“A lot of social science research reinforces the idea that changing your context helps you change your behavior. Imagine if there was a bot that makes everyone more active. It can egg you on and make the passive dynamic a little bit more active to remind you where your goals are. These are not technically complex and you could see them becoming ubiquitous. These little voices are trying to nudge you in a specific direction.” - Shiyan Koh
“Generative AI is really helpful because there's a trust component. If you have multiple accounts, you traditionally need an agent or representative to support you. There's also a privacy component. To some extent, you could trust an AI a little bit more because it's not talking to a human who could potentially talk about the topic over dinner, despite the NDA. That's a good idea especially because you can do everything and it can be like a real wealth advisor.” - Jeremy Au
“What’s interesting about the things people wish existed is that they weren't that crazy. There were a bunch of requests for things that are not that far off but the challenge is not the technical execution. We’ve seen various versions of these but they've never really taken off because of distribution or monetization challenges.” - Shiyan Koh
In this conversation between Jeremy Au and Shiyan Koh, they share their insights on entrepreneurship and forward-thinking perspectives on innovative ventures. The discussion highlights three pivotal areas:
1. Exploring Language Learning Models: Shiyan and Jeremy delve into the potential for developing language learning models (LLMs) for underrepresented languages such as Bahasa Indonesia, Vietnamese, and Tagalog, and the unique challenges and opportunities that come with it.
2. Community Unbundling: They discuss the concept of "unbundling the community," where niche communities can be broken off from large platforms like Facebook and Reddit, giving way to unique, focused platforms.
3. AI Applications in Wealth and Health: The conversation moves to how AI can revolutionize wealth management, functioning as a personal financial advisor, and its potential applications in health, functioning as a personal health assistant that keeps track of our health-related behaviors and nudges us in the right direction.
The discussion also covers other startup ideas they brainstorm. They touched on recent Thailand and Cambodia political news and responded to a listener's question on how to gain access as a diverse founder.
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 esevel.com and get a demo today. Use our referral code, "BRAVE" for three months, free. Terms and conditions apply.
Jeremy Au: (01:07)
Hey, Shiyan. Party time!
Shiyan Koh: (01:09)
Yeah, we're partying.
Jeremy Au: (01:12)
It's just like, it's just like, it's like
Shiyan Koh: (01:14)
No.
Jeremy Au: (01:14)
That's a random conversation. If you're watching on YouTube, she just pulled out a glass of wine and a can of La Croix. There go. That combo, I guess, is a sub form of hydration, right?
Shiyan Koh: (01:27)
Yeah, it's always important to stay hydrated and so I recommend if you're drinking wine, you should also be drinking water. In my case, sparkling water.
Jeremy Au: (01:37)
Sparkling water. Wow. There we go. So, well, you wanted to do a few shout outs? I think a few folks who said hi, they enjoyed the recent episodes. So, Shiyan, I think you want to say something to Lina.
Shiyan Koh: (01:47)
What's up, Lina? Lina from NonPublic. They're working on something cool over there and so wanted to give her a shout out for reaching out. It was funny because I pinged her and I'm like, we should catch up, and she was like, I feel like I just spoke to you recently. And then she realized, no, she had just been listening to the podcast.
Jeremy Au: (02:06)
I know. Okay, there we
Shiyan Koh: (02:07)
So we hadn't actually spoken. She had just heard our voices.
Jeremy Au: (02:11)
So this, it's like my wife receives a podcast episode from me at the end of the day. There we go.
Shiyan Koh: (02:18)
We don't need to talk to each other.
Jeremy Au: (02:19)
Marital communication complete, right?
Shiyan Koh: (02:21)
Here are my thoughts from the day. Please listen to this at 2x speed. So actually it's more efficient than actually talking to me.
Jeremy Au: (02:28)
Oh boy. I don't think it, it'll go down like the power bricks. There we go. David He at Gunderson, thank you for enjoying your run with us and hearing our family office episode. So, we are looking forward to your upcoming webinar on startup down rounds. I'm all ears. I think there'll be a really fascinating training that you're holding for the Singapore Venture Capital Association, S looking forward to that. I think we got one more shoutout from Sheryn Ascott who was asking questions about diverse founders. And I think, let me just read out exactly what she said. She said, "Hey, I loved the episode. I felt that there was a lot to learn about the reality of the business." And she felt like there was a part about blind spots that she really appreciated and yet she was also very thoughtful in asking questions about saying, Hey, I would love to hear more about how diverse founders can improve their chances of success in raising. She had said her friend had all the advantages and she found it so tough.
It's even harder when you don't have those advantages. So, yeah. Shiyan, I guess that's addressed to you. What's your quick response to that before we get into the rest of the episode.
Shiyan Koh: (03:31)
Yeah, I think to refer back to that episode, we talked about like. what is it like when the person you're pitching may not necessarily have an intuitive understanding of the problem that you're trying to solve and how that might lead it to be sort of an easier pass because they're kind of like, uh, I don't really know about this problem. It can't be that big of a problem. And I think this applies to all founders, not just diverse founders. And I think part of it is really trying to put your listener into the shoes of your target customer. Walk through that customer journey to give people that virtual sense of what is that pain? When do you feel it? What are your current not so great alternatives? Why is the thing that you are building so much better and so much more compelling to this person? And I think the example that we had given previously was someone who was doing diagnostics through the use of menstrual blood, and there was like both the ick factor and the like, Hey, this isn't something I'm like super familiar with.
