Singapore 2023 Budget Winners, Crowding Out Startups, Generative AI Barbell vs. Secrets vs. Judgement, Localization to Southeast Asia Opportunities, Agent/ Concierge Displacement, Entertainment Love Avatars & Digital Ghosts - E243

· Blog,Singapore,Southeast Asia,Podcast Episodes English

“You have to make sense of what the facts are telling you. That's still quite hard for ChatGPT to do. It has the entire internet, but it doesn’t have a voice. That’s a human quality. It might get there because it can imitate the voice of other things it finds, but insight is the challenge. It’s like investing wherein people always think we already have the technology to automate numbers, but there's a vast field where everyone's looking at the same numbers but making different decisions because judgment and point of view matter.” - Shiyan Koh

In this episode of BRAVE, Jeremy Au and Shiyan Koh delve into the fascinating topic of crowding out younger startups and its impact on the startup ecosystem. They also explore the potential of generative AI to shape the future generation, and how Singapore's 2023 budget applies to Southeast Asia.

Supported by Esevel
Do you manage your own IT for distributed teams in Asia? You know how painful it is. Esevel helps your in-house team by taking tough tasks off their hands and giving them the tools to manage IT effectively. Get help across eight countries in Asia Pacific, which includes onboarding, procurement, device management, real-time IT support, offboarding, and more. Gain full control of all your IT infrastructure in one place with our state-of-the-art platform. Check out and get a demo today. Use our referral code, "BRAVE" for three months, free. Terms and conditions apply.

Jeremy Au: (02:20)
Hey Shiyan, two big topics that seem to be on everybody's minds in Southeast Asia. The first is, the Singapore budget since so many people are impacted by all the cash that's floating around. The second of course is generative AI. So lots of stuff about the press. And everyone's asking like, how does it apply to Southeast Asia? So let's talk about our favorite dollars, dollars, dollars. What do you think?


Shiyan Koh: (02:43)
Show me the money!


Jeremy Au: (02:45)
Show me the money!


Shiyan Koh: (02:47)
Yeah, I mean, I think it continues to be a fairly pragmatic budget, which is in line with what we would expect are a number of schemes that have been, I guess, topped up is what we like to say here in Singapore. And I think maybe aside from the dollars, what is perhaps interesting is to ask the question like, what is the desired outcome of the schemes?

Do we believe that they actually drive the behavior we are trying to get? And so that's, I think, the more interesting question for me. If you think about ecosystems, we always say capital, talent and market. Those are the things that you need for a thriving ecosystem. I would add one more, which is mindset. I actually think Singapore and the region actually has a fair amount of capital. And so I don't necessarily always think money is the problem. It's not that lack of money is stopping people. But on the margin maybe you can incent some additional behaviors. I think the money is actually more important to impact the mindset rather than the physical dollars themselves.

It's more of a signal to say, "Hey, this is important. We care about it." Parents, don't worry if your kids are doing this crazy startup thing and innovation's important to us. I think the signaling value is actually more important than the actual dollars. I'm a little bit skeptical that Steve Jobs didn't write a hundred page grant proposal before deciding he needed to invent new stuff.

That's not the path. So yeah I'll just put that out there first, which is I think the question is on the margin, does it change behavior? If you weren't going to expand overseas, does the existence of the market expansion grant make you expand overseas? Should it? And if you were going to expand already, then it's just cash and it didn't really change your decision making.

And so I think that's like an open question. I think some of this money actually funds more knowledge sharing and network programs to help people build networks in new markets. And I think that's probably to me, more valuable than the literal cash, although no startup is going to be like, yeah, don't give me extra money. But yeah. I'm curious what you think, Jeremy.


Jeremy Au: (5:04)
Definitely agree with you about signaling and I think it's not just externally and to the founders, but also to the government as well. I've definitely known a founder that was expanding to Japan. Some of it was financial support, but a big chunk of it was actually getting help from the ambassador and the local embassy to actually build the connections because Japan's not an easy market to enter from outside.

I think there was a nice priority signalization that is kind of like money talks and action is really important. So like you said, I think a lot of it is knowledge sharing, I think where I defer view slightly is this part about does it really matter as well in terms of creating outcomes.

