OpenAI Board Acceleration vs. Doomers, Job Displacement vs. Replacement & Parasocial Winners vs. Losers

· Podcast Episodes,VC and Angels,Popular

“Chess is one of those things where there's a very finite set of things. It’s a probability calculation exercise. Every piece can only move in one way, and your opponent's pieces can only move in one way, and so you can calculate the probability of that. And so I think that is one thing where you would expect a computer to take over all of it. Maybe I have a little bit more optimism which is, I do think that most things are not like chess. People can still do crazy things and be unpredictable, and so sometimes I do think that human judgment is still required, but you can augment it with machines to do a lot of the much more tedious, repetitive tasks. That's what you should let machines do.” - Shiyan Koh

“It was a unique set of circumstances because, in the business world, there are very few things that were out of the blue as this. If the company is underperforming for five years, then you kind of know someone's going to step off, but this was a very interesting dynamic. I remember reading the news and thinking that this has been driven by a philosophical difference because if you look at the composition of the board, a lot of them are effective altruists, very focused on the opposite of accelerationism, which economists would call ‘doomerism’. There's a saying they're more safety first. From a policy perspective, they are the regulatory-friendly folks who are looking to slow things down. And so, the jury's still out whether they come across as naive to idealists, to principled, stubborn, zealots. History will write that story maybe in 10, 20, or 30 years.” - Jeremy Au

“Romantic relationships are dropping over time. There's a loneliness epidemic, as they call it, but my perspective is that people say they want relationships and more importantly, they want things that come out of a great relationship: companionship, affirmation, or a buddy. The games and entertainment industry are giving a lot of that. We used to make fun of Japan, the “otakus” and all these shut-ins, and I thought Japan is just an early adopter. At the end of the day, the truth is real estate prices are still the same, but the cost of compute has dropped by an order of magnitude. And so, I was reading a YouTube comment saying, ‘This is the cyberpunk dystopia that we had,’ which is true. It just takes time to get there, but in 100 years, real estate will still be about the same price, but the cost of compute, whatever you call it, AI, VR, will drop by several orders of magnitude again.” - Jeremy Au

In this episode, Shiyan Koh, Managing Partner of Hustle Fund, and Jeremy Au discussed three major themes:

1. OpenAI Board Conflict: Jeremy & Shiyan delved into the recent leadership changes at OpenAI, including Sam Altman's reinstatement as CEO after he had been fired by the board. They highlighted the philosophical divide within the board regarding OpenAI's direction between accelerationism (pushing for rapid acceleration and commercialization) vs. advocating for cautious progress. They noted that the accelerationists had prevailed, signifying a continued push toward commercialization. They also touched on OpenAI’s extraordinary growth, scaling from zero to a billion dollars in revenue in three years, and the practical challenges of halting AI development, given its open-source nature and widespread adaptation.

2. Job Displacement vs. Replacement: Jeremy and Shiyan discussed the Coatue AI report on the dynamic interplay between AI and human labor, the capability to scale human-lean organizations with compute power and the possibility of ultra-lean companies, like a 'three-person unicorn', driven by AI efficiencies. Specific professions, such as those in marketing and customer service, are being competed by AI faster. They also speculated on job displacement vs. replacement dynamics, and the urgency of individuals to personally adopt AI to stay ahead.

3. Parasocial Winners vs. Losers: Jeremy and Shiyan shared that the rise of AI will create distinct winners and losers. They discussed that corporations that effectively leverage AI will likely dominate and significantly transform their operations and productivity, rather than individuals or laggard companies. They also touched on the increasingly para-social nature of human relationships in the age of AI. Just as how the invention of the birth control pill revolutionized family planning, social norms and economic productivity, AI relationships will transform how humans interact and form relationships with AI, corporations and other humans.

They also talked about the global loneliness epidemic, the marketing choice to frame AI as a co-pilot and other historical parallels with past technology waves.

Supported by Acme Technology

Are you a business owner, CFO, or engineering lead who's tired of grappling with outdated finance processes? Are you frustrated at the high costs of card payments or find yourself bogged down by manual financial tasks? It's time for a change. Meet Acme technology. Our software enables you to connect directly with your bank of choice to automate all your finance and payment processes. Enjoy real-time reconciliation and direct bank payments and payouts. No lengthy integration. Transform your banking experience into a Stripe experience. All with easy integration through streamlined APIs. Learn more at www.tryacme.com.

