Sep 19, 2018 | 3 min read

Conversation with Thuc Vu

Podcast #27: Robotics, Blockchain, AI and Game Theory

Dr. Thuc Vu is an entrepreneur and co-founder of OhmniLabs and Kambria, a decentralized open innovation platform focused on accelerating the development and adoption of the world's most advanced robotic technologies. Our conversation covered a range of topics, including the differences between industrial and consumer robotics, key advances in AI, the role of game theory in designing collaborative systems, and the promising opportunities for robotics in health care, education and senior care. We also explore the broader market dynamics in the robotics markets and the burgeoning tech scene in Vietnam.   

 

Book Recommendation:

Originals: How Non-Conformists Move the World by Adam Grant

orange-line.png

We'll notify you weekly about new podcast episodes, upcoming guests, and news. You can subscribe to the podcast and if you'd like to be considered to appear on the podcast contact us.

 

View Transcript

Good day everyone, this is Ed Maguire, Insights Partner at Momenta Partners, with another episode of our Edge podcast, and today we have with us Dr. Thuc Vu who is the co-founder OhmniLab and Kambria. It’s interesting that we were able to arrange this podcast; a good friend of mine Mai Luong who I know from the New York blockchain community, started working with Kambria, and said, ‘You’ve got to talk to these guys, they’re doing some amazing stuff’. As it turns out another friend of mine has also been working with you guys. We haven’t met in person, but I’m very excited to talk to you and hear the story. Thanks for joining us for the podcast Thuc. 

Hi Ed, very nice to be here, and thank you for inviting us. 

 

Great. First, I’d like to start off just by understanding a bit of your background. Could you talk about your background, the work you’ve done that has got you into technology, and what has brought you to become an entrepreneur? 

I grew up in Vietnam, and I came to the US for school, I have been in the US for about 20-years already. I did my undergrad at Carnegie Mellon, PhD at Stanford both in Computer Science, focusing on AI. After Stanford I started a company with a couple of other people, we were focusing on the social analytics, basically trying to understand relationships of user on social networks such as Facebook. We started the company in 2010 and then we got acquired by Google at the end of 2011, and I stayed at Google for 3-years working on Google+ and then Android. After Google I started a company called OhmniLab with Jared and Tingxi. Jared was my roommate back then in Carnegie Mellon, and so we worked very well together, we did a bunch of research projects in robotics. 

Our view at the time when we started OhmniLab was that we wanted to do a product of robots for the consumer space, because we hadn’t seen that much traction there. The only robot if you will, that sold millions of copies is the Roomba, the vacuum cleaner, but beside that we only have maybe toys or some entertainment robots. So we asked ourselves why? We have a lot of amazing technology already developed in the field for industrial manufacturing space, but for consumers we don’t see anything. Our hypothesis was that maybe because the cost was too high, the value that these robotic products offered to our end-user is too low for the cost they must pay. So, as long as the value cost is not clear, and its not high enough, people are not going to buy these products. 

We set out to develop something that had very concrete value proposition, and low cost so that people can buy them and can adopt them easier. So, that’s how OhmniLab started in 2015, and then at the beginning of last year which we started, the whole blockchain crypto space exploded. Very impressively a lot of research I did back then at Stanford was involved with game theory, and so that’s what really caught my attention right away with the blockchain, because in a sense the token models allow one to pre-design like a new game for an existing industry, or business model, to really incentivize all the stakeholders so that they can perform or behave in certain ways. So, we thought that’s an opportunity for us to design this new open platform for AI in robotics, so that with the token models, with the game theory we can really accelerate what we are trying to do with formulas, which is developing technology in the AI and robotic space that can be adopted widely by people in society. 

So, we decided to think a lot about Kambria, and feed it off with a separate project towards the end of last year, so now we have OhmniLab and Kambria running in parallel. You can how see Kambria is like Linux, non-profit building ecosystems, and in OhmniLab it kind of moves into providing different services back to Kambria, but also helps commercialize the technology coming out of Kambria. I can go more into detail of Kambria. 

