As the company got bigger, the company's vision and the CEO's vision didn't matter as much as making sure the employees are happy where they are, and what the company's doing must also be best for the careers of everyone working for the company - Bhargav Sosale
Bhargav Sosale is the cofounder of Medios Technologies, a startup building AI tools for screening retinal conditions. It deploys its algorithms on portable retinal cameras that every general practitioner can afford. Using Artificial Intelligence to analyse images of the human eye, Medios is able to safely detect Diabetic Retinopathy (DR) in its earliest stages to prevent vision loss in patients, and generate instant results for a fundus photo examination. Founded in 2017 at Entrepreneur First and backed by SG Innovate, Medios has begun working with leading hospitals to test their software. Today, Medios is being used by doctors and community health programs across India, and has screened over 30k+ patients. In 2019, Medios was acquired by Remidio Innovative Solutions, a leading ophthalmic device company based out of India.
Previously, Bhargav co-founded Lisn, a music streaming app that provides users with a collaborative music streaming and listening experience in real-time. At its peak, Lisn trended on Product Hunt, was getting over a 1000 downloads a week, had over 120k songs streamed, and the app was used by over 10,000 users.
Jeremy Au: [00:02:03] Good to have you, Bhargav.
Bhargav Sosale: [00:02:05] Hey, thanks a lot for having me today.
Jeremy Au: [00:02:07] Yeah, it's awesome to see you after the acquisition and see you on the other side of things.
Bhargav Sosale: [00:02:15] Yeah. It's actually been quite surreal, to be honest. One half of me, I think, feels like nothing's really changed post-selling the company, and then the other half was like, oh wow, everything is different now. You went through it, too, probably. You're able to get what I'm talking about.
Jeremy Au: [00:02:29] I think there's a part of you as a founder and operator where you're still running the business, so that has never changed. The other part is that you're no longer the owner of it. So as a result, you're a non-executive, and so that's an interesting position that you just made me reflect on in the last 10 seconds.
Bhargav Sosale: [00:02:46] What's super interesting is, this is, I think, my second real job. Since I've been starting a company, I've always been the founder and the one behind it, and except for a small stint right after I graduated college, I think this is the first time I've actually been employed somewhere else after the acquisition. That's also been something that I've been getting used to.
Jeremy Au: [00:03:05] Well, we're definitely going to dive into that, and we're just going to go and have the opportunity for you to share your journey at a high level for those who haven't had the opportunity to meet you.
Bhargav Sosale: [00:03:17] Since I was a kid, I've always wanted to make people happy and change their lives in some way. So when I was 14, I started out actually as a musician. I joined a band in high school and we actually grew fairly big in India. We were touring around the country. We were playing sold out gigs in front of thousands of people.
I hadn't actually even considered tech or startups. I didn't know anything about it until the movie, The Social Network, came out. I saw that and I was like, wait, so you've got this one guy sitting in a dorm room in Harvard, changing the way the entire world works and he's creating like such huge amounts of impact. Here as a musician, I'm still able to do that at a small scale, but with tech, I can scale it up to even a billion or 7 billion people, everyone in the world.
That's what kind of piqued my interest. So I decided to study computer science. Now the Singapore government was gracious enough, they gave me a scholarship and asked me to come study at NTU. I came to Singapore and that was the first time I learned how to code. Because I knew I wanted to get into creating impact through tech, and startups were actually one of the best means to that end.
So I spent about a year after graduating, where I also did some machine learning research with my professor back in NTU. He wanted me to do a PhD, and that's when I kind of turned down that option. I joined a company called Feature, which was like a Foursquare for Singapore. I joined there because I thought the team was tremendous. Actually, the CEO of that company is now the president of Sale Stock in Indonesia, which is a unicorn.
I learned so much from that team about how to actually build products from just an idea, all the way to launch and everything around it. How do you acquire your first set of users? How do you achieve product market fit? I learned so much from that.