(01:55) Jeremy Au:

Morning, Shiyan.

(01:56) Shiyan Koh:

Good morning, Jeremy!

(01:58) Jeremy Au:

There we go.

(01:59) Shiyan Koh:

Energy, that caffeine boost.

(02:02) Jeremy Au:

Oh, there we go. Can't go around that multiplier. Well, today's a big week.

(02:06) Jeremy Au:

I mean, you know, obviously AI has been around, but the week has been full of OpenAI news about so on and so forth. It also ties in nicely with AI discussion that Harvard and Stanford hosted in Singapore and that were a panelist with and I was helping organize.

(02:20) Shiyan Koh:

You tricked me into doing it and then didn't show up. Let's be honest.

(02:23) Jeremy Au:

Well, unfortunately, the only day that worked for you and the other speakers was when I was on National Service reservist. So I was like, okay, there we go. But sure would have been nice to have been there. And obviously, there's a Code 2 AI report they want to go into. So let's talk, today's episode will be all about AI, OpenAI, the business of AI.

But before we start, I just want to do a quick shout out to Todd Westra. So he commented on the last episode in tree 11, down rounds and startup failure patterns. So he said, Hey, you know, I really loved episode because the podcast covers so much ground from fundraising to strategy, and he felt that, he's glad that we talked about how customer segmentation and emotional intelligence are such important and often overlook aspects of startup success. So thanks so much, Todd, for your wonderful comment and praise. Thank you.

(03:06) Shiyan Koh:

Yeah, appreciate the feedback. I tell people that I labor in the delusion that no one listens to us, otherwise we'd never do this, but then it also is gratifying when people say like, I thought that was pretty useful. And then you're like, okay, I guess some people listen to us and we don't sound like complete idiots.

(03:24) Jeremy Au:

It's like today, we would deliver on sounding like idiots.

(03:27) Jeremy Au:

So let's talk about, maybe about the first one is like, you know, there's the big who are about, you know, OpenAI

(03:32) Shiyan Koh:

The five day coup.

(03:34) Jeremy Au:

The five day coup.

(03:34) Shiyan Koh:

The battle of Sam Altman.

(03:35) Jeremy Au:

It's like what I say, it's if you aim for the Emperor, you better not miss kind of a basic rules of coup planning is plan better. Oh boy. What a, I mean, it was just like, literally you woke up and then my phone is like full of, "Did you see this?" And then secondly, 30 minutes later, all these memes start coming, dropping in, you know?

(03:50) Shiyan Koh:

The first job AI replaced was Sam's.

(03:53) Jeremy Au:

Oh, that's a good one. Yeah. And then it was like, Oh, maybe all kinds of jokes, right? It's oh, if you're having a bad day, think about Sam, right? So that was interesting. But I mean, like the high level effects of the thing was that chronologically seems the board was very much organized as a nonprofit and he had a for profit entity underneath reporting to it.

(04:11) Jeremy Au:

Microsoft invested in a for profit entity. And then the nonprofit board received word of an advance in artificial intelligence called Q Star. And then that precipitated a move where they decided to oust Sam Altman and then Sam Altman, after five days and with the help of Microsoft and other shareholders in the corporate entity eventually came back onto the board. guess also the open AI staff was signed a petition and tried to walk out as well. So yeah. And now Sam is back as the CEO of OpenAI. So that's a rough high level point of view.

(04:40) Jeremy Au:

Shiyan, any thoughts about the situation?

(04:41) Shiyan Koh:

I mean, I think there's no clearer demonstration of "Hey, your talent walks out the door every day, and you need to make sure they walk back in". Then, I don't think the board anticipated that when they removed him. The bulk of their 700 employees would be like, We're out. Right? And so, that was just like the, I mean, even, Ilya, who had been part of the board that voted to oust him, was like, I changed my mind and signed the petition, and I'm also out. And so I thought that was particularly striking. Of course I think we should caveat this. We have no inside information, and we don't really know what was going on in the boardroom and all those sorts of things.