But I think everyone has had an experience with, maybe not feeling super well and going to the doctor and maybe someone not being able to understand how to figure out what's wrong with you, right? So trying to find analogies to help people connect with the problem is important, just to make it simpler. It isn't just sort of an ick factor thing. I think it also happens in enterprise software where you might not understand the buyer. Why would a compliance person have this problem? Or why would an HR leader have this problem? So I think contextualizing the problem is a really important part of connecting with your audience and helping them see like, Hey, I am solving a big problem, and here's some information from customer development that I've done that helps give even more color around what that problem and the size of that problem is, or better yet, why my solution is so compelling.
Jeremy Au: (05:11)
Yeah, I think it's even more important in the sense that every founder is an outsider, fundamentally, because if you're a founder and you're trying to build something that's going to only happen in 10 years of hard work to become a true reality, the truth is you're always an outsider because everybody who's an insider doesn't want it to happen.
That's some degree, right? 'cause change is scary. The futures had to build. There's a bunch of embedded economic interests that are part of the current reality in terms of incumbency. And so, whenever you're trying to build that future, you're always going to be an outsider, and you're always going to be like that prophet or that Cassandra in the sense that nobody really wants to believe you. I mean, for my past two companies, when I started out, it was like I got rejected lots of times as well, but I also acknowledged that I worked very hard to create those bridges. In my first company, I didn't have some of these logos that I now have.
And so I would just be like, okay, these are the other ways I relate to you and these are the ways that I would just very explicitly spell it out for you. Whereas I think, I have to admit that a second time was a little bit easier because with those badges, I was able to open up the doors relatively easy. That's my personal experience, was like the door would open and then after that, you still face a pretty tough crowd, honestly, once you're in the conference room, but at least, the first door got opened, right? I think my reflection is that where it shows up a lot, I think if you are a diverse founder, is that it's really that first gate.
You even get that conversation going, right? And so I think my perspective is you have to leverage your network as much as possible. A warm introduction, because I think diversity is always like a certain plane, so the better is agenda or a single plane, but what are the other ways you connect? You connect based on school. Do you connect based on prior industry of expertise? Do you connect based on hobby? Do you connect based on a warm connection? So really, being very super specific to be like, okay, I am different from this person in one or two different planes, which is most people, right? I mean, we all differ. I differ from you Shiyan, in a sense that I'm male, you're female. You're cool. I am super cool, you know, so so forth.
Shiyan Koh: (06:56)
I've never been cool in my life.
Jeremy Au: (06:58)
Okay. All right. Right. We're both equally fine. This is similarity. We're both nerds who are trying to be cool. I'm just going to say we're both cool. Okay, great.
Shiyan Koh: (07:04)
I'm not trying to be cool, man.
Jeremy Au: (07:06)
Okay. You’re not trying to be cool.
Shiyan Koh: (07:07)
I’m embracing my life.
Jeremy Au: (07:08)
No, never cool. And I'm a nerd who's trying to be cool. There we go. But I think finding those planes is going to be those key moments. And I think that helps you get the door open. And I think the reality is that door, wherever you are, you could be a Southeast Asian founder trying to open up doors in the US. That's always hard. You're an American founder trying to build doors in Southeast Asia, that's also tough as well. Anyway, so that's my quick response to that listening question.
Shiyan Koh: (07:29)
This reminds me of trying to get through US immigration. Trying to keep the door closed.
Jeremy Au: (07:34)
Yeah.
Shiyan Koh: (07:34)
So whenever I land in the US, I try to wear something that has a local sports team logo on it. I'm trying to communicate to the border control official that I am just like him. I, too, am a Warriors fan. You know, I like basketball. I try to blend in. And so, they will quickly stamp my passport and waive me instead of putting me into secondary inspection and asking, making my life very difficult. So, I think this applies in a lot of different places, right? Which is like, what is the commonality you can connect with someone on and how do you signal to them that we're on the same team, man, we love Stephen Curry, who doesn't?
Jeremy Au: (08:10)
So there's a bunch of news that was happening in Southeast Asia that we felt had to be mentioned. Obviously, for those who are curious about Southeast Asia. I think the first piece of news was that Cambodia's Prime Minister Hun Sen is going to step down and is handing over power to his son Hun Manet.
So that's a big one that happened this past week, and the second one, of course, is that Thaksin, the Thai billionaire former leader is going to return in August after 15 years of exile. And so those are two very big pieces of news. I think this is pretty important because these are obviously important countries in Southeast Asia. And I think that I think folks are really thinking true about how these political transitions are really important because they are part of the political stability, right? And you have can only build an economy based on political stability and then you can only build technology and startups and venture capital that have a tenure horizon based on that overall economic stability.
So I think there's a little bit of a step ladder here that you got to have political stability at some level in order for the future to be built on those stacks. So yeah, those are my thoughts.
And now we go to our next stage. I think we want to talk a little bit about requests for startups. I think you and I had a bunch of ideas that we wanted to brainstorm and think of. So, I put together a little note list of ideas.
Shiyan Koh: (09:19)
Oh good. What do you got? I can't wait.
Jeremy Au: (09:21)
Well, you got you got some, right? I'm sure.