And I think I look at it in two different ways. I think the first is Singapore obviously is a strong country, actually in terms of GDP per capita in many ways, and education outcomes to Malaysia and Thailand, right? In Southeast Asia. But I think the capital availability has been pretty high, right? And so some of the moves for a government to not only bring in investment, but also to incentivize that investment to go toward startups has jump started the ecosystem with a high keep capital availability, which again kicks us the whole flywheel of accelerators and cultural change for founders to come up on one hand.

And the second hand is, I think the world is moving towards more subsidies and protectionism. I was discussing with a Harvard professor and his point of view was, for example Singapore has historically been a semiconductor place as part of the global supply chain.

Especially when semiconductors are relatively global and distributed. Since the decoupling, America's desperately trying to restore it into themselves. And then China's also trying to restore semiconductors to themselves as well. And basically both China and America are both throwing a ton of money. And so I think the truth of the matter is that America and China can throw way more money at their own factories than they would. Singapore could, right? For its own domestic semiconductors. And this is paralleled in terms of cash subsidies. But I think the other angle that Singapore primarily has done is through tax minimization, right?

So tax rebates and credits. And with the recent global tax reforms driven by president Biden good for America for making sure there's less tax avoidance from MNCs. But I think it removes a strong pillar of economic attraction, right? For multinational companies to come to Singapore and actually a very large chunk, 4 billion under this national productivity fund is actually just devoted to attracting MNCs that were primarily in the past been attracted by tax credits. So I think this might be the new normal. Yeah.


Shiyan Koh: (07:44)
Yeah, I mean the semiconductor point, I think that's a much longer conversation, right? Which is that I think on the US side, you've got the R&D talent, but you don't necessarily have the production talent or the process knowledge. And then of course, the US is actively trying to deny that process knowledge to China on top of denying access to the actual machinery that produces semiconductors, right through their targeted sanction, sort of blocking programs. I can't remember the name of the actual program. But yeah, I think it's a worthy question to ask what Singapore's role is going to be in that? Well, we did have a big semiconductor project in the eighties.

We had National Semiconductor which we sold to the Europeans. And so I think there is a question of what is the sliver that we can own that makes sense and is strategic and isn't just buoyed by subsidies. But yeah, I think the tax, the decline of the tax credit, I guess as a carat is an interesting one.

But cash alone I think doesn't bring people because you still need these other pillars, which is you still need talent and I think this has been a complaint on the tech side, which is we offer a bunch of carat for MNCs, big tech companies to come. We tie it to employment goals. You need to hire X number of Singaporeans or people onshore in those roles and the MNCs run to hit those targets.

They can pay way more than startups. They essentially crowd out startups from the talent market or raise salaries across the board, which is really challenging for startups as well. So I think they're the second and third order consequences of some of these actions that I don't know if we necessarily always think through or see when the policies are first rolled out.


Jeremy Au: (09:37)
A hundred percent agree. I think it also changes the quality and the type of jobs that are available. So a lot of multinational corporations, they're building out sales or local marketing, but not necessarily locating engineering or key product management roles in Southeast Asia.


Shiyan Koh: (09:52)
That has changed.


Jeremy Au: (09:53)
That's starting to change, yeah.


Shiyan Koh: (09:55)
Stripe is an engineering center. Google, has a bunch of product and engineering teams here around their NBU business. And I have seen a number of people whose programs are actually tied to technical talent. It's not tied to sales and marketing.


Jeremy Au: (10:10)
Yeah, I mean, I think the roles are changing.


Shiyan Koh: (10:12)
For the most actually.


Jeremy Au: (10:13)
At a junior level. But now I was talking to someone who is in the government and his perspective was the local startups that have grown right from C to series A to growth stage and to going public. They have the whole spread.

But they also have the executive leadership and they also built out a whole bunch of talent that kind of have that networks. And they also have a long-term interest in growing the local talent pipeline in a different form and fashion. So I think there's an interesting dynamic, like you said, where it's like you said, there's a crowding out of young startups, but the issue is not not crowding out young startups.

You're crowding out future unicorns or large companies that could come out of this crop of young startups, right? That can really build up that full executive leadership. So I think it, that's that crowding out I think is much less obvious, right? Because nobody cries about a young startup not being able to do it. But if you crowded out a future creative or Razor or Grab or Gojek, then I think it's much less obvious.


Shiyan Koh: (11:29)
Well, I mean, even Grab and Gojek, all these guys, like they don't have the bulk of their technical talent on short in Singapore because there's just not enough people, they all have engineering centers in India, in China, in Vietnam, because you just can't get enough bodies. Of course Singapore wants to attract talent to Singapore, right?