I did chat with a number of folks that I'm close to in the Valley and I was like, "What's going on?" And even folks that like, you know, competitive companies and things like that, they're like, we have no idea. We're just as surprised as everybody else. And so I think it was one of those things was so out of the blue. And so I think the board overplayed their hand both on anticipating Microsoft's reaction, who's like a major investor and partner, and their own workforce's reaction to not think through how that's going to work. But I guess now we can say the accelerationists have won the day, the battle. And I think the drive towards commercialization just continues apace, right? So, we're in for interesting times.

(05:54) Jeremy Au:

Yeah. And I think that's the key takeaway here is that two parts, right? One is, it was out of the blue because it very much was out of the blue. It was a very unique set of circumstances, which is uncommon because in the business world, there are very few things that were out of the blue as this, right? If the company is underperforming for five years, maybe then you kind of know, like someone's going to step off, et cetera, but this was a very interesting dynamic where it was very much, I remember reading the news and I said, this has been driven by a philosophical difference that's very there because, if you look at the composition of the board, a lot of them are effective altruists, a lot of them are very focused on, you said the opposite of accelerationism, which is, I think, economists would call them doomerism, which is, I think, very negative.

(06:30) Shiyan Koh:

But they're not, right?

(06:31) Jeremy Au:

Yeah. There's a saying they're more safety first. You might even say it, from a policy perspective, they are the regulatory friendly folks who are looking to slow things down. And so, I think the jury's still out whether they come across as naive to idealists, to principled, to stubborn, to zealot, right? I think that's going to be a very, I think the history will write that story maybe in 10 or 20 years or 30 years.

(06:53) Shiyan Koh:

I don't know. I think it's sort of like last year when they first released ChatGPT, and there were a bunch of people saying we should pause.

(06:59) Jeremy Au:

Yeah.

(06:59) Shiyan Koh:

And I was like, how are you going to make everyone pause? It's just like, regardless of what you believe, just as a practical matter, how would you actually stop people?

(07:07) Jeremy Au:

Yeah, but they did pause it, right? The reason when it was publicly released was effectively ChatGPT 3.5, right? That was really had that public release. And if we think about it, ChatGPT 1, 2, 3, that was a long time that they spent effectively. You wouldn't call it stealth, but they kept it private. And they did some pre-wiring warnings and I remember reading the why article and they were like, Sam Altman's really worried about this ChatGPT 2 because it's going to change the world, right? And he was, they were right. I mean, he did change the world, but, it was just kind of funny to read this article and I was like, oh, thank you very much. And I scrolled to my next piece of news, right?

(07:40) Shiyan Koh:

Yeah no, but my point is just the cat is out of the bag. You can't actually stop people from building. There's open source models that people are going to take and adapt. So even if all of the CEOs say we're going to pause, you're not going to stop the open source community from going. So I just think, as a practical matter, it's like actually quite hard for people to be like, "Hey, we're going to just stop development." The idea is already in the universe.

(08:01) Jeremy Au:

Yeah. I mean, it's kind of crazy, right? And I think it's just such a fast uptake as well. I remember I was reading, it was like, for a couple of years, it was zero revenue. And after that, it became like a couple of mil in the next year. And then this year, it's effectively a billion dollars of revenue. So, it's the fastest you can say, what is the fastest unicorn? I mean, it's already a unicorn before that, as well.

(08:19) Shiyan Koh:

Fastest to 100 million users.

(08:20) Jeremy Au:

Yeah. And fastest to a billion dollar revenue as well. And for me, I think I'm actually a little bit sympathetic, I wouldn't say empathetic, but sympathetic to the doomer side. I was reading this wonderful Malcolm Blackwell profile of the birth control pill, right? And this guy was walking around. He's just a scientist, thought he was going to do something that was nice. He was a Catholic and so he very much wanted to help improve reproductive health. And then he ended up reshaping society because the birth control changed society entirely, created in the 1960s, you know, it changed working people's roles, family structure globally. So, from his perspective, he was just doing the research needed to help people regulate their hormonal systems for birth control. So, I'm kind of like mildly sympathetic to the point of view which is we just unleashed something and we're pulling back, right? So that's an interesting dynamic.

(09:04) Shiyan Koh:

Makes sense. Should we go into the report? Because I think there's like a lot of really good stuff here.