I think after that was when I decided to take a dive and build my own company, right? It kind of worked out as an opportune moment because I had a side project that was an app that lets you listen to music with your friends. I built that because me and my friends wanted to use it, and once we put that on and we started using it, we loved it. We were like, hey, let's do this full time. We put it out on the app store and I would say it kind of blew up in the early days.
We were getting actually about 1000 to 2,200 downloads every week. We actually trended on Product Hunt. We had a bunch of users in the Valley and Mexico and Spain. I think that was my first taste of, okay, look, I'm actually creating the impact at the kind of scale I wanted to do it. While that company didn't work out for many reasons, but one of them being, we were actually very inexperienced founders. We didn't know how to build a business, we just knew how to build products. We kind of tuned that down, but that's when I knew, okay, building a company was for me.
The second time around, I was recommended to EF by a friend who was in the SG2 cohort. For those of you who don't know, EF is an accelerator/investor that invests in people, not companies. So I was recommended to the SG2 cohort, I took the interview process and I got in. That's when I built my second company, which was a health care company that looks at photographs of the eye and diagnoses eye diseases.
I built that with my co-founder, Florian, who I met at EF. Well, long story short, it's been three years since we built the company. We've actually sold that company. We've had some war stories where we went head on against Google, competing directly against them, but we're happy with the way things turned out. We live in a hundred different clinics in India, different state Akella, state government screening program in a state in India, and we're also being used by a very large NGO called The Vision Foundation of India.
So we've come quite a way since then. I think our original goal starting this company was to prevent blindness, and I think we're on the way to making it happen. Now after selling, we're doing that with our parent company.
Jeremy Au: [00:06:52] What an amazing journey. How did you first get started on in the startup journey? How did you catch the bug?
Bhargav Sosale: [00:07:00] Well, there are two aspects in general with startups. I think it goes back to what I previously said, I've always wanted to create impact and make people's lives better. You can do that in many ways. I was doing that as a musician at first, but along the way I realized that startups just had so much more leverage. When I saw that two people with computers in their dorms could create this kind of impact and startups were the way of doing that, I started naturally leaning towards startups as well. I think that's the point where I caught the bug, because it's very addictive.
I wish I could share my screen. I don't know if I can. But I think when we built the music app, Lisn, and we put that out there for people to see and use, we started getting a bunch of fan mail. It was crazy. I think it was an app that lets you listen to music with your friends, that was the first app we built, and people started sending us emails saying, "You know what? You've helped save my relationship with my long distance boyfriend." And we were getting dozens of these emails on a very regular basis. I was like, wow.
Because I feel like the startup journey reinforces itself. You catch the bug because you have an idea of what you could do through startups, but then as you do it, you start getting these pings of reinforcement, whether it's a really happy customer or someone telling you that your product has changed their life, right? Those pings of reinforcement gave me dopamine hits, like it just made me want to keep doing it more.
Jeremy Au: [00:08:19] Who was influential for you in starting the startup journey?
Bhargav Sosale: [00:08:23] That's a very, very good question, actually. So I can think of a few different people, but one person I can say had a profound impact is a college mate of mine. His name is Pulkit Jaiswal , right? We're actually not in touch now. I want to get back in touch with him sometime.
When we were in college, he was, I would say, the outlier out of every student. You know, in Singapore, college is mostly about, oh, we've got to get the best grades. We have to do all the ECAs to check a few boxes so we'd get a job at a bank, and I was down that path initially, as well. Pulkit was a complete outlier, right? He was the kind that he eventually dropped out of college because one day he said, "Hey, you know what? I want to move to Silicon Valley and start a company."
For someone who was, I think at that point, 18 or 19 years old, to be able to say and do that, I found absolutely ridiculous. I was like, wait, how are you going to do that? Are your parents going to be supportive of it? What about money? What about the rest of your life?