That's the goal of a government. But there is a sort of positive knock on effect for the region as a whole. So if there is a model of a Singapore startup where maybe you have some core senior talent in Singapore, but you are much bigger, sort of engineering teams are actually in Vietnam or other places, it's not the perfect, it's not the platonic ideal perhaps that the policymaker thought of when they designed it, but actually for at the ecosystem level it actually could be long term.


Jeremy Au: (12:04)
I agree. And I think, side note, of course, was I thought it was interesting to have a lot of these productivity solution grants. Local, small, medium enterprises to upscale and change. I think productivity continues to be a bugbear for, not just for Singapore, but Southeast Asia as well. In terms of productivity for the small medium enterprise, I think.


Shiyan Koh: (12:24)
I think this is a mindset problem, I don't think this is a software problem, honestly. I mean, have you seen some of these grants? They're like, "Hey make a website. Hey, use a CRM." I mean, I think software's the least of their problems. It doesn't matter if you've implemented the software, if you don't have the mindset to think about, "Hey, how does this software accelerate my business?" I think in some instances, these grants, while well intentioned, lead to kind of people trying to game the system because it's a subsidy companies who are accepted into the program will mark up their list price and then the net to the company is the same, right?

As if they bought it not through the program. But the company providing it gets this markup that the government is funding. I would be very curious to see how we measure after these SMEs implement this software, is there really an increase in productivity? I'm skeptical.


Jeremy Au: (13:27)
Yeah. I think it's subsidy, but also I think competition is needed. And I think one example was I was working with a second generation in luxury distribution and sales and there was a big competition because he's actually trained in tech and he's just in a catch-22 where he's like trying to implement omnichannel online sales. And the old guard, and his father and so on, so forth. They're very focused on saying, well, we'll let you do online sales if you can prove to us with data that online sales works. And then he is like, "That's not doable unless we change the system. If we change, we have a website and all these things." So like you said, it's a mindset, but there's also a generational change dynamic, right?

And I think it just boils down to the leadership saying like, "Hey there's a burning platform and we have to change." And that's when the subsidies would be helpful.


Shiyan Koh: (14:18)
Yeah, I mean, I don't know. Do the thought experiment, right? Would you be better off sending all the towkay to management training versus subsidizing them buying software?


Jeremy Au: (14:30)
I think they're subsidizing the towkay’s kids who are ready to take over. An d that's how you get your mindset changed. You're the ones who are willing to prove that there's a new fresh way and take it forward. I think there's a generational shift actually.


Shiyan Koh: (14:44)
I think that's probably, hopefully what will drive it. But I think the SME mindset also is how do you go from a sort of scarcity breakeven type of mindset to like, okay, what is a step function change we can drive in this business? And I think that's actually much harder because you don't necessarily have the talent, even if the next gen is very capable, they have to revamp the whole staff, not even their own father or mother or whatever.

It's like whoever is in the business today has some older way of doing things and you have to basically, retrain them or you need to turn over that staff population. And I think that's hard. But maybe it's an opportunity for a rollup.


Jeremy Au: (15:34)
I have faith, we have in this generation's towkay kids and rollups for everybody else, you just get quiet and cash out and you sell it to a nice roll up approach that will give the efficiencies and tech.


Shiyan Koh: (15:47)
No, because everybody always thinks their thing is worth more than it is, and they think they want to give it to their kids. So I think, yeah, even roll up. Not so easy.


Jeremy Au: (15:59)
True. But sometimes the kids don't want to take over and so they say, might as well cash out and play golf and then divide the cash and let the kids do whatever they want in their life. That's, I think what the roll up folks are finding is a compelling sales pitch to it.

Which is like, hey your kids don't want to do it, right? Let us take over. I think we've definitely seen that pickup actually quite a bit in Singapore, for sure compared to the region for now. So it would be interesting go-to-market, I think for a lot of startups as well. And on that note, obviously the big question about labor and turnover and so far is ChatGPT and generative AI. And I was talking with some creators today and they said like, "Wow. This felt like landing on the moon kind of moment, right?"