(09:08) Jeremy Au:

Yeah, I mean, it also ties into what you shared with the Harvard, Stanford community. So why don't we just keep it high level? So Coatue AI is an AI full report. It's 115 pages. So lots of slides, but each slide is nicely done. It's not very dense McKinsey style, very clear headlines. I'll say about one third was context setting Wanted about it was like what we're seeing today, right now, as a result in terms of the landscape and the one-third is like, how does it change the future? And I think you actually tie up back to some of the societal changes that we think will happen eventually. But any favorite aspects, Shiyan, that jumped out at you.

(09:39) Shiyan Koh:

There's a few favorite slides that I had. One, this is slide 29, which is it's already delivering value, because I think there's a lot of people who are like asking the questions like, Oh, is it a fad? You know, hey, people were interested in one hour delivery. Then people were interested in crypto. And then now, people are interested in AI and all these different things are getting funded. And is it just hype? And I think AI is actually fundamentally different in that it is a platform shift. It's like a technological change.

And the way that we talked about it at the Stanford Harvard event was to say, like as technology progresses, what happens is you get higher and higher levels of abstraction. So if you ever go back to the really early computers, remember punch cards or like those photos that you see of people when they were like at NASA trying to program the rockets to go to the moon? And it was a very like low level machine line code set of instructions where you had to write out everything. And with the development of software, each level, we get more abstractions. So we no longer have to write every line of code. We have libraries that abstract functions. And this instant, this chatGPT, and what it's basically able to show us is that in some circumstances, we don't actually have to write the code ourselves, we can just speak in natural language and say the thing we want, and it can write the code for us. And we can do this in multimodal, not just text, but image and video. And why that is exciting is because when you are able to abstract more things out of it, you actually make this power accessible to more people. So more people can go build things and put things together. And you can make existing workflows cheaper, faster, better, all that sort of stuff. And so I think that was like, hey, this isn't like pie in the sky. It's already happening.

So that's what I like about that slide. Which is Google, Microsoft integrated it into their co-pilot, so with Github, so people are using it for programming, already becoming much more efficient. I think they're integrating it into their Word suite, so whether it's like writing tools Excel, Office, all that sort of stuff. So you're already like seeing the benefits of it, and they've seen, they've like rewritten physician communication, and people actually prefer the rewritten, GPT communication. It actually makes your doctor sound warmer, more understanding. You're getting benefits from it already. And so that was like, one slide that I really liked, which is I do think it's going to have a major impact on productivity, and transforming types of work.

And the really extreme version of this is there's a slide where they talk about well, could you build like a three-person unicorn company, which sounds a little bit extreme, right? But the idea that like, hey, there are many functions that you used to have to hire huge teams to do, but if you can actually build agents autonomously, instruct them and they're learning. You can build really lean businesses, which I think is a super interesting idea. But there are 120 slides, so I could do this forever. So, Jeremy, what were some of your favorite nuggets out of the deck?

(12:23) Jeremy Au:

Yeah, I think one favorite nugget in my head was I've always laughed at this framing of calling AI co-pilot and I always thought it was hilarious because my big rule about marketing is that either they describe exactly what's happening or they entirely articulate what's entirely not going to happen. And the co-pilot always felt like one of those very nice words that felt very reassuring psychologically to humans and be like, Oh, AI is a co-pilot to you. And then there's all these crazy quotes out there. It's "Oh, it's not going to be an AI job. It's going to be an AI and human is replacing a job. And I'm like. What? That's not how productivity technologies work, right?

I think the big story that I always remember is it goes back to the mechanical loom. It's like the weaving industry, right? Everybody used to make their own clothes in every village, etc. And at some point, the mechanical loom arrived, and basically people were able to weave with much more productivity. And then everyone's, oh, okay, great, all we have to do is defend ourselves by having our own looms. But the truth of the matter was that some cities just became loom centers where they had not just thousands, but hundreds of thousands of looms: the infrastructure, the mechanics, the repair folks, the weavers, et cetera..

And then you end up in a situation where the British colonial empire is super scaled, and then they basically made everybody else in the world make cotton, ship, it was more economical for people to make cotton across the world, and then ship that to Britain for it to be weaved and then for the clothes to be pushed out back into the world, right? And so basically, if you were like a normal seamstress or a tailor, you just got swamped by these manufactured goods. And so I think to some extent, it's not really a co-pilot. It's not like a one to one dynamic. And I think this deck kind of put it nicely, right? It's look, there was a co-pilot phase for chess where human and AI was better together for about less than effectively 10 years. And then afterwards, the just machine just crushed all humans in chess, right? So, I think you can say oh, it's a co-pilot in the sense that maybe it's one human, percentage of human effort and 99 percent computer side.