He was very nonchalant. He's like, "You know what? I'm going to do this." So he went to the Valley and he actually started a company over there. He became a Thiel 20 out of 20 finalist. He came back to Singapore after that, I think, once his visa ran out, because he was on tourist visa there. So he came back, and I think I had a catch-up and he told me about his journey.
Even then, I was still very, very skeptical. I was like, are you bullshitting me? Like you being part of the Thiel 20 out of 20, just quitting your usual life, quitting, dropping out of college, doing this, right? I was very surprised. He again said, "No, this is the best thing for me, and I'm creating the kind of impact I've always wanted to create." That's when I was like, okay, there is something more to this.
Again, it was the pings of reinforcement. He went back to the U.S. and then he came back again sometime later and I had a catch up with him. This time he was getting ready to start another company, which was called Garuda Robotics. It's actually a pretty big company in Singapore. He said, "Hey, I'm starting this second company with drone tech, and I'm doing this with Professor Mark Yong." I was like, wait, you actually convinced a professor of our school to drop out and start a company with you? Why would he do that? He was like, "No, he's excited about the vision and he wants to do this with me." I spoke to Mark, and Mark said that he was not dropping out, he was resigning from NTU as a professor to go ahead and build this company.
The third time that happened was also, he built the company for a while, we met again a few months later, and he'd just closed a round of funding. I can't remember the exact amount of funding, but that was when I realized, okay, look, I have to get out there, to get off my ass and what society pushes on you as a very traditional way of living life. The traditional societal checkboxes are probably not the best way to do things when I've actually seen a lot more success with people like Pulkit who have been very outlier-like in their thinking.
So I would give him a lot of that credit in terms of just helping me shape my mindset on basically saying that I shouldn't follow what society tells me to do. Of course, other things reinforced that. A lot of books I read reinforced that. But I would say that was probably the earliest time.
Jeremy Au: [00:11:22] What have you learned about leadership along the way?
Bhargav Sosale: [00:11:25] That's a very good question. I think what I've learned about leadership is that there is no singular definition of leadership. What is leadership to one person may not be leadership to someone else. It changes from person to person, from context to context, depending on the kind of company, the stage of the company you're at.
One of the things I always believed when I was a kid was, hey, a leader is someone who's like the boss. He says something, everyone's like, wow, if this guy said it, we must do it. Now that I've gotten a lot older and hopefully wiser, I feel like leadership is kind of the opposite, where I see that leaders are those who serve their people and act in their people's best interests and not themselves.
At an early stage startup, I've seen that leadership, at least I believe right now that leadership is primarily about having a vision so compelling and exciting and being super passionate for why you're doing what you're doing that that is infectious towards everyone around you who get behind you on that vision. I think that was what I believed leadership to be in an early stage company.
Now, as the company got bigger and you started having 10 employees, 20, 50, 100 employees, things started changing there where you now suddenly have to start caring about your employees and their wellbeing. Make sure they're happy. Make sure they're, I would say, content with their lives. We have to make sure that the company's vision and the CEO's vision doesn't matter as much anymore as it's about making sure the employees are happy where they are and what the company's doing must also, in a way, be best for the careers of everyone working for the company, because now those are questions that are coming up.
I'm in the process of really figuring all of this out, to be honest, because for me, leadership has always been accidental. I've always been the kind of guy that says, hey, I really want to do this. I'm going to do everything. I'm super passionate about it and I'll go behind it. I think until now that's been infectious and other people have been supporting me for it and like that. But as I'm now dealing with a larger number of people working with me, I'm still figuring out what leadership here looks like.
Jeremy Au: [00:13:25] What hurdles did you personally face and how did you overcome them?
Bhargav Sosale: [00:13:28] I would say I had many different kinds of hurdles, of course. I'd say one of the biggest hurdles, I would say, was societal/parental. I come from a fairly traditional background, right? My dad's a doctor. Actually, everyone in my family is a doctor. I was faced with the same choice when I was in class. My parents were like, hey, do you want to be an engineer or do you want to be a doctor? What do you want to do? I had those two options in front of me.