Shiyan Koh: (09:23)
I got a list, but you go first.
Jeremy Au: (09:24)
I go first. I'll go first. Okay. Okay. Idea one.
Shiyan Koh: (09:28)
And just to clear, are these ideas you feel have business merit or just things you wish existed because those are not necessarily the same thing?
Jeremy Au: (09:35)
I mean, if you wish something exists, it implies that a consumer like me wants it. And the question that you have to believe is, how much of a weirdo do you think I am? Is this a generalizable problem to a large enough segment that has some business merit? It may not be a VC-backable merit, but it's some kind of business merit. And unless you going to say like, this is a very idio-Socratic request that Jeremy has.
Shiyan Koh: (09:54)
Well, no, there could be other things. Even if it's something a consumer wants, it's like too expensive to acquire these consumers. There's just not enough margin to do the thing. You know what I mean? Like there's a lot of other reasons other than you're a weirdo, which we can debate separately.
Jeremy Au: (10:07)
Yeah. separately, extensively for the entire hour. There we go. Okay. So why don't we say, let's just throw them out. Nobody wants to hear about this theoretical framework about this. I'll try out one. So I think what's interesting is that we have lots of generative AI models that are all tackling different languages, right? And LLM models are not AI models, they're just probabilistic language models.
And so they're required to be trained based on language. And so the first wave, were all about English, right? So OpenAI. And obviously, that's gone viral, everybody's using it at least amongst my friends and colleagues. Then if you take a side step out, then you know, in China, which is I think part of that global decoupling, there are four different AI models that raise lots of money to basically build the Mandarin version because of two reasons.
One of course is that Mandarin is a totally separate language that requires very different set of training materials from English because a lot of their stuff, the Chinese internet is actually in order magnitude smaller than the whole english universe in terms of time, length, depth, as well as variety because of different countries and so, so forth. So I think they have to figure out different types of training as well and different kind of parameters and reinforcement training, but also because they're worried adding think rightfully so that they're not going to have access to the English model in a sense, because of that geopolitical tension.
Then, now actually we're starting to see the European. You know, we have two of them, right? We have one for the French, has a European focus obviously, but as a French team. And our one is a German team, and it's kind of building for the whole of Europe. So I think there's two of them. They both raise, you know, hundreds of millions of dollars as part of the seed rounds. And that kind of got me thinking a little bit, which is like, okay, what does it look like for Southeast Asia because we do have Bahasa, which is about 300 million folks. And then, obviously there's some commonalities with Malaysian Bahasa there. And then of course there's also Vietnamese, there's also Tagalog. So these are all Southeast Asian languages that frankly, nobody's going after yet, from a model basis. So I think there's an interesting dynamic where at least there's a metaphorical piece you can imagine that somebody could say like, Hey, the OpenAI for Southeast Asian languages like Indonesia, Vietnamese, Tagalog, which is I think, a bundle, some synergies based on the end users in terms of what they're going to be using for. So there's some customers' bundling that can happen together there.
Shiyan Koh: (12:14)
But is that just like a language translation middleware or like, if conceptually, if GPT is trained on everything that's on the English internet right now, is it that we think there are things that exist in other languages that are not in English internet? Or is it just that we think that there are certain idiosyncrasies about how those languages describe certain problems, that there needs to be some sort of like translation middleware, essentially to the GPT model.
Jeremy Au: (12:39)
Yeah, I think some approaches, I mean, people in China were complaining, that basically, they could tell that some of the models were, if they couldn't figure it out, they would basically translate the requests from Chinese and push it through the English models. And then after that, translate it back to Chinese and the results were a subpar because imagine it's just like, you're going through two versions of that filter as well. So I think people feel like the performance isn't there and of course, there's also an expense to it. And also you're not building the fundamental model reliability at a bottom of it. So that's an interesting dynamic that's there. I mean, you can so I think it's interesting, right? I mean, poetry would be a good example, right? There are different forms of poetry in different cultures. It's very idiosocratic. It's not going to be, you know, you translate a poem and all that stuff, it's just not going to work out. So I think that's one of the further age cases that's there. But also, I think in Bahasa for Indonesia, a lot of people are using short forms, so they're using brief abbreviations to represent the whole thing, especially in WhatsApp. And also, I think for a lot of emerging countries, a lot of that language is not on public internet because it's not on websites or MySpace or Reddit. A lot of it is actually messaging apps, which are all private communication. So again, it's even more invisible for a robot as scraping the web.
So I think there's an interesting dynamic. There could be a party round. Again, I think it would be something that is obviously in terms of race might be an order of magnitude smaller to some extent. Because again, you know, GDP per capita is different from the Europe and us, so the economic use case is going to be one order magnitude smaller. But you can imagine that population size does actually make it a compelling side to be like, Hey, what does that look like? So like the Grab or Gojek localization of OpenAI. So that's what I'm thinking. That's one idea I have. So yeah.
Shiyan Koh: (14:14)
So are you asking someone to build that or you think someone should?
Jeremy Au: (14:17)
Why not? Yeah, I've been asking everybody to build it. Because I think a lot of AI folks are basically building applications and you know, they are finding some success obviously via are English user, but you can't generalize down to SMEs for example, because they're going to be communicating consumers who are speaking Bahasa, for example, or Vietnamese, or Tagalog, right? And so that's something to be aware of. Okay, Shiyan, you have another idea? Let's kind go. Otherwise, it's going to be like one idea show, right? What's your idea?