But it kind of goes back to what you talked about. It's like the the new organizational structure which is that, historically a company needs to scale if you want to increase marketing. I'm seeing an example. You would have to hire 10 more people if you want to grow at 10x. But now you're saying that you can scale the marketing division using more compute power. And I think it was a great Twitter thread that happened just now. And this guy basically was like, Here's how I created a blog with like 1.4 million articles and crushed the SEO game by cloning all competitors and generating AI. And I was like, yeah, that's exactly what I expect, right? It's not going to be a million SEO people each using the AI. It's going to be like five corporations that figure out how to build a million pages of SEO-friendly pages. And of course, they're going to kill Google eventually with this kind of approach. But in the meantime, it's like an algae bloom. It's still beneficial to grow as fast as you can because computers can scale so much faster than humans. And so I think it goes back to the previous conversation we had about job replacement and job displacement is being very thoughtful about where jobs are going to be.

I think that if you're something that's going to be attacked by AI, it can grow very fast. And I think marketing is one of those non-protected professions. I think being a doctor is a protected profession.

(15:12) Jeremy Au:

But I think there's gonna be an interesting dynamic and I can imagine a lot of human labor friendly. I think unions are going to come back in the context of this legislation because you're going to have a point of view.

(15:20) Shiyan Koh:

Yeah, I mean, it is interesting though. Chess is one of those things where there's a very finite set of things. It literally is a probability calculation exercise. Every piece can only move in one way, and your opponent's pieces can only move in one way, and so you can calculate the probability of that. And so I think that is one thing where you would expect a computer to take over all of it. Maybe I have a little bit more optimism than you, Jeremy, which is I do think that most things are not like chess. People can still do crazy things and be unpredictable, and so sometimes I do think that human judgment is still required, but you can augment it with machines to do a lot of the very more tedious, repetitive tasks. And that's what you should let machines do, right? There's no reason why humans should

(15:57) Jeremy Au:

I'm actually bullish overall. I think it's a productivity gain for human society. I think it's a net benefit assuming there isn't a singularity event, but maybe it is beneficial. But I would say, but I would say, but I think we just had acknowledged that I think they're going to be some job professions that is going to hear, see huge amount of displacement. I just don't think the job SEO or marketing, a lot of those jobs are going to be.

(16:18) Shiyan Koh:

They've been agents,

(16:19) Jeremy Au:

Yeah. So, you know, customer service, call centers, to some extent BPO as well. So there's a lot of displacement in those jobs. And the good news is that most of them, they don't have a union, right? So, so it's gonna be a issue. I think laws, I always think to myself one of those interesting ones where I love the job is go is replaceable, but it's a very strong union dynamic and is locked into every country. So it's a very interesting. I think that's where you're going to see a lot of action in terms of the pro versus I was going to say pro versus anti human labor activists, but he called it pro human, pro AI, productivity dynamics. I think that's going to shape up quite a bit.

(16:50) Shiyan Koh:

It was interesting at the Harvard Stanford talk. There were a number of lawyers in the audience, and they cited the case where, some lawyer in the U. S. had used ChatGPT for his arguments, and ChatGPT had hallucinated cases that didn't exist, but that supported his case. And the question was like, how do you stop that? And my response was, that isn't a ChatGPT problem, that's a human laziness problem. The lawyer shouldn't do that. The lawyer should check his work, right? The lawyer shouldn't just be like, blindly. And so, I don't think every technology has people trying to like, skype off and do things. And, you know, I don't think it's the technology's fault. It's kind of the people who are using it have to decide how they're going to use it.

(17:26) Jeremy Au:

Well, the awkward reality is that human laziness is the driver of productivity across the world. I mean you can make the argument is that, oh, I can just you know, I mean, oh, come on. Right. It's just, would you rather maintain a horse or you drive a car? Right? Would you rather have a self driving car where you can watch TikTok versus drive a car? Right. You know, it's just a huge amount of difference there.