Because they typed in and they said, "Hey, can you generate a script for us?" And then put a bunch of localization requirements and it got sped out in three seconds and it was not great, but it was pretty serviceable as a first draft. And they were like more importantly it was done for effectively free, right? In three seconds. So versus if they were just to brainstorm, it'll take them at least an hour just to get started after the procrastination. So, yeah. What are your thoughts about generative AI and everything at a higher level first, before we talk about Southeast Asia?


Shiyan Koh: (17:10)
I mean, I think it's awesome, right? Have you seen the graph of the ChatGPT adoption? It's like literally straight up, it's for startup usage of products. How long did it take to get a hundred million users and there's like Facebook, Snapchat WeChat, whatever.

And then there's literally ChatGPTs. It's just like a vertical line. That's how fast it's gone. And so I think that's like crazy. My view of it is that it basically is going to make the bottom of the, or the left end of the distribution. Like it's going to move the bar on that, right? So if you are a manual, lower-skilled person, you will be replaced by ChatGPT because the quality of your work was sort of like that, very serviceable left end of the distribution anyway.

And now you've been made free, right? So there's a free alternative. I think the high end is still going to be fine. So if you are Picasso, you're not going to get replaced. Because you are coming up with original stuff. You're not just doing something derivative of everything else that has been available on the internet. And then in the middle, ideally you can be using ChatGPT-driven tools to make yourself more effective to increase your productivity. And so then it's about like other skills that are not core to the functional expertise, managing teams, delivering outcomes, blah, blah, blah, that kind of stuff. But it's fascinating.

I mean, I was talking with an investor and she was saying, oh, I have an analyst that you know, does translation for me. English, Chinese, Chinese, English, and I tried some of the documents in ChatGPT and it was 95% there, and she said I might replace my analyst because ChatGPT doesn't have hopes and dreams to manage, doesn't have a bad attitude.

So that's what I mean, right? I think the soft skill stuff, which if the stuff you do is like pretty routinized and you don't have a good attitude, you're going to get replaced. But then that's when the other stuff becomes more important, which is ChatGPT is just a tool. It can't deliver an outcome unless it is harnessed effectively.

And so how effectively can you harness it and other resources to deliver an outcome for your enterprise, your team, whatever it is? But I'm really excited about how this is going to open up lots of things. And there are things where somebody built a little demo where there was a gardener who was not very good at English, and basically a ChatGPT interface for his customers. But he's a good gardener and it basically just made it much easier for him to interact with customers to build his business, to invoice on time, things like that because it sort of removed that language layer. And so I think that's pretty cool, right? That you can improve other people's earning potential just by easing these communications. And so those are some interesting sort of initial demos that I've seen separate from the, like, hey, make crazy art or compose bad love poems on Valentine's Day.


Jeremy Au: (20:37)
Well, those are always fun. And it was like you said, It was mind blowing. I was playing around with Midjourney. For the picture, and I'm not a designer or illustrator, but there was some stuff that I made really great in 20 seconds. And I was just like, mind blown. And I think the superpower here is you make somebody who has no expertise in something, get to a pretty average level. But if you're, like I said, if you're average or displaced, and actually I'm a little bit more pessimistic about what you said because I think you implied like, okay, yeah, if you're bottom you get replaced and then in the middle you still have a chance.

And the top, obviously you still have a calling and passion. But I think we're going to get barbelled a lot more. Which is I think this is I don't know the mechanical weaving loom. It's like historically, everybody all weaved themselves. Everybody grew their own cotton. Everybody weaved themselves the fabrics and everybody sold their own clothes and suddenly, I think you know, this mechanical loom basically displaced all the weaving. And I think there's just going to be like factories where there's going to be like ChatGPT being scaled to millions of billions of pages.

I'm just saying. And I think there's going to be giant factories of content mills that's like spamming the internet. And I think you're right. I think, and that's, I think going to eat up more. So the middle, I think the middle is going to disappear a lot more within that 10 layer. And I think, like you said the high end will still exist, but I think it's much thinner. So I was discussing with a writer, right? And he writes business article. So he's basically saying like, okay, this month the macroeconomic change because of these three key factors and whatever. And we talked about it. He said, yeah, that job is gone. He is like, he can do that job a little bit more in Southeast Asia for another one year or two.

But there's nothing stopping an American person to simultaneously create a daily updated version for every single country, every single market, every single vertical at the same levels as he does. And so he was just chatting like maybe what he needs to do is he needs to go into like autobiographies, right? Because it's hard to do.