Shiyan Koh: (14:46)
You are the idea guy, man. I'm just here to drink sparkling water and poke holes. Well, I mean, we put out, I pinged a couple sort of groups just to hear what people wish existed, and I guess what was kind of interesting was that they weren't that crazy. There were like a bunch of requests for things that probably are not that far off but the challenge is not actually, at least in my mind, the technical execution. The challenge is more like how do you make money and distribution? And that I've seen various versions of that, but it's very, they've never really taken off because of these distribution or monetization challenges.
So if you just think about what would make people's lives easier and better, there were a bunch of people who were like, I actually just wished I had a personal assistant that worked like a virtual personal assistant that worked. And then, I asked for examples of what those things might be. It's like, I want them to read my email and send responses to things that are like really common requests that I get.
Jeremy Au: (15:41)
Right.
Shiyan Koh: (15:41)
And you're like, that's actually not out of the realm of possibility. If you think about that, that shouldn't be actually that far away, but then you always run into this problem where it's just like, would you really trust your email program to write all of your email? How much control do you want? If it's a really standard thing, it's like, Hey, I want to submit a pitch deck, then yeah, you can have an automated response like, Hey, please submit our form or like, Hey, I'm busy right now, blah, blah, blah, blah, blah. Most things don't necessarily fall into those standard buckets, and so it's like how much judgment, how much leeway you think you can give.
It'd be kind of interesting, but you could see something. You could imagine something that trained on your responses, right? There was sort of like a generic model and then you plugged it in and then it looked at what you did over a period of time and it started trying to train on you and basically, initially, it would just queue things up. And you would approve them or amend them and it would learn again and you'd kind of keep going through it. But it seems like those things are like possible. But the problem with this is just initially it would be shitty so people wouldn't pay for it. You need it to become better, but this is like the whole sort of distribution monetization problem.
And so, it seems like one of those things where like an existing email provider will do it because they already have distribution, so it's like not a great startup kind of thing, right? But yeah a lot of these ideas were things where you're just like, that's actually not that far off. It probably will exist. I'm not sure a startup is going to do it. It actually seems like an incumbent is better positioned to do some of these things.
Jeremy Au: (16:57)
Riffing off that, I think there's a version of that that has a lot of economic value and that you just reminded me of is basically like an AI version of Hubspot or Pipedrive or some kind of sales dynamic, which is I think the same level, but I think it's much more tinly, which is at the end of the funnel, a sales meet in terms of a contract is signed and then, of course, customer service takes on from there. But you know, all that work is such like all bashed up, right? I mean, you have lead generation then there's lead qualification. Then you got to do the onboarding call, and then you got to do a bunch of emails to schedule stuff, and then you got to go and do the call, and then you got to close and you go to the contract.
And I'm like, of course HubSpot is like looking from a marketing and all in one perspective, but I think this is HubSpot's design philosophy, rightly so, is called there's a human driving this entire process because guess what? We already generated a lot of these communication tools and so I'm kind of waiting for HubSpot to maybe either implement it, that's one, but it could just be someone else, right? We're just basically saying like, look, you don't need to buy HubSpot or PipeDrive or whatever it is, or Streak CRM, you know, look, you just want close sales. We have this digital agent that basically is your secretary in terms of user shared or personal assistant, but it was going to help you drive sales, lead generation, typing your parameters with scrape LinkedIn for you do everything, just make is KISS, right? You know what I mean? And I think so many folks that are not company level, you're like a one man, five men, 10 men sales team. Just be like, okay, this is a great way for us to basically double, triple, quadruple productivity rather than more CRMs are about trying to avoid losing productivity because you're busy filling out notes on everything. So you could even have like meeting bots, every salesperson hates writing sales notes. Good. I understand cause you could be selling instead or you could be doing something else in your life.
Shiyan Koh: (18:30)
But that should already exist, right? You already have the like news, the recorder, it already summarizes and you can just insert it straight into the CRM.
Jeremy Au: (18:38)
Yeah, but then you had to go and if you put it in and after you had to go there, Ctrl + C, Ctrl + V, and then you got to edit a little bit and you go to press the submit button because even the transcription stuff, they don't summarize it so they just transcribe it. So what I'm trying to say is I think you just, everyone's just speaking on these like foundational things.
Shiyan Koh: (18:53)
They're doing it
Jeremy Au: (18:54)
They's starting.
Shiyan Koh: (18:56)
Have you seen Rewind?
Jeremy Au: (18:56)
Yes. I have seen Rewind. I thought it was pretty cool.
Shiyan Koh: (18:58)
It summarizes meetings and it's pretty good.
Jeremy Au: (19:01)
I mean, I wish I should've told somebody , almost like I want to say seven years ago, I said, I literally told, someone's asking me about something like transcription, et cetera. I said, can you just record meeting minutes and just tell me what the to-do is? Because we are always assigned this to either the most junior person who normally does a terrible job at it, or someone like who's kind of like middle to senior who's good enough to do it, but we need the engagement and the conversation and not taking meeting minutes. So I was, literally told him, I said, I'll literally pay you like $50 an hour meeting just to get this done because we're having important meetings. Anyway, you have to take my word for it, but I literally told the person that I should have recorded it in a request for status episode. Now it does. It's amazing. It's amazing.