(17:44) Shiyan Koh:

Yep. Yep. I would agree with that. But it is funny, right? Because I feel like the lawyers were feeling that their jobs were a little bit threatened. And so I think that's like a push to say you should actually learn how to use this for your benefit, not just be like, how do we regulate it?

(17:56) Jeremy Au:

Yeah.

(17:57) Shiyan Koh:

It's not allowed. You're like, well, you're not gonna, people are gonna use the tools. So, you just have to figure out how to deal with that.

(18:02) Jeremy Au:

Yeah, and I know it's interesting because, I also teach, right. This is for fun as an adjunct lecturer, and it's interesting because I had still have response to ChatGPT, I had to change a lot of the assignments towards group work or video format, but I still had some level of written dynamic that was there, and it was interesting because I realized that I just can't have it next year, like a written assignment format, because all these have to create a lot of what's the word higher level and I was with reasoning dynamic because the summarization aspect of it, I could feel, and I'm going to say this, I'm not saying that, but I felt that some of it was what I call coin AI fishy, which is it reads like AI wrote this and I can't say it was written by AI. I'm just

(18:40) Shiyan Koh:

You didn't run it through the tool?

(18:41) Jeremy Au:

The checker doesn't work, I mean, is this, it's like, the problem is that it can give you an indicative thing, but it doesn't have his false positive, so you can't go back to the student and say hey, this is, you know, not kosher in that sense, right? So anyway, the long story short is, now there's a lot of stuff I'm getting, there's this AI fishy, it's I'm pretty sure it was written by AI, not only I got it, I got an apology letter, and I'm pretty sure it was written by AI. So, I don't know how I feel about the apology letter that was written by AI

(19:05) Shiyan Koh:

Can you just make them write by hand in your class? Timed in front of you?

(19:09) Jeremy Au:

That feels a bit of a Luddite response, no? I don't know, I feel you just gotta change it to like group work that's, you allow ChatGPT I don't know. You allow peer evaluations. I think that's how I'm thinking towards more towards otherwise, you know, is this impossible, really impossible. The good news is that everybody's becoming smarter as a result, right? Because, instead of writing a shitty apology letter, you write a pretty decent one that looks kind of like AI.

(19:31) Shiyan Koh:

My friend showed me something really funny, which was that his father was starting to use ChatGPT to formulate his responses in the family group chat.

(19:41) Jeremy Au:

That's so funny.

(19:42) Shiyan Koh:

So he was like, this isn't how an old Chinese man talks. Like, where did all this come from? But then he like, couldn't get him to admit it. But he was like, clearly he's my dad doesn't talk like this.

(19:51) Jeremy Au:

Yeah, it's a bot has taken over. I mean, it's like the pod people . They did take it over. No, I mean, I I shared before last time, it's like there's a guy I know who who didn't know how to break up with somebody. So he asked ChatGPT to create a script on what to cover right on his call. Which is kind of bonkers. I mean, you mentioned breaking up somebody, but of course, you know, if you think about it, it's understandable. I mean, if you're young and you don't know how to break up with somebody, and then you ask ChatGPT, ChatGPT will give you 10 bullet points. anD

(20:14) Shiyan Koh:

But then how does it feel if you're being broken up with and

(20:18) Jeremy Au:

You don't know, that's why it's a phone call. The guy's ringing off like, you know,

(20:21) Shiyan Koh:

No no no, but it's a phone call but you've been in a relationship with someone and they suddenly start talking about this stuff and you're like, where did this all come from? We haven't talked about any of this before. Doesn't it feel fishy?

(20:31) Jeremy Au:

You're like, you've been emotionally unavailable for the past three years. And suddenly you have an emotionally available breakup. Sorry. That's so funny. Yeah that's kind of screwed up, but it's hilarious. I mean, yeah. Reminds me of the time I was in a train recently and I was just, you know, I was just like in a train and it was a bit crowded. And so I look over around and there was this guy, clearly a white collar, young guy probably say a fresh graduate. And he was like, no, honestly, he was just busy flirting with a bot called Cheating Crush.

(20:57) Shiyan Koh:

Cheating Crush?