 

Yes, we’ll definitely get into it, but I’m interested in your take partly on why consumer robotics haven’t really taken off. I can think of some of the work, like there was a Jibo robot and a number of crowdfunded projects over the past few years, but what have been the big challenges with consumer robotics, in your view? 

The biggest issue is that the cost is too high, compared to the value that the product will deliver to the user. So, most of the robots out there are still like entertainment used-cases, and so people are probably not going to spend more than $1,000 just to buy a product that they can talk to, see some dentists, the novelty is going to wear off very quickly. Then another issue is that technology is sometimes very optimistic, product is out there that promises the world to the user! But its high, robotic development is very-very high, whether it’s because involvement is a different expertise say, and so not only on the software AI part, but also on the hardware manufacturing part. So now putting all these challenges together, all of a sudden you either blew up the car in the time to develop the product, or you cannot deliver on the promises, and that in turn kind of harms the consumer profession of the state as well, which is something that’s unfortunate. 

 

Are there meaningful differences in the technologies that are being applied between industrial robots and consumer robots? How applicable have some of the innovations in for instance collaborative robotics in industrial technology, been able to translate to… or, how do you see the value of that applying to some of the work that you’re doing around Kambria? 

That’s a fantastic question. Industrial manufacturing robotics have been making a lot of good progress, so we had robots in all the big factories in the world, robots working in warehouses, logistic purposes and stuff like that, so, there’s been a lot of good progress there. But if you think about the cost, it’s too high, for each robot’s arm it’s going to be a couple of hundred thousand dollars, and so when we translate that to the consumer there’s a huge gap there, and so that’s the biggest challenge. The technology kind of stays around in the industrial space and hasn’t really trickled over to the consumer space yet, just because of the cost and the mindset in there. 

However, we see a lot of good progress in terms of software for AI applications, controlling for examples these arms, collaborative robot, to make it special so that its safer in the environment with humans around. Those software can be readily transferred over to the consumer space, to deploy in different applications, so I’m quite excited in that sense, and also one optimistic view is that a lot of these costs are coming down, people are more conscious about a cost factor now, and so hopefully we’ll see more highly capable arms at a much lower cost for the consumer space available soon. 

 

Great, could you talk about your vision for how robotics and robots can enhance our daily life, and the value proper to consumer? How have you been thinking about that, and are there some potential used cases that you see in the future, which will become a lot more practical as costs continue to decline? 

There are a couple of ways I’ve been thinking about this. One is, I believe that affordable, capable robots will buy access to affordable labors for the vast consumers. So, if you think about labor’s cost as arising, especially for paths in the home, or in hospital, restaurants; the wealthy class would be able to hire care-givers, or people that can help them clean the houses and stuff like that, but the cost is high. But now with the affordable robotic application, we can provide this kind of labor to a much larger branch of arms of people. This in turn would free them so that they can focus their energy and effort on more interesting paths and activities. 

The second angle is, robots can augment the existing services that are provided around the world. So, for example, a couple of different verticals that we’ve been pursing, and can see huge traction is senior care. So, in the US a lot of the seniors live on their own, the cost of caregivers is on the rise, it costs maybe on average, $5,000 to $6,000 a month to hire a caregiver, and on the other hand there’s a huge shortage of care-givers. But were saying that the caregiver services can be augmented by robots, so that they don’t have to sit in the home for 10-12 hours, and most of the time I’m not being productive. So, now we can deploy robots there and allow the caregivers to dial in remotely, control the robot providing the services, maybe just come over for a short time. So, that’s one of the verticals that we see with a lot of potential there. 