Shiyan Koh: (19:39)
I actually had the first person who actually told me to turn off the recorder, the first person in three years, told me to turn off the recorder today.
Jeremy Au: (19:46)
Well.
Shiyan Koh: (19:48)
Citibank does not allow their employees to be on any recorded calls and also does not allow them to use ChatGPT. This is what I learned,
Jeremy Au: (19:56)
Interesting. I mean, I generally say no to being recorded, actually. I mean, you know.
Shiyan Koh: (20:01)
We're being recorded right now.
Jeremy Au: (20:03)
Cause I know we're going to be recorded, but then they're asking me stuff like, okay, we'd like to do a conversation about your financials. Are you okay being recorded? I'm like, No. Yeah. I'm just saying if you can't say no to, why would you say that?
Shiyan Koh: (20:15)
Yeah. Totally. But like a standard business meeting.
Jeremy Au: (20:17)
Yeah, like an internal business meeting, we're talking about stuff. But then of course, you know what, if you're like business humor. You're joking about A and B and C, I don't know.
Shiyan Koh: (20:25)
I don't know what kind of jokes you're making, man. Better watch out.
Jeremy Au: (20:27)
Okay. I mean, I love comedy, right? I think the big thing I realized about comedy is that it's always context and audience specific, right? So everything's funny within that time for the audience. But comedy kind of breaks down the moment you get outside that audience, or, yeah, outside the context. The problem of recordings is that it's not that, it's not that you don't have a good conversation with like your three people and you're joking about something. Let's say I'm in the army, right? And then I'm like, yeah I used to have lots of army comedy, right? But I wouldn't really want my bosses, I'm just saying, cause like they say like, oh, this is what kind of dark humor is this about life and you know, blah, blah, blah, wanting get out and all this other stuff. I'm just an example, right? That wouldn't play out well. And you zoom that out, outside Singapore, then it is a different form of joke.
So I think that's why everything is getting like canceled, right? It's because comedy is about some level of understanding, shared understanding of the audience and context, but also some noticing or some absurdity of the outside environment and that ends outside environment. Yeah. It's the audience changes, right? And so I think from my perspective, that it's not about, I don't have the issue being recorded if, as long as that recording stayed for the three of us, for example. But once it scaled out to like 1 billion people on TikTok, nothing you say can really pan out.
Shiyan Koh: (21:33)
Aren't we on TikTok? Jeremy,
Jeremy Au: (21:35)
We are on tikTok.
Shiyan Koh: (21:36)
This podcast?
Jeremy Au: (21:37)
No, we're not going to get a cancel on TikTok because we know what we're talking about and people can opt into a conversation or people can opt out of that conversation. They can choose to listen to this on, swipe away from us, right? So this is a different context, but I think comedy is always context dependent, right? Yeah, there's a circle of trust. So yeah. So, I think there's lots of things that as all these meetings become transparent, and then of course, maybe you're discussing, I'm giving an example, performance reviews, I'm using an example, right? So yeah, you can imagine lawsuits happening for that.
So what I mean by that is that protected categories, so there's obviously in the US there's age, gender so, so forth, all these other things, but one of the big issues that we saw was that the reason why nobody gives feedback after interview rejection is that, people get sued, because they say, Hey, you said that you're looking for a high energy person based on their feedback. And I am older and so therefore, you're saying that you're discriminating on me based on my energy level. It's code for, and so so forth and I think as a result, I think we basically saw the collapse of post interview feedback effectively, because, people just say like, Hey, we have lawsuits. So you imagine like meeting meetings about performance reviews, those would never be recorded because how do you have a frank conversation where you know that's, I don't know. I think there's that dynamic of transparency, which is a perfect virtue, but then also there's a trust circle, and that's at some level not the same thing.
Shiyan Koh: (22:51)
Yeah. Yeah. I don't know. People say crazy stuff. I had a friend get feedback was like, you seem great, but your lifestyle requirements are not, are too much for us. And like the lifestyle requirement was basically like I have to pick up my kids from school twice a week.
Jeremy Au: (23:06)
Yeah. I mean,
Shiyan Koh: (23:07)
You know, it's not like, not like she didn't want to do work. It's just like, Hey, like I have to do this thing and I'm happy to log back in after.
Jeremy Au: (23:14)
Yeah.
Shiyan Koh: (23:14)
Thought we were kind of like past that stage where it's basically you can't have a working parent. I mean, that seems like an insane selection criteria.
Jeremy Au: (23:21)
I think it's true. And I think that's the thing, right? It's just like, there's fair feedback and there's unfair feedback and I think transparency makes it super clear, but it's not in everybody's interest to make it super clear or it could even be misunderstood, right? So I think there's a crux of it. Okay. I think another tool I have would be figuring out if something is human or bot-made. I think I'm starting to see all these images, these texts, and I know it's so hard, but.
Shiyan Koh: (23:46)
You want know if it was artisanal.
Jeremy Au: (23:48)
Was it organically grown by a human? Did somebody sweat and then told to himself like, I handcrafted this horrible dad joke pun, and I laughed at myself. Hee Ha a at my laptop before hitting my submit button, instead of like, someone who just automatically generated, it was just a total bot. No one even, nobody even, you know, nobody even auto automatically generated. Is this a bot farming commcomma right?