Shiyan Koh: (22:32)
Well, no, no. But I think there's a difference though between information and insight. So yeah, if his writing is just regurgitating macroeconomic facts, yeah, he's going to get replaced. But presumably for a business reporter, you have some insight, right? Maybe not everyone is Matt Levine from Bloomberg, you have to make sense of what the facts are telling you. And I think that's still quite hard for ChatGPT to do, right? Which is it can pattern match. I mean, it has the entire internet, right? So it can pattern match off the past and it can say this thing is like this thing, but it's not that funny necessarily. It doesn't have a voice.

Those very human things, human qualities. And maybe it'll get there, right? Because it can imitate the voice or of other things it finds. But I still think insight is the challenge it's investing, right? People always want to think it's like numbers driven and it's like, we already have the technology to automate numbers. And there's obviously like high frequency traders and things like that for people who trade very quantitatively. But there's a whole vast field of investing where everyone's looking at the same numbers, but they're actually coming to different decisions. They're making different calls and different insights. And so judgment, I think, really matters, point of view matters. And so I'm like maybe like slightly more positive.


Jeremy Au: (23:57)
My counter argument is that I think you think too highly of most of the writing out there.


Shiyan Koh: (24:01)
Oh, no, I don't like most of the writing out there. I would say most of the writing's not that good. They gotta step up their game, but they're actually capable of that. It's just nobody has asked 'em to do that. But it's like sports writing, right? It's Team A played, team B, player team A, player team B.

It's like boring, right? Like yeah, a machine could do that and they could do that before ChatGPT people were already producing sports news with automated scripts, but really great sports writing. Like it's funny, it's exciting, you know it, it's like it's worth reading. But I agree, the vast majority stuff today is dribble. It's boring and it's mediocre. It's not interesting. But this is maybe a call to these folks to actually make it more fun and interesting.


Jeremy Au: (24:45)
I agree with you that insights for me, by the way, I don't call them insights. The way I’d be framing is, I call it secrets, like even secrets which is a form of insight. I think that's where you'll beat the AI, right?


Shiyan Koh: (24:57)
People are insightful, is what you're saying.


Jeremy Au: (24:59)
That's what I'm saying. And I think a lot of, like humor, for example, I love improv and comedy. I think a lot is actually formulaic. And for most people it's great. I mean, I always remember watching America's Funniest Home Videos. And I love it as a kid, very formulaic. When I watch this as an adult, I'm just like, okay, A lot of people getting injured but not dying. It is kind of like a very good formulaic humor type for kids. And I also agree with you that judgment is key. I just think that's hard. And I think, as you said, people have to flee away from descriptions and pattern matching to insights and judgment.

But let's see. The question of course is, how does that apply to Southeast Asia? Because I think a lot of folks are like, okay all these news things be coming from the US and there's a lot of US press and a lot of these, a lot of US-affiliated folks, including ourselves in Southeast Asia, are playing with it. But what does it mean for Southeast Asia? For me, I'll just start off with my first thought which is, a lot of this fundamental research is being done in the US and in China, right? And so there's an awkward reality where it is a lot of the gains for the fundamental learning models, the fundamental research is being done in the US and China.

Russia's nowhere near a good example. Like historically it's been good at computers, but they're nowhere near artificial intelligence or these general understanding machines. But lots of countries around Southeast Asia just they're all in the US or in China, right? Building these AI machines. So I think a lot of value is going to be captured by the folks who are actually building these learning models. So I'm thinking through what else is the implication for Southeast Asia? What are your thoughts?


Shiyan Koh: (26:35)
What does ChatGPT enable people to do? It enables people to generate ideas, right? By drawing on a body of historical content. It allows you to like, summarize and synthesize blocks of text effectively. And so I think if you look out where it's been rolled out pretty quickly. Notion rolled out an AI feature. Microsoft, of course, with their investment has talked about how it's going to be integrated into Word and all of their productivity suites. And so I don't know a very silly outcome is maybe people will write better emails, maybe there will be better English.

It's not exactly world changing. I'm curious to see how it shows up in sort of business tools, right? So it's not going to be a standalone product. It'll be integrated into business tools. Maybe it'll enable better interfaces and communication. I think on the art and generative art stuff, like I would expect you to have it in Canva, in Adobe, in all of these like standard authoring tools to get you started on thinking on your project before you kinda actually do the work.