The second vertical which is also very interesting is education. There are a lot of children missing school because of serious illness, because of injuries or disability. No matter how much the parents try to provide tutoring at home schooling for these children, its not the same as going to school, having the same curriculum with the teachers, and interact with their classmates. It’s a much different world when the kid can get out and interact with people, and learn in the school environment. So, we have been deploying several robots I did for a kid that had to stay at home, to remotely dial into the robot, control it, drive around in the school, attend classes, ask questions, interact with friends, even have breakout sessions, discussions, things like that. 

Really fascinating to see how much we can change, and have the emotions they have, when the kid can go and interact with their friends. So those are the two verticals that I’m very excited in. Then a couple of other verticals, for example in the food industry we see an increase in robots helping with prepping food, especially in restaurants, or coffee shops, a robot can made coffee, those are also quite interesting. Or in hospitality vertical, robots that can deliver things to the room, which are also being piloted in a couple of places. 

 

Those are some great used cases, I know the telepresence robotics certainly made a lot of advances. I think you may have seen the Big Bang Theory show where the main character Sheldon creates his own ;Shelbot telepresence robot, it’s hard to get that image out of your head once you’ve seen it! I know that the beam robots for instance provide such an enhanced experience for people who are working remotely too, but you’re right, these earlier generations have been expensive, and not that easily accessible. 

Yeah. But another issue is, I think they’ve been focusing a lot in the enterprise marketplace, so they can be a remote worker, but there are a lot of automotive technologies there too, video-conferencing has been around for a much longer time. So, we see the opportunity in kind of semi or non-structure environments, for example in the home or in the restaurant where it’s a little bit more chaotic, and there are not that many alternative technical solutions available yet, so that’s where we believe that we can bring a lot more value to our user. 

 

I’ve been fascinated to follow some of the thinking around creating software for robots. And as you think of robots, I’d love to get your view on what some of the big engineering challenges are, at least from a software standpoint. I know there’s the concept of more effect paradox where it’s really simple for robots to do a whole bunch of often very complicated tasks, but the simple tasks that say a three-year old human child could figure out, like if something’s a door or a bookcase, or just navigating around a room, that those are some of the most difficult problems to solve for robotics. 

What do you see as some of the most interesting technology and engineering challenges, when you’re looking at building the next generation of robots? 

I think that 90 percent of effort would go into solving the 10 percent of the bad cases! Though actually you know the devil is in those corner cases, because of something that the robots haven’t been trained on, they will fail miserably, and people next time will create some consequences, and people will say, ‘Oh, no, that’s not going to work, so I’m not going to buy any robot ever again’. So, I think that’s the biggest challenge of how do we figure out these corner cases, and give people the mindset that, ‘You know what? 99 percent is going to work okay, just be a little bit patient!  

But going back to your question, I think getting the right trending data for these robots in the environment like home, or hospital, is the biggest challenge. If you think about driving a car, they make huge-huge progress in such a short time, I believe that’s because they’ve so much training data out there that they can generate or get their hands on. They can just have people driving enough miles on the street, and collecting all these pictures, images, videos, lidar information, all of that, and trend the algorithm. But if you put us in the home environment, that is not available today, that’s such a big data set, and so its going to be hard to really get some of these used cases in the home environment to be polished 

But, I’m quite optimistic because a lot of good progress has been made in areas such as machine learning, and in a way that it allows that you provide less training data to the robot, or to the AI component of the robot, so that they can learn faster and be able to have a more complex path environment. 

 

That’s interesting. You touched on another topic that I wanted to focus on, which is the application of artificial intelligence in machine learning, as driving some acceleration in the ability to solve problems. Could you talk a bit about just your views of where we are currently in the market, some of the machine learning and AI, and what you consider to be some of the most important developments that are providing the foundation for the work that you’re doing? 

Yes, I believe as I mentioned, the deep learning area has been very-very interesting, enabled us to do different things that we wouldn’t be able to even dream of, like self-driving cars just five years ago. But besides that, there’s a couple of really exciting areas that I’ve been looking at, for example collaborative robot, that’s a new mindset in which developed within the last five to ten years, that robots should be really developed to fit within an environment with humans around, instead of in a factory where the robot is automation, does its task, repeats it millions of times without regard for the environment. So, collaborative robots open-up more applications and used cases in the first semi-structure environment, for example warehouses, and the next step is to totally unstructured environments like in the home, or outside on the street. So, that’s quite interesting. 