Starting a company was something that was just not on the table, right? It was not even up for consideration. When I told my parents, hey, you know what? I don't want to work. I had my first job after college, and I told my parents that I don't want to do this. I want to start my own company and see where that goes.
My parents thought, well, I think at first, they thought it was a joke. Then they're like, oh, you're serious? But what about your future? You've got a degree from NTU, and what about all of this? You can get a great paying job, why don't you just relax and enjoy life? Why do you want to start a company?
It's interesting, because they had many different concerns. I think financial was one part of it. I always hear from my mom, "Oh, you should just enjoy life. You can earn so much money, why do you want to go ahead and take all of these risks by building a company?" Right? So that was one part of that.
I think the second thing was just around general instability. Many people who try companies, 99% of companies fail. Even those that do well are still at risk of closing down in case we have an economic downturn, for example. So they're always like: wouldn't you want something not as lucrative financially as building a company, but would you rather have something that's a lot more stable? Right?
I would sometimes tell my mom, "Hey, you know what? I have a 1% chance of becoming a multimillionaire by building a company." My mom's like, "Do you want 1% chance of being a multimillionaire, or would you rather have a roof over your head every day for the rest of your life, and there's a 99% chance of that."
I would definitely face a lot of pressure from my parents with respect to that. I think, interestingly, I also see a new angle now. As I started building the company and I started scaling it up, one thing we realized was my mom and my dad are very concerned about how much little time I have left. So now it's like, "Okay, so you've done this, you've sold your company, but why do you still want to keep doing this? Look at how much time you have. You have no time for your family. You have no time for your wife. Right? Wouldn't you want to take it easy and relax a little bit and just enjoy life and just give time to all of us?" Because now the financial concern is a lot less, but, yeah, now they've found this new angle.
I just feel like we have to keep managing society's expectations of us. If they weren't my parents, I could have easily said, yeah, I don't care about you, I'm going to do what I want anyway. But because they've raised me since I was a kid, I do feel like a hurdle is I have to manage their expectations for them, and similarly with my wife as well. I mean, but they've been extremely supportive, they've been understanding, but it's still something that has been a hurdle.
Jeremy Au: [00:16:09] What support or resources would you recommend for others who are looking at a journey similar to yours?
Bhargav Sosale: [00:16:16] I can definitely share some resources that helped me a lot when I was starting out, right? I think one of the first things that got me into this journey, or helped, was basically Paul Graham, and all the essays on his website. I think that his essays, today, are something I still reflect on, I sort of reread them a lot. There was one particular essay about how to generate wealth. I actually owe a lot to those readings.
Aside from that, there are about three more books I can definitely recommend. So actually, four more books I would definitely recommend. So one of them is Zero to One, by Peter Thiel. I think it does a great job at talking about the bigger picture of economics and how that ties into company buildings. It really talks about how companies can really change the world and what separates the best companies from the worst ones.
I like that he also touches a lot on psychology and certain beliefs that he has, such as contrarian thinking, that were very new to me until I read the book. I feel that helps me be a better founder and be a better startup, if that's a word, right? So definitely for sure, Zero to One by Peter Thiel.
Another book I really liked was You Are Not So Smart by David McRaney. For me, I come from a purely tech background, right? I think for a longest time, I'm an introvert, I was a very stereotypical nerd who sat in front of his computer and coded all the time. Understanding people and empathizing with them, understanding why we act the way we do, why we think the way we do, was very, very new to me. A lot of times, I think that makes sense, right?
For me, I actually once had a breakup where a girl had said something like, "It's not you, it's me." I was like, if it's not me, then why are you breaking up with me? That doesn't make sense logically. I never understood people, right? At all.
So I think the book, You're Not So Smart, taught me a lot about Psychology. It was the first time I ever heard the word of bias. I didn't know what it was until I read that book. So it taught me a lot about human tendencies, our biases, our logical fallacies. It helped me understand people better, which helped me build better products for them.
