR
- 56:36glasses or whatever, You know, I don't think they'll be as
- 56:38lightweight as the glasses we wear every day.
- 56:41It'll probably still kind of look like a VR headset in many
- 56:44ways. But I do think people are going
- 56:46to spend like hours per day inside a high quality Apple
- 56:49Vision pro type experience. Because I think again, I mean, I
- 56:53mean if the average Americans watching five or six hours of TV
- 56:56a day, which they are like. Why wouldn't you watch two of
- 56:59those hours on 100 foot screen in front of Mount Hood,
- 57:01Washington or whatever, right. Which is basically the pitch
- 57:03from Apple, you know or why wouldn't you watch personalized
- 57:07AI generated TikTok content like from Eric's, you know, feed or
- 57:09whatever, right. I just think that, you know,
- 57:12screens getting better has been one of the true constants of our
- 57:15lifetime, right? And the Vision Pro which is sort
- 57:19of the infinite screen, the infinite canvas for visual
- 57:22content is, is kind of the ultimate expression of that,
- 57:25right. So.
- 57:26I think. That I'm more on the tenure.
- 57:27There will be quite a lot of time.
- 57:29And I feel like that space has been dogged by limits of optics
- 57:34and just sort of hard constraints.
- 57:36And so a lot of our intuitions about software level
- 57:39improvements don't translate and I would say I would take the
- 57:43longer time horizon. On that maybe, Yeah, I don't
- 57:47know. Apple seems to think they got
- 57:48it, so I guess we'll all see. Yeah.
- 57:49This is a smart Max take. When's Apple been wrong before?
- 57:52I don't know. Trust.
- 57:53I'm just saying that when Apple ships something like.
- 57:56I mean, again with the iPhone, like, was there a touchscreen
- 57:58that had ever worked before? No, there never had been, right?
- 58:01I mean, I just think that are you gonna get?
- 58:02That wrong? Is that out yet?
- 58:04No. Next year.
- 58:05Oh, I. No, you can't get it until next
- 58:06year, but I'll definitely get one for sure.
- 58:09Yeah, you're like, I can expense.
- 58:10I mean, for God's sake, right? Like you're like, we're a game
- 58:13company. I would buy it out of my hard
- 58:15earns personal money. OK, I would.
- 58:18I I don't know. I I guess, yeah.
- 58:20I I trust Apple when they make a.
- 58:23I mean, this is a decade long hardware bet and they waited a
- 58:26decade instead of shipping five or seven years ago because they
- 58:30didn't feel like it was good. Enough is there, besides
- 58:32obviously that Chad GBT was getting us thinking about this
- 58:36technology, Do you see any sort of AI connection or AI utility
- 58:41or how do you see? I just again, think with these
- 58:44characters that we're talking about, whether it's an assistant
- 58:47or whether it's an in game character or whether it's a
- 58:49virtual companion or whatever, a pet or a girlfriend or
- 58:52boyfriend. Like if they are the size of a
- 58:56real object that you know, if they're the size of an actual
- 58:59virtual boyfriend, girlfriend, or a size of a, you know, a
- 59:02Pokémon, whatever that actual size is and you can talk to
- 59:06them, that's going to be a better experience than.
- 59:08Looking at a tiny little screen in your hand, right?
- 59:11And. I wanna.
- 59:12Yeah, I wanna. Write I mean the.
- 59:14Companion I've got. Yeah, yeah.
- 59:16I mean, have you read the Golden Confidence?
- 59:17Right? They all have.
- 59:18Like, yeah, yeah. Yeah, how do I get that?
- 59:21But you want that's part of self projection though also.
- 59:25I mean, you want everybody to see that kind of thing, not
- 59:27just. So then, yeah, sure, if you're
- 59:30both in Apple vision class headsets.
- 59:33Right. Amazing.
- 59:35All right, this is what this is basically our first episode.
- 59:39We're gonna come out with probably five more.
- 59:43I think next episode will be, what do we think of sort of the
- 59:47apocalyptic vision of AI? Yeah.
- 59:50Anything you guys would add on what to look forward to you in
- 59:53the next couple episodes? I'm excited to run this
- 59:59conference. I think it was amazing the first
- 1:00:01time. I think it's gonna be even
- 1:00:03bigger and. Better this time.
- 1:00:05Yeah, I genuinely believe this is like the biggest thing since
- 1:00:08the Internet or the iPhone. So I think that we're all pretty
- 1:00:12authentically pumped about what's happening in the space.
- 1:00:14And I think the amount of stuff that's happened in the last six
- 1:00:16or seven months has been mind boggling.
- 1:00:18And it just feels like things are moving faster than at any
- 1:00:21time in my entire life in any technology space.
- 1:00:23So it's just super exciting to. Even be like remotely adjacent
- 1:00:26to any of this, so. I'm at the heart of it at
- 1:00:28Cerebral. At the heart of it, yeah, at
- 1:00:30least physically. At the heart of it, working on
- 1:00:35working on the other elements and being at the heart of it.
- 1:00:37But yeah, physically for sure. Great.
- 1:00:39Well, that's our episode. I'm Eric.
- 1:00:41Newcomer, Max Child, James Wilsterman are my Co hosts of
- 1:00:46cofounders of Volley. Thanks so much to Scott Brody,
- 1:00:50who's been producing the episodes.
- 1:00:52Shout out to Riley Kinsella, my Chief of staff who's super
- 1:00:55involved, Gabby Caliendo who works at Volley, whose quarter
- 1:01:00to the conference and making everything happen.
- 1:01:02I think she made sure you guys had microphones and lights so
- 1:01:06you could actually see and hear you.
- 1:01:08Thank you to young Chomsky as always for the theme music
- 1:01:13Please got to build the feeds? Like, Comment, Subscribe on
- 1:01:17YouTube, give us a nice review on Apple Podcasts and.
- 1:01:21Go play Yes Sire or song quiz. And for me, subscribe to the sub
- 1:01:28stack newcomer.co That's the most important thing.
- 1:01:32Thank you. All right.
- 1:01:33We'll see you next week. Goodbye.
- 1:01:36Goodbye. Goodbye.
- 1:01:37Goodbye. Goodbye.
- 1:01:38Goodbye.