We have a portfolio company here in Singapore actually, that integrated some of these features into performance reviews, so people hate writing performance reviews. And it basically assembles all the things that feedback you've given to an employee over the course of the review period, and then it generates the draft performance review.

So you just go in, and edit it rather than have to write from scratch. But I think the fact of the matter is we're still in the early stages of SaaS here in Southeast Asia, and so I know it's harder for me to imagine where would Grab integrate generative AI into calling cards. I don't want you to get creative. I just want you to get the car to me now and I want you go to my destination. I don't want you to do anything else. But I think down the road, I could see use cases in underwriting. So FinTech, in medical, right? Taking sort of all this like data and structuring it. But I think maybe the second or third order effects is have you seen GitHub's co-pilot. Like helping more people level up and write better code. Maybe down the line you take all these bootcamp grads, you take all these young junior people, you give them this kind of tools to supercharge them. And then that kind of accelerates their learning and enables them to create more and better things. But I don't know, my imagination's kind of sucking this morning. What are you picturing Jeremy?


Jeremy Au: (29:32)
I think there's some interesting opportunities. I think one that I can think of right now is concierge services. So historically, you could build a relationship with someone and you have repeated transactions over time and you build like a portrait of UN that happened human to human, right? Insurance agents. And it's a good example. And most companies gave up on that because that's for high net of individuals, et cetera. But for most people there's transaction by transaction, right? Because it's easier to automate the slice of the transaction, weirdly generative AI plus knowledge of history, et cetera, could actually generate quite a high fidelity concierge that couldn't have existed.

And it can feel like a concierge and more importantly, would be much more instantaneous and reliable than the actual concierge, because concierge unfortunately are only available nine to five Monday to Friday. Saturday at 4:00 AM, your insurance agent isn't going to answer a question for you, but it generative AI, I don't know whether your insurance agent is using a generative AI at that point of time, but it could help the insurance agent project that, oh, I totally saw this at 4:00 AM and I totally send my message at 7:00 AM with a very targeted thoughtful response and then the person walk out in the morning, edit it a little bit and send it out. I think there could be a real super powering of the concierge where it used to be a human actual phase, but in the background it's actually a bunch of scripts to make them five x smaller.


Shiyan Koh: (31:01)
Sure. For sure, for sure. Yeah, so all of these sort of like agent-driven, service based industries, right? But I think it always goes back to this control, like how much control do you have over the core training set, right? How much do you allow the individual human to change things? And because it's definitely not a hundred percent AI, right? Because you would take the human out of it altogether, and then I think humans can tell it's not a human. But is it 5%? Is it 30%? I think those things are interesting. Like I did a project for Stitch Fix where there's a stylist, but there's an AI that generates the clothing recommendations. And I think the scientists would want it to be a hundred percent AI. But I think the reality is that humans actually like having a human who will pick out things that an AI won't notice and tweak things and adjust them. And maybe over time you hit a point where you don't need a human anymore. But I don't know. I still think that people can tell the difference when it's not over repeated interactions.


Jeremy Au: (32:10)
Yeah. Well I think that might be like the fuddy duddys who care about that. And I was like watching Dota 2 finals, and I was like, I remember it was like this whole arena, right? And there's all these nerds. And I was there because I used to play a long time ago. I'm not as active these days, but I was like in this arena and I was thinking myself, oh my God, the nerds have become cool, right? Like we have the million dollar prize, we have the fireworks. And a generation ago my parents just thought that computer games were like a really bad sport and nowhere an acceptable game for kids to play. And the next generation now has mobile games, right? They don't even do computer games. And so I feel like you're right to say like the current consumer who's buying consumer stuff wants some human content or more human content on average. But I expect that the next generation, as they get younger and they become more AI native, in that sense, I think they require less of it.


Shiyan Koh: (33:04)
Well or no, or it becomes handcrafted artisanal stuff becomes even more prized thing.


Jeremy Au: (33:12)
Oh, for sure.


Shiyan Koh: (33:14)
Like you can generate anything, but like, hey, you spent 100 hours knitting this thing, right? It's hand knit. Like that stuff costs more, right? And so that performative quality of making the thing yourself will become more valuable.


Jeremy Au: (33:31)
I think so too. It would be much smaller. I was just thinking to myself like back in the day leather handbags were just mainstream. You had a cow and then it, it was nice to you for 10 years and then you killed a cow and then you made it into a bag. That was like our parents, our grandparents' time, right? But now leather is this luxury good in handbags. Nobody uses ballistic nylon or set cloth or their handbags. But now, there's only premium letter left. Mainstream letter called old Betsy doesn't exist anymore. I think it'll be a much smaller quantity. it'll be much more luxury and much more performative.