The other book that I really liked was Principles by Ray Dalio. So what I really liked about this book was that it reinforced a lot of the way I learned things. I personally always believed that we are like neural networks, our brains are like neural networks and like machine learning algorithms. We have a certain set of beliefs, we take an action based on those beliefs. The world will respond in a way that will either tell us that our beliefs are correct or our beliefs are wrong, and we will learn from that and change those beliefs. This is exactly like how the neural network works, and that book essentially talks about that.
I think there was one last book, which was Pitch Anything by Oren Klaff. What that book did really, really well for me was that it was able to break down the emotional aspect of making the sale, emotional aspects of making a pitch. Which me, again, as someone from a completely tech background, I think needed to read that to understand how to pitch to investors better, how to sell to employees better. As a founder, you're always selling, right? So it taught me how to make that sale no matter who we're selling to.
Jeremy Au: [00:19:22] In your viral Medium post about how you took on Google and won on this health AI space, what do you think about that? How do you reflect on that startup journey against a giant and coming out the other side?
Bhargav Sosale: [00:19:38] That's actually a very, very good question. Actually, a lot of people have asked me this before. I think one of the biggest things that I now realize in hindsight is that when we were originally starting out our company and everyone said, "Hey, why are you going to beat Google?", I would go to all our investor pitches and would say, yeah, Google are doing this, this, this, this, this, this wrong. Like everything they're doing is wrong. Right? I would be so confident in that. Every investor would be like, that's bullshit. Like how can a company with that much money and that size and they are literally one of the best AI companies in the world, how can you say they're doing everything wrong? I think a lot of this came from me my co-founder's experiences going and speaking to a bunch of doctors.
I come from a family of doctors, I understand them better, so I just kind of inherently knew. But deep down, I also had a lot of doubt. Like Google does have a lot of money, all these investors are right. Am I just being very arrogant by thinking they're doing everything wrong?
Now I stuck with my gut. That's one of the biggest things we learned was we had to learn to stick with our gut and have that conviction no matter what went wrong on the outside. We stuck with our gut and our beliefs on why we were better, and we executed on those very same decisions.
Now, if you had asked me, there was probably still only two or three percent chance of success, right? Despite everything. Now, we came out on the other side and we're really glad that people are picking our solution over Google, I actually realized that, okay, look, it reinforced the lesson as a founder, you are always going to be dealing with a very small percentage chance of success. You have to have conviction in that 1% of the chance you're going to win, right? There are 99 reasons you will fail, but there's going to be one reason you're going to succeed, and you have to give a hundred percent of yourselves into that 1% chance of success.
You asked me how I felt about it. While I never had the chance to put that belief of mine into practice, now, in hindsight, I realize how important is to put a hundred percent of ourselves into that 1% chance of success, because that is the only way you're going to come out successful the other end.
Jeremy Au: [00:21:33] There are so many AI companies now, and they're all B2B, B2C. For yourself, you went on the B2B approach in the healthcare world and you competed against Google's solution and other solutions out there, right? What exactly do you think you executed better on or sold better?
Bhargav Sosale: [00:21:54] So that's a very good question. For me, I always looked at AI as a means to an end. Today, you have a lot of companies using the word AI as a buzzword to pitch to investors. They say they've got a lot of AI in it, but they really don't kind of thing. So what we did was we always knew AI is a means to an end. If there's a problem to solve, if there was a way to solve this problem without machine learning, without neural networks, we would have, right? It just so happened that neural networks was the best way to solve the problem. We were obsessed about the customer, obsessed about the problems they faced, and we used the tools we needed to solve that problem.
Now, a lot of other companies, they don't look at that. They look at AI as, "Hey, you know, it what's going to give us the money from investors." Right? Or they're excited about the technology itself, but not enough about the problem, so you have a lot of these AI companies that build a lot of tech that no one really wants. We see that happening a lot. I think if you can draw the separation between the tech and the problem, that's what would eventually make a good AI company. Where you have to realize that tech is just a means to an end.