Shiyan Koh: (24:09)
But why does that matter? Why does it matter if it's funny? Who? Why do you care who made it?
Jeremy Au: (24:13)
I'm an old person. I'm a geriatric millennial who likes my stuff human generated. I don't know. I mean, if you look at your social media feeds as a way just to be entertained, and it doesn't matter whether it's bot generated human generated, but if you're looking at it as a way to be part of the community, you wouldn't want this to be ghostwritten by somebody else. So, you, you know, if your secretary wrote that email, saying, oh, I really appreciated your new line of clothing, and it was really amazing. It'll be come across as so much better if it was written by you rather than written by a n AI personal assistant.
Shiyan Koh: (24:45)
Yeah, but you as the recipient, like you got already the good vibes from the message, right?
Jeremy Au: (24:50)
But that's why people want the signature, right? Somebody wrote handwritten cards. I mean, handwritten cards are luxury these days, otherwise it's just emails and texts. Happy Birthday is very different on WhatsApp than you wrote it down. You wrote whole birthday card.
(25:01) Shiyan Koh:
Jeremy Au: Yeah, level of effort.
(25:03)
Yeah, and it means more like the more busy you are. If you're the president and you wrote a written card, it means a lot more than, for example, Jeremy at the age of 12. Well, actually Jeremy, at the age of 12 was pretty cute, so I think it was worth a lot as well. But I'm just saying like, to some level, you know that when someone's really busy and it took the time to do that extra human touch, it's meaningful, right?
Shiyan Koh: (25:23)
Yeah. How much would you pay for that service though?
Jeremy Au: (25:25)
To find out if stuff is real or not. I think. Okay. To me, it would be a Chrome extension. Would just be a Chrome extension. I'll pay like 10 bucks. Oh yeah. I'll pay 10 bucks a month. Why not?
Shiyan Koh: (25:33)
All right. You heard it here, guys, $120.
Jeremy Au: (25:37)
Right? Okay. There's a great Instagram account of follow. It's like real, or not some equivalent of it. But basically, It shows what celebrities post on the Instagram of themselves, et cetera. And then it shows the getty images. But you know, like public photographers, public domain photographs of just them at the same event. And then you can just see the difference. And I just use that to remind myself, there's really so much like photoshopping, removing of blemishes, distortions, and all these filters to me. And you know, it, it, you know, like there's a lot really attractive people, both male and female on the internet all the time.
And that Instagram kind of like always fact-checks me a little bit, which is like, yeah, guess what? You know, like, I don't need to look like that because it's totally impossible to look like that. But I mean, the truth is, if you are on the internet all a time, it's like, you know, your standards get whopped, right? Like 80, 90% of the images are like that, you know, then you're just like, oh, okay. It's totally normal to have like perfect skid, you know? And then you're like, oh. You know, people have pimples, people have whatever, and he is just like, you know, I mean, I think that's fine. I mean, there's like mouth blemishes and mouth freckles.
Like I can imagine that, like, I can't believe I'm saying this, but it's like, okay, that's not even the worst thing I've seen, right? I think the worst thing I've seen is you know, those waists are like super skinny, I don't know.
Shiyan Koh: (26:44)
That's why I like Zoom. It makes me look better than Google Meet.
Jeremy Au: (26:47)
How does Google Meet do anything? What are you talking about?
Shiyan Koh: (26:50)
No, Zoom. Zoom has a correction feature. It makes you look better.
Jeremy Au: (26:53)
What?
Shiyan Koh: (26:54)
Yeah, you didn't turn it on?
Jeremy Au: (26:56)
This is a correction filter. Wait, what? Okay, wait, I need to check this out. Is it like a good correction filter?
Shiyan Koh: (27:02)
Yeah. It's good. This is like, so you know, you're on Zoom all day long and you like look at yourself and then you get up and go to the bathroom and you're like, wait, I look terrible. I don't understand. And you're like, oh, it's the Zoom correction filter. It makes me look better when I'm in meetings.
Jeremy Au: (27:17)
What? Okay. Tomorrow, I will try it.
Shiyan Koh: (27:20)
But you know what? Right now here, whoever's watching this, there's no correction filter. This is how we look.
Jeremy Au: (27:26)
I know we need to complain to the Riverside platform. It's like we need correction filters all the time. We need it.
Shiyan Koh: (27:30)
Yeah. We need to be approved. The wrinkles, you know, everything, smooth out our skin.
Jeremy Au: (27:34)
Oh boy. Yeah.
Shiyan Koh: (27:36)
Yeah. So another thing that people sort of asked for, which was just like they wanted a comprehensive view of their finances and they wanted to be able to like.
Jeremy Au: (27:43)
Yeah. What happened to them?
Shiyan Koh: (27:44)
They wanted me to interrogate it, like ask questions of your data.
Jeremy Au: (27:51)
Yeah.
Shiyan Koh: (27:51)
Without having to do the analysis yourself or having to do a ton of manipulation of the data. So like mint.com was just like, Hey, I'm going to log in and put everything in the same place, which was, you know, step one, but, first of all, it's hard to get aggregation across geographies. So let's say you had accounts in different places, that's pretty hard. It's easier to do in a single geography, but then, the data itself is not particularly insightful, right?
It doesn't really tell you anything. So if you had to take it to the next level, you'd want to be able to talk to it the way you talk to a financial advisor.