Shiyan Koh: (34:07)
But maybe you know all those people who don't need to do their jobs 40 hours a week anymore, cause generative AI is doing it for them will have free time to then handcraft more.


Jeremy Au: (34:17)
Free time to what? Cause they have free time to fall in love with their generative AI romance partners. Which by the way I think it's going to be a real thing. I literally just read an article about, this person, he was grieving over his dead fiance and he trained the AI to speak with the voice, right? To some extent, right. And the style of his dead fiance. And so he used that to process his grief and so forth. And obviously the story ends with him moving on because he uses that to get his life in control. But then I was like reading the story, I was thinking to myself, I don't know. My grandkids could fall in love with generative AI.


Shiyan Koh: (34:55)
Yeah. I mean, you could fall in love. But it's like anything else, right? You could fall in love with a voice and a point of view, but like you can't have lunch with generative AI. You can't play tennis with generative AI. Generative AI is not going to keep you warm in your bed at night. Do you know what I mean? Come on, man.


Jeremy Au: (35:14)
I feel like I know for sure that I'm going to be part of the generation that's going to be like leading, this countercultural movement to keep love human or something like that, right? It's less about the fact that I think it's going to happen to some extent, but I just feel like there's going to be a counter reaction to it as well. So I think it's going to be an interesting dynamic, right? It's what we talked about.


Shiyan Koh: (35:32)
Yeah, but I mean, it's like people are already obsessed with anime or movie stars or whatever. Like the human brain has incredible capacity to like, create connection to distant things that have no relationship, no real relationship to you, right? BTS. Does BTS know you exist?

But BTS loves to sell you stuff, right? But you can still feel all this stuff, right? But I mean, there will just become more and more versions of this thing, and maybe that becomes atomized, right? And it's not that everyone is obsessed with one rocker, you are obsessed with your own produced version of things.


Jeremy Au: (36:06)
Wait, you and I just came out with a billion dollar idea. We sell a generative AI that pretends to be BTS. We have the video, the syncing you feed it your biodata, and then the BTS guy just falls in love with you organically. That's a billion dollar idea there.


Shiyan Koh: (36:23)
I mean, why do you even need Cameo anymore? Right? Like you could do Cameo without the human talent. There's enough material out there that you could make videos of any famous person saying whatever you want. Maybe you give them a license fee, but you don't have to pay them the whole thing cause they don't actually have to spend any time anymore physically.


Jeremy Au: (36:45)
And I think I believe in that, and I think one of my things, I believe is that it's going to displace a lot of creative talent. So Darth Vader already sold his voice rights, right? His voice can be used forever and ever and is being cloned forever. But I think unfortunately, of course Leia she passed away. But there's a lot more Star Wars movies in the future. And they already generated several versions of her being de age or CGI-ied already, and so to some extent, actually I think historically a movie could not continue the franchise with that actor or actress because that person's passed away.

And that role would have had to be given to a young, new person, new blood. But that role is going to be continued to be filled to some extent, amount of screen time. And so I think one thing that hasn't happened is since the original trilogy. Luke Skywalker, Han Solo, and Leia have not reappeared as a reunion since the original trilogy. And my big prediction is that it's going to happen again. Disney is going to pay enough money to make this digital recreation. It's going to happen, in five years or 20 years. But, I think all three of them will sell their digital rights and then we're going to have a reunion and they're all be a hundred percent generated. It's crazy.


Shiyan Koh: (37:58)
Yeah, it's an interesting question, right? Do we value those as much? Does it matter? It's like how, you can have sequels, I always feel sequels are not as good as the original, but maybe that's me being a cranky old person. And then does it get diluted, you're on the 10th Star Wars or the 20th Star Wars episode like that's a hard one.

And then you lose the promotional activity, right? If you don't have a physical live actor, then who do people go see? You could only see the movie. You don't have other sort of spinoff things to do with it. It'll Be interesting though. I'm curious to see how that goes.


Jeremy Au: (38:42)
I mean, Disney has no problem with infinite copy and paste of Mickey Mouse, right? And various characters, right? And lifetime value is long, right? You still have kids and you keep going, and there's an amusement park. There's so many ways to consume that same piece of content in a different context.