Now I have one more point to add to that. So I start at AI being a means to an end and a lot of people being very obsessed with the technology. I mean, many people are very excited about the technology itself and not the problem. I see this most happening with data scientists, and people doing their PhD actually see that a lot, where they're very excited about it. One of the key learnings that we had is AI in academia is very, very, very, very different from AI in the real world, right?
Actually, a friend of mine read that post, and I think to quote what he said, he said something like, "I can't believe you had to say that. If I meet another data scientist who thinks that the real world is like Kaggle, I want to punch him in the face." Right? It was something along those lines.
We see this happening a lot, where everyone thinks the real world is like some kind of a Kaggle contest. But the real world is not like a Kaggle contest any more than real life programming is not like a competitive programming contest. They're two very different things. Many AI companies I see also get that wrong, where they go ahead and they think of the real world as a Kaggle contest and they try to do that. A lot of times, it's not always about accuracy, a lot of times, the data sets you're working with a very different.
Yeah, focus on the problem and solve the problem for the user and you'll have a better chance of success that way.
Jeremy Au: [00:24:14] AI seems to be this wonder solution that gets it 100% correct in an instant, right? We all know that AI is fuzzy and there are accuracy bounds that they can go up to. We also think a lot about the data that needs to be collected in order for us to train the AI to get better over time, right? How do you think companies using AI should be approaching that problem in terms of it seems to be a Catch 22, right? Like we need more data to get the results, but because we don't have the results, we can't get the data we need. So how would you advise them to think about it?
Bhargav Sosale: [00:24:56] That's a very good question. I think I've seen many approaches succeed with respect to gathering the data we need to solve that Catch 22 situation. For us, we never had a product out. In healthcare, we have to be nearly 100% accurate from day one, so we can't put a product out and get better over time, unlike many other companies. So we hustled, right? We went, we built relationships with a lot of doctors, we used our networks a lot, and we were able to get doctors to support us and our mission. We were very lucky that way. Right? I do owe it to them.
In some cases we did a lot of bartering. There was a doctor that wanted a tool to train his students on learning how to diagnose some diseases, like a very standard piece of software. Me and my co-founder, the two of us, spent two weeks building that piece of software because he said he would give us data in exchange of it. So we went ahead and did all these things in terms of bartering, and I see a lot of companies doing that. In healthcare, that is an approach that almost always works.
Another thing that works is you find another problem that isn't related, I mean that doesn't need AI to solve, but one that lets you gather data. So you could, for example, in healthcare, just build a workflow tool. You could build an EMR software, right? An EMR is an electronic medical registration software. You can build that for hospitals to use, but because they will now be linking all that patient information to it and that data is going to be stored on your servers, you now have data to work on the next problem with.
So that is another approach that I kind of see working well. Start with a workflow tool that solves another problem, but you get the data you need.
Jeremy Au: [00:26:29] Yeah, we see that with the Google Waymo versus Tesla competition of self-driving cars, right? Google has a fleet of cars that if leniently built over time to scan the roads, while Tesla is solving a very different problem where they are just having cars that happen to have auto sensors in them and just racking up an exponential growth of the miles driven and letting them be able to learn real life road conditions for driving cars. So it's definitely interesting to see those two different approaches.
Bhargav Sosale: [00:27:00] Yeah, for sure. Actually, what's interesting is I now remembered another approach that I thought was very creative for self-driving cars. I think this was in Stanford. What they did was they used GTA 5, right? Because a lot of times you can't have enough of the real world scenarios captured. Like what if you need to recreate a scenario where a pedestrian just runs like a maniac across the road when you're driving? You don't have that kind of training data. So I know a lot of people use GTA 5 to simulate various different weather conditions, conditions that you don't normally see on a day-to-day basis on the road, simulate certain kinds of accidents, and they use a game engine. That became the trading data for the self-driving car model, and it actually worked really well.