(20:58) Jeremy Au:

Yeah, I know. I mean, I could just call it Elizabeth or something. I put it in a WhatsApp. He works, but TA was called Cheating Crush. He was throwing a water bottle at her. She was kind of giggling and saying, why are you so mean to me? And he, you know, he was like, oh, I like being doing this to you and teasing you, and then shoot. I use the word she in a sense that she's a

(21:13) Shiyan Koh:

The, the bot.

(21:14) Jeremy Au:

Yeah. And then I was like, well, that's how humanity ends. I guess. Anyway.

(21:20) Shiyan Koh:

It framed as a game? Or is it?

(21:22) Jeremy Au:

No, I, this guy was serious. Yeah. You know, it was a 15 minute MRT ride where I was just like, you know, and it was interesting actually, because if you think about it. It's actually socially acceptable, I'm just giving you an example, to, to WhatsApp your wife or your girlfriend, I mean, say somebody, right? I mean, if you had WhatsApp somebody, you know what I mean, and you said all that stuff, and the person was a human it'd be totally socially acceptable, right?

But I was just like, my eyes popped out, because again, it's crowded, I mean, and this guy's So to him, maybe it was like socially acceptable to do this. I don't know. But for me, I was like, whoa, this is like in the wild, man. This is you know, driving your billion dollar revenues, you know, for character.ai and all these other folks, right? Which is this person knocking off work and he's feeling lonely and, you know, he's busy flirting with AI avatar, right?

(22:01) Shiyan Koh:

That's so interesting, I guess. So what do you, I mean, normally, I mean, I would say I'm not a very good flirter.

(22:07) Jeremy Au:

Good news for you. There's somebody who's going to be 100 percent receptive.

(22:09) Shiyan Koh:

But the goal of flirting generally is to advance your relationship with someone, right?

(22:14) Jeremy Au:

Yeah.

(22:14) Shiyan Koh:

So what, in this case, what's the objective? Is it to get better at flirting so that you can then apply these skills to a human?

(22:22) Jeremy Au:

I don't. I think that's a great reflection I had recently. And I was just going through some of the science here, but, I think relationships are dropping. Romantic relationships are dropping over time. There's more loneliness growing over time as well. There's a loneliness epidemic, as they call it, but I think my perspective is that as a human, people say they want relationships, but more importantly, they want things that come out of a great relationship, so they want companionship, affirmation you know, a buddy, right?

(22:45) Shiyan Koh:

And so you're saying you can get all of this from an AI bot?

(22:47) Jeremy Au:

I mean, the games and entertainment industry is giving a lot of that. I mean, we used to make fun of Japan, right? Otaku and waifus and all these shut ins. And then, they're all lonely and they're all at home and they're stuck at home. And I was like, you know, Japan's just an early adopter, right? It's just, at the end of the day, the truth is real estate prices are still the same, but the cost of computer has dropped by an order of magnitude, right? And so, I was reading this fun YouTube comment, but it was like, don't be tricked! This is the cyberpunk dystopia that we had, which is you know, and I was like, you know this is true, actually, it's it just takes time to get there, but in 100 years, I think real estate will still be about the same price, but the cost of compute, whatever you call it, AIVR, will drop by several orders of magnitude again.

And, so as a result, I think by nature, people are going to move more of their life into social parasocial relationships, AI, so, so forth. So anyway I actually am a big believer that I saw this guy and I was like, you know what? I totally get it, man. I marry it. And I totally, I see the value I'm just saying, right. It's

(23:41) Shiyan Koh:

Does this help the incels? Does that make people less frustrated and violent?

(23:44) Jeremy Au:

I mean, I don't know. It's what do you feel about pornography, right? You know, we went from zero to widespread on demand availability. So, but you know, pornography was the sex. So I think ChatGPT relationships would be to relationships. You know, is this if on demand affirmation and acceptance, I don't know. Just think it's going to change society the same way birth control did. Will be interesting.

(24:02) Shiyan Koh:

It's a topic for a different podcast, maybe?

(24:05) Jeremy Au:

I mean, we're keeping this to the AI side. I mean, no, okay. There are so many business models. Okay. There is a literal model company startup that's pitching itself as only fans with AI is basically,

(24:15) Shiyan Koh:

Yeah, you don't even need to pay the people, right? It's like a much business.