Shiyan Koh: (38:57)
That's true. There's that Bob Iger interview where someone asks him can IP be overexposed? And he was like, no. Good stories, like you can't have too much of them. And I was like, do you have a small child who's made you watch Frozen a million times? Because I really think you can have too much exposure to IP. But yeah, it's, a good question. Exciting times ahead.


Jeremy Au: (39:21)
I mean, children actually psychologically cannot get overexposed stuff. They love repeating stuff. Adults have a shorter attention span slash /ability to see repetitive content. On that note, actually, I was thinking to myself recently, you and I actually now are immortal already digitally, so that's something I was thinking about. Both of us have appeared online. We have our voices, we've talked, we've bantered, we have our WhatsApp messages, we have our email.


Shiyan Koh: (39:49)
You want to do an episode just with generative AI versions of ourselves? Like we don't talk, and we just see what it comes up with. And then we have the audience judge, like, does this sound like Jeremy?


Jeremy Au: (39:58)
I mean, there's no law preventing you from generating a generative AI version of someone else, a public figure or you in a public domain. So I think what's interesting is that the other day I realized there's going to be a generative AI version of ourselves already. Like I'm already immortal because in 50 years time, someone just pulls the podcast and the notes and just pulls it. The only decision is what level of resolution I granted, right? Either I put up, I don't know, like a creative comments and say, do not create a generative AI version of me, please. Or you can use public domain stuff of me, or you can hear a nice curated subset of materials I can use to populate your generative AI version of me. Or I'm going to give you everything, my Chrome history and everything and all I'm just trying to say is that there's enough digital footprint today to basically breadcrumb our way into immortal versions of ourselves.

So it's kind of a weird thing. Didn't realize that I'm now immortal. As long as I grant the rights of my digital footprint for folks to feed into AI. It's weird actually.


Shiyan Koh: (41:01)
It is super weird. You should do it. You should build a little bot. Cause you have like two years worth of podcasts and stuff. It'd be really curious to see what comes back out.


Jeremy Au: (41:11)
It was like notice like we're not Frankenstein. Like the story, where the person creates something.


Shiyan Koh: (41:17)
I think it'll be interesting. I mean, I don't have as much material as you do but it'll be a fun sort of weekend project to play with.


Jeremy Au: (41:23)
Well, only for the public stuff if you choose to grant it your WhatsApp and your emails generative AI.


Shiyan Koh: (41:30)
I think that's too much.


Jeremy Au: (41:32)
It reviews all your catchphrases, it reviews all your hopes and dreams.


Shiyan Koh: (41:36)
I don't know what's in your WhatsApp, Jeremy, but mine is generally coordination and logistical. My hopes and dreams are not communicated a lot in WhatsApp.


Jeremy Au: (41:44)
Yeah. I mean, but I think remember it is another thing which is like digital legacy, right? When you die your accounts and all that stuff goes to someone that you nominated in your will or as a custo-guardian, right? And there's a weird piece, I can't imagine the future is your guardian gets all your stuff and then the person's like, I'm just saying, okay, here's another startup idea, right? You're a generated AI. The Guardian get the deposit stuff and creates a generative AI of you.


Shiyan Koh: (42:10)
I think that's actually a great idea. I think that's a great, someone should do that.


Jeremy Au: (42:14)
Like a digital tombstone. But it's like a digital and oh, it's like a seance.


Shiyan Koh: (42:19)
Oh my God, yes. You're like talking to the dead, right? You're asking them questions. Oh my God, yes. A seance.


Jeremy Au: (42:29)
And then you do that at Hungry Ghost Festival. Now you have literal ancestral spirits. You can literally be from a hundred years ago. What do you think about me feeling lonely today? And then Shiyan is like, I believe in you, you can do it. Don't worry, dude.


Shiyan Koh: (42:53)
Don't worry. Go outside. Make some friends, talk to some people. That would be really interesting actually. That be a fun weekend project. We should build a bot. Yeah, I think that'll be very entertaining.


Jeremy Au: (43:07)
Well, on that note glad we got to discuss a lot of different things. Budget winners for Singapore, crowding out startups, generative AI barbell versus insights and secrets and judgment. And then we started brainstorming about how we can localize this southeast Asia agent and concierge, creative industry, digital tombstone, ghosts. So that was a fun chat.