There's another company that I know, actually Singapore-based, called Bifrost that are also solving this problem. They're using, at least the last time I spoke to the founder, it was maybe six months, ago about the tech, they were using Unreal Engine or Unity to create artificial worlds to train models. Right? The use cases are crazy.
So when I was speaking with him, I asked him what a hypothetical use case would be. He said, "Look, everyone today is using machine learning. What if the Singapore Metro System wants to place cameras in the stations and they want to detect whether a train is damaged. Maybe you have a dent in the door, maybe a window is cracked, maybe a door is not closing properly. These incidents happen so rarely, but it's not like they can get that train data. But they want to keep cameras and prevent this from happening because SMRT when has downtime, everyone complains on the internet about it. How do you generate that kind of data? That can be solved with 3D modeling. Just create an artificial world that can do that, and that artificially generated data will work almost as well as real data."
So that's a very creative approach, I believe. Yeah, I see that happening now. It's an upcoming trend.
Jeremy Au: [00:28:49] Amazing. I think that's so true to see how simulations and analog solutions can really transfer from situation to situation. What are you personally excited to see more AI in? Do you see that in more the medical field? Do you look at it as real world applications?
Bhargav Sosale: [00:29:08] I don't think I'm excited by AI in a specific application. I think deep down I'm still excited by the technology in general itself, and I'm excited about AI as a field, as a scientific field. So I'm now, full disclosure, I'm being biased because of GPT-3 that just came out and everyone's talking about it. I've been trying to get access to it for the last one week or so. Haven't got it. But I'm excited by that.
Every time I think, okay, we are kind of saturated in what today's AI can do and today's set of neural networks can do, something new comes up that kind of blows my mind away. That itself fundamentally excites me because it tells me, look, I don't know about so much, and I'm potentially still underestimating the potential that's out there. I'm underestimating the potential of the technology. If tomorrow we could take GPT-3 even three steps, five steps further, I'll be very fascinated to see what that looks like. Even just a part of that imaginary future exists, that is what excites me.
In terms of applications, I look at it as, again, a means to an end of solving problems, and I just think it can solve certain problems better than others. But yeah, so that doesn't excite me as much as the tech itself.
Jeremy Au: [00:30:17] Yeah, definitely. I think it's hard for even myself to grasp like we went from AI solving chess, to Go, and now being able to be the world's best auto-complete for texts very generically with GPT-3. I think obviously I'm as excited about GPT-3, but I think the crazy part is how fast we've kind of evolved from point A to point B to point C. Like I said, now it's going to be point D, E, F in the next 10 years, where are we went from chess to GPT-3 is what we're going to see in 10 years from now what GPT-3 is to 2030, 2040, right? I think it's harder to even put your mind around what does that now let us reinvent?
Bhargav Sosale: [00:31:05] For sure. One of the things that's true with almost any kind of technology is the early days, you kind of see very small progress, but over time it starts compounding because the amount of knowledge people have compounds technically. Solutions build upon previously existing solutions, so things start evolving at an exponential rate. Which is the first chess-playing algorithm if you look at it today, it wasn't that advanced in terms of tech, that is by modern standards, but the rate at which we've gone from that to GPT-3 is crazy.
Me and Florian, my co-founder at Medios, we were sitting and playing around with nothing, we were looking at a bunch of demos online about GPT-3, and we stumbled upon one where GPT-3 generated its own neural network architecture. I was very surprised. I was like, so now we've got AI building AI. Is it just an optimization problem? Like, is it something that we do this for a couple more iterations and we have Skynet, right? I actually asked myself that question. Probably not. We probably won't reach Skynet realistically, but I do think where we will get to will be very, very impressive nonetheless. That rate is astounding.
Jeremy Au: [00:32:16] Yeah. I think it's interesting to see kind of like what we've seen in the semiconductor industry, where there's almost like an inevitable understanding of world's law. You know, how I would describe it where there's this inevitable march of increasing, improving computational power, density, power consumption, reliability, all these other attributes that you don't see in many other industries, and we're kind of seeing that now for AI. It's just been an interesting to see that in our generation.