(24:19) Jeremy Au:

And, you know, they've already grown to 10 million of revenue in three months. I mean, it just turns out there's a lot of lonely people out there. You know, so, I mean, you know, so I think that's a, is a real business opportunity, but it was a real. Change your society, right?

(24:31) Shiyan Koh:

It's interesting.

(24:32) Jeremy Au:

Yeah. And I think that is great slide in the Coatue AI report, right? It was like looking at character dot AI, right? There's a saying like, hey, this thing has grown like crazy, but nobody wants to talk to Martin Luther King, right? It's kind of a great speaker, but, so they're saying, within effectively one year from zero, they've now effectively got to about, 35 minutes per visit, about what 12 million monthly active users. So there's all kinds of crazy stories that's happening and this company never raised much venture capital either for character. Yeah. Something really interesting.

(25:02) Shiyan Koh:

The world is crazy.

(25:03) Jeremy Au:

Yeah. I guess from your perspective would be what's the one takeaway that you think that's pretty important to walk away with?

(25:10) Shiyan Koh:

It's happening, and you can't stop it, and so you should play with it.

(25:14) Jeremy Au:

Ah I've this cyberpunk quote line. There we go. Anyway.

(25:18) Shiyan Koh:

I mean, I think you should use, I think everyone should use it. I think that it will actually really change how they think about their work and what can be, what is possible. I think it actually changes your mind about what is possible. And that, I think, is a fairly exciting and liberating thing.

(25:31) Jeremy Au:

Yeah. I think for me, the key takeaway is similar to yours, it's like the S curve for this would be much faster because building out the internet took a ton of work because got to lay cable and build wifi networks, smartphones took a lot of time because you just got to get phones into people's hands and make the price lower. But AI is just like a digital online only function and it's creeping into every app. So I think the adoption curve is going to be much faster. So it would be interesting to see where the value capture goes as well in a chain.

(25:55) Shiyan Koh:

I mean, it's gonna be frickin amazing.

(25:57) Jeremy Au:

Unless you're an SEO marketer.

(25:59) Shiyan Koh:

I guess.

(26:01) Jeremy Au: X or Z.

Yeah. Anyway isn't that a thing? Is this basically everything we see in technology about steroids? Honestly.

(26:06) Shiyan Koh:

Yeah, so I think, I don't know, I'm pretty excited about it, but I would just encourage everyone to try it and use it.

(26:12) Jeremy Au:

Yeah.

(26:13) Shiyan Koh:

And new stuff is coming online all the time. So I think there's just like remixing And there is an aspect of teaching yourself stuff too, like you can upload PDFs and start interrogating the PDFs.

(26:23) Jeremy Au:

Right.

(26:24) Shiyan Koh:

Which I think is also an interesting thing. Sometimes, I read scientific papers, and I don't necessarily understand all of it. And it's like a way to help get deeper into it as well, and so don't know it's like very exciting, right? We grew up, well, I'm older than you, but I grew up before the internet. I still remember the day of encyclopedias, whereas my kids don't think about that at all, right? My kids think Google knows everything. That they can google anything and they can find it and that's like pretty much almost entirely true. And then I think AI just makes that orders of magnitude better.

(26:50) Jeremy Au:

Yeah, I agree. On that note, I'd love to summarize the three big takeaways. First of all, I thought it was fun just to talk a little bit about open AI and the board shuffle and the eventual return of Sam as CEO. But I think it was interesting just to hear the dynamic about. What we think will happen as a result.

And I agree with you. We believe that there'll be increased acceleration and commercialization as a result on SMS full control about it. The second part that we talked about was, I think we talked about the Coatue AI report and we talked about what we liked. I think we liked the part about scaling organizations in terms of their functions.

We talked about some of the facts that we had around the growth and productivity of various jobs that's going to happen. So that was a interesting discussion about also where value is going to accrue and how fast you need to be part of that journey. And lastly, we talked a little bit about society, about how there's going to be some winners and losers.

I think winners is if you're able to be part of the corporations that turbocharge and use AI a lot. If you're a profession that's easily displaced by generative AI, I think we're starting to see the changes in that happening, right? And we can also talk about how as a result, society is starting to change all the way from how courses are going to be graded in the future, away from papers all the way to how people are going to have relationships as well. On that note, thank you so much, Shiyan.

(27:54) Shiyan Koh:

Thanks Jeremy!