Jeremy Au: (28:22)
Yeah. Exactly.
Shiyan Koh: (28:23)
You'd want to be like, Hey, given my current spending patterns, how does this set me up for retirement?
Jeremy Au: (28:28)
Right.
Shiyan Koh: (28:28)
Or like, it seems like we've been spending more on X, Y, Z thing, like what's going on there? And it would just tell you, right? It would basically analyze what was going on, like give you answers to it like a personal financial advisor you could interact with. But basically, it's just a ChatGPT’s bot sitting on top of your own personal data that was aggregated across like a bunch of different institutions and geographies and like, the difficulty is also being able to track the cache between everything.
Jeremy Au: (28:55)
Right. Yeah, that's super fair. And I think that's where some of this generative AI is really helpful as well, because there's a trust component if you have multiple accounts, et cetera, you traditionally probably need an agent to support you, right? So some sort of like representative to walk you through so, so forth. But there's a privacy component. I can tell them everything. These people tend to be, I wouldn't say fresh graduates, but they're relatively young. These are wealth advisors, and so I think to some extent you could trust an AI a little bit more to some extent, because it's not least talking to a human who could potentially talk about it over dinner no matter their NDA, et cetera. So yeah, that's a really interesting idea. I think that's a good idea especially because you can do all the things right? You can look at your property, you can look at your tax rates and how to apply for tax credit, et cetera. So this, that whole encompassing dynamic, right? Yeah, that's a great idea. Like a real wealth advisor, right?
Shiyan Koh: (29:42)
Yeah, but on your own data, which is basically saying like, Hey, I'm going to give you all my data and I just want to be able to ask you questions like, Hey, am I overspending my budget? If you set your own budget, you say, Hey, you know, I only want to spend $500 a month eating out. Could like halfway through the month be like, Hey, how am I doing right for the rest of the month? How am I doing against my budget? I think that would be like a sort of on-demand kind of useful intelligence.
Jeremy Au: (30:06)
Yeah, like basically giving you, he kind of scrapes your credit card statements as well, so you connect it in automatically, your bank balance, et cetera. So he just, he knows your, oh man, this is this. I've signed up for so many of these personal finance apps in the US and they've Yeah.
Shiyan Koh: (30:20)
Shitty.
Jeremy Au: (30:21)
Why? Yeah.
Shiyan Koh: (30:22)
Well, it's a variant. It's a variant of a bunch of startups that I've seen, which is like trying to replace the SQL query, right? So it's like the constant challenge of like business analytics teams, right? Is like they're getting questions from all over the business that are like, how's this cohort performing? What was the process marketing campaign like? Did it actually lead to an increase in conversion, blah, blah, blah. And then, some poor analyst is sitting there like querying the database and it's like, well, do you need that? Is there a way people use natural language to like query the data, set themselves, and self-serve, rather than having to rely on your, on a giant data analytics team to answer some of those questions? And I think that's like the sort of personal finance analog, right? Can I query my own data?
Jeremy Au: (31:01)
I think you also do like an idea that came out to me a long time ago was like a self-tribal reinforcement, brainwashing, I don't know what you call it, goal. Okay. Sounds scary. I don't know what rephrase is. It's okay. So for me, when I was like losing weight, I think a big part I had to do was I had change my tribe, right? And what I mean by that is I had to like, it was a pain in the ass, but basically I subscribed to all these health newsletters, health podcasters, and YouTubers. I'm just saying, right? I was like change my algorithm.
Shiyan Koh: (31:27)
Listening? You're listening to Huberman?
Jeremy Au: (31:28)
A little bit, but only when he's a guest. I don't listen to his, like regular stuff. Yeah. So I mean, I think these are all content, so I'm training YouTube, it's like, Hey, I'm trying to get healthy, right? I'm I don't want to click on those things, but I click on them anyway, just so that, I hope that the signal kind of like I'm putting like this bad signal on the sky saying like, help, I need Batman to come.
And I mean, Instagram is why I follow these things. I also had unsubscribe from other things, right? And so I think there's a bit of a cultural, tribal dynamic, identity dynamic, and I think to some extent, some diet nutrition apps do that. They used to do that. I used to sign up for these, these health apps with these nutritionists humans who would, you would talk to them about, take photos of your food and then they'll positively reinforce you and say like, great, Jeremy, you did a great choice about that food. I'm just an example, right? But obviously, the problem is that, it falls apart the moment you get a big, giant piece of pizza and you don't take a photo of it because if you're guilty about telling someone about it and then they don't respond to it because they don't know about it and blah, blah, blah, all that stuff, right?
Shiyan Koh: (32:20)
So how about you just wear a camera and it like constantly tracks everything you're doing, so you can't even cheat yourself.
Jeremy Au: (32:25)
Yeah, but it's kind of like similar to personal finance assistant, but personal health assistant, right? It's like looking at your spend and all these other data sources that would never have been there. And there's the AI that basically just says like, Hey, we just noticed that you ate a lot this week at these restaurants. You, you ordered a giant Hawaii pizza. An example, right? I mean, you know, it is my kids, I swear it was like 2:00 AM my kids were asleep. Swear to God, you know? So, I'm just saying like, you know, there's some sort of the AI that could send me on WhatsApp to be like, you don't need to log into an app or over WhatsApp is if basically sends you a message that, Hey, we just know this, blah, blah, blah. Here's some nice positive message for yourself the day.