Bhargav Sosale: [00:32:44] For sure. I think to add to that, one of the things I also think that means good for AI in this industry is that there are so many ancillary industries that are now feeding into the development of the AI industry. For example, every industry now is online, right? 10 years ago, even 10 years ago in 2010, I don't think that was the case. But now everyone's online. That means there is so much data being generated.
Now this is true for industries like you have traditional SMBs, like small bakeries, having their own websites. But I also now look at the last mile population. When you go to like a village in India, and this actually did happen. I was in a tiny village in India and people were using Tik Tok over there. I was very, very surprised. So you have now have so many people who've never been online before now coming online at an exponential pace generating the kind of data that needs to be there. A lot of other industries are adopting tech, and again, that's generating a ton more data that needs to be there for AI to grow. This is now developing better and better AI, and hence overall better online products and better solutions, which makes more people digitize much faster again.
So it's almost like a vicious cycle, right? Where people come online, AI gets better, more people come online as a result of that, AI gets even better, and then, yeah, I see that happening.
Jeremy Au: [00:34:04] That's fascinating. That ties back to the earlier point about how Tesla is able to have cars that generate data, even though they're selling cars and then the data is just like a gigantic by-product. But I think what you're saying here is that everybody's digitization efforts in every industry, I assume, from like Nest cameras, to our cars, to our IOT, and air conditioning, and everything is all spitting out data, and that generates a ton of datasets for AI to really play with.
Bhargav Sosale: [00:34:37] Yeah, for sure.
Jeremy Au: [00:34:38] So how do you stay on top of the field in terms of industry trends in terms of technology? It sounds like you've been playing around with different technologies. How do you stay on top of that?
Bhargav Sosale: [00:34:49] All right. I think it's mostly been a combination of the internet and the people I surround myself with. Florian, actually, my co-founder, he's actually a computer vision genius, right? He has spent his entire life doing computer vision since he was back in college, even before it was really a field of its own. So he generally tells me about the latest things in the vision space.
Now, similarly, EF has a great community. So when we all have our EF meetups and just catch up, even though it's been a couple years, tech is probably the central point of discussion. That's my moment where I learn a lot about things that I haven't previously discovered. That's in the offline world.
In the online world, I would say it's almost always Twitter. I didn't know about GPT-3, until I heard about it on Twitter, and that's when I did more Googling, I read the paper. But yeah, Twitter is probably number one, and then Reddit would probably be a close second. Like Reddit has some of the best subreddits. Like everything's on Reddit. So, yeah.
Bhargav Sosale: [00:35:46] Probably not. The kind of person I am with Reddit is I Google for stuff and then Reddit just happens to be on the top link. So I don't go actively looking into Reddit channels itself.
Yeah. I would say the one on machine learning is where I probably get my AI updates, but nothing specific comes to mind.
Jeremy Au: [00:36:07] Last question. How does a founder, CEO, and now post-acquisition executive de-stress?
Bhargav Sosale: [00:36:16] How would I have fun? That's a very good question. Well, I have my Xbox right here next to me. I find playing video games something that de-stresses me a lot. Outside of that, my wife has been teaching me how to swim in the last few weeks. I'm embarrassed to say I've never been a good swimmer, so she's been teaching me a lot. I think both of these things are probably on the top of that list that comes to my mind.
There are other things that I do outside of work. I try to spend some time reading. I like writing. I like wasting time on the internet and bumming around and Netflix. Sometimes I'm just scrolling through stuff on Twitter. There are those things, I wouldn't call them particularly helpful in de-stressing me. Yeah, like I browse the internet because I'm bored, but I don't find that stress relieving. Playing video games and exercise, for sure. Yeah.
I try meditating every now and then. When I have the patience and I do meditate, it's wonderful. When I don't, yeah, it's one of the things, meditation is, if I do it, it's very helpful.