Shiyan Koh: (32:58)
I take photos of my food and I send them to my nutritionist and my nutritionist's constant comment is fish balls are not real protein. Please eat real protein.
Jeremy Au: (33:08)
Okay, so when I did those health apps in the US, it always broke down because they couldn't handle Asian food because you're like, you're taking a plate of taco. It looks the same as healthy noodles, honestly. But there's a big honking difference. And so from the following perspective, it's just totally different. They couldn't handle Asian foods, right? So it's just like, you need to go carb-less. And I'm like, you need to stop eating potatoes. And I'm like, I don't eat potatoes. I want rice, I want noodles. And they're like, eh, sorry. You know, our training SOP for our army of human therapists in terms of nutritionists. Yeah.
Shiyan Koh: (33:37)
A shoutout though, Novi Health which is a Singaporean startup, their database has Asian food.
Jeremy Au: (33:43)
Yeah.
Shiyan Koh: (33:44)
So you can put chocolate there., You can put Ramen. And it also has Western food. You can also put like a tuna sandwich or whatever, a bagel. I could do it, but it has a Asian food. So shoutout to Sue Anne and Novi Health.
Jeremy Au: (33:55)
Awesome. Well, that's a plus one to our Monk's Hill Ventures part for the company, so thank you for the very kind words. Yay. Check out Novi Health. So yeah, it takes the tricky part, right? But it could be other things like, I'm trying to, I want to really get more into I think the Google and entire internet algorithms are based on who we were, but I think, who do we want to be?
Shiyan Koh: (34:14)
Oh, I think, if you could put a bot into the group chat.
Jeremy Au: (34:18)
Ooh, like on WhatsApp.
Shiyan Koh: (34:19)
So I mean, I actually like this and there's a lot of social science research that reinforces this, changing your context helps you change your behavior. So like my college friends, we all have this WhatsApp channel called fit workout or whatever, and people post their workouts. And so, we all live in different geographies. Basically, you wake up and you're like, oh, I don't want it. But then you can see all your friends already worked out that day and you're like, oh fine, I'm going to do something. But imagine if there was a bot in that chat. So it was like even more active.
It was like, Shiyan, don't be lazy. You know, Sarah already did this. And it can even go back and be and it could be like, why don't you try to beat Sarah's distance or time?
Jeremy Au: (34:56)
Yeah.
Shiyan Koh: (34:57)
And it can egg you on, but it just makes like the, it makes the passive dynamic a little bit more active to kind of say hey, and remind you where your goals are and things like that. And I think these are very, these are not technically complex and you could sort of see them becoming ubiquitous, right? Like these little sort of voices that are just like trying to nudge you in a specific direction.
Jeremy Au: (35:17)
I mean, that's what FitBit was trying to do, right? They're trying to create these communities, right? So that you could, cause at the end of the day, if you're a user of FitBit, you either could and it circles back to your point earlier, right? Is the newsfeed, is that feed of images of people being healthy? For example, is it something to be consumed? Is this a utilitarian perspective or is it for a sense of belonging? And the truth is that if you're just looking for that information, both. I can imagine that, I mean, I think that's the issue, right? It's like historically, every community allows you to consume in that sense, but also to allow you to belong. And so Reddit, for example, when you have bots, is that, they're breaking the social contract in the sense that something that you consume was produced by somebody else who also belongs, right? In this case, now the production is done by bot and the, belonging is still another looker or whatever. So I think there's that breaking of the social compact, but I think that's for the social clubs. But obviously, if you're talking about trying to get fit and you can't find a community right. Then this is a great, I don't know, intermediate way to get there or accelerate.
Shiyan Koh: (36:11)
Yeah.
Jeremy Au: (36:11)
On that note, any parting ideas, any you want to shoot out there before we wrap things up?
Shiyan Koh: (36:16)
No, I think these are just always fun exercises to kind of get people's juices flowing on. Like when you look around the world, like what do you feel is like pretty broken and you know, what would it be interesting to go solve. I'm actually curious to hear from the listeners. I would invite them to submit like requests for startups and we can do an episode with listener ideas and just bat them around a little bit and talk about what would be hard or what would be easy or how would we test it. I think that could just be like a really fun way to exercise our sort of investor brains.
But yeah, I think one thing is wanted to kind of share an event that's happening this week. It's Founders' Friends.. It's a series that we've run in the US in San Francisco and New York, and always had pretty good reception on it. It's the first time we're doing it in Asia. It is Wednesday, August 2nd, and we'll be interviewing a founder, Max from Pantas, which is a Malaysian carbon accounting software. He started out as an e-invoicing software and then pivoted to this and has seen really interesting traction. And so would love to invite folks to join us to hear a little bit about Max's story and then meet other founders. And I guess Jeremy, we can include the Luma link in the show notes for folks.
And if it's successful then, we would love to be able to do more and I think the goal of this is to have really authentic conversations and not just talk about the raw stuff, but all this other really hard stuff.
Jeremy Au: (37:35)
Yeah.
Shiyan Koh: (37:36)
I think that's kind of a conversation. Would invite you guys to come join.
Jeremy Au: (37:40)
Awesome. On that note, see you next week.
Shiyan Koh: (37:43)
See you next week.
Please register for Founder Friends in Singapore here.