Then the third trend which is also exciting is, the cost of a lot of these hardware components are coming down, and so that allows to create much more affordable products, and the advance of free printing additive manufacturing is amazing. I’m going to put a big bet on that as the new way to manufacture in the future, because it should significantly cut down the time and cost for us to iterate on the product. So, it can shorten it amazingly, this one, but also it can save a lot on manufacturing time as well, once we have free printers that can print multiple materials, or have multiple nozzles, and things like that. So, at OhmniLab that’s something we’ve been applying ourselves. We have been pushing on using free printers as our way to process and manufacture robots. So, 100 percent of robots are manufactured inhouse in California, using free printing. So, that’s been an interesting journey. 

 

I think what will be fascinating is the combination of these technologies that are working together to drive next-generation solutions. I’d love to get your perspective on some history of what we’ll call lower-end trainable robots; there was Rodney Brook’s project Baxter, and I know Universal Robots has had some success with more affordable trainable robots. But then you also have the Willow Garage which I guess have been working on a robotics operating system at the time, they didn’t manage to flourish, I think they shut down a few years back. 

What do you think was the reason that some of these earlier approaches may have struggled, was it an issue of the market, the technology, or other factors? 

I think it’s all the above. Willow Garage has been probably one of the iconic robotics companies in the States, most of us look up to them as the pioneer for the robotics platform. But if you look at some of the products they come out with, like the PR2, its very expensive, a couple of hundred thousand. And because it’s a general platform it’s too bulky, and the applications are not quite clear yet, so when you make a general-purpose robotics platform, you’re going to include a lot of functionalities or components that are probably not needed for a specific used case in mind. Then on the other hand, the market hasn’t seen a killer app yet for robots in the consumer state.  

So, coupled with those two, it makes it a lot harder for these approaches to get traction. So, in my opinion, I think probably the freer way to go forward is to create a platform in which it cannot have a modular approach that we can switch in, and switch out in confidence, so that developers can design and develop a specific robot for a specific used case very quickly, to keep the cost in terms of time and capital much lower, and then test the market very-very quickly to see whether there’s traction there with the user, and if not then iterate from there. Otherwise you’d spend tens of millions of dollars, and two years of developing a robot for doing ‘x’, then people say, ‘I don’t care about x’, and then all the effort and money is going to be wasted. So, I think we need to shorten that a lot, and reduce the resources that are needed to do that. 

So, this is what we really want to achieve with Kambria, growing this kind of eco-system in which people can work together from all different angles and create this explosion of robotic applications. Which is why we named it Kambria by the way! Cambrian explosion for biodiversity. 

 

That’s where we’re getting a bit of innovation, and it’s a great lead into talking a bit about your philosophy for Kambria. What’s the vision that’s really motivated you to pair up with your two co-founders? I think what’s really interesting is the way you guys describe it, each of you has different areas of expertise that come together, your co-founder Jared I see its robotics and blockchain, you’ve got expertise in AI and game theory which I want to ask you about, and then Tingxi Tan has expertise in cloud computing and blockchain; so, between all three of you, you cover a lot of different bases in terms of the technologies. What are your foundational principles for Kambria, and where do you hope to go with the platform, and the company? 

Well firstly, I’m very lucky that I get to work with this fantastic team, the people have been very helpful. So, a bit of a background story of why we started Kambria, as we’re building our robot formula, we have experienced that we had to review most of the facts, down to border controller, or the battery charging. So, if you think about it, that’s astounding, even the smart battery charging, we cannot find anything out there on the web that allows us to do what we do, there is only maybe 80 percent of what we need, but because it’s so close we have to redo the whole thing. So, we’ve been thinking a lot about there’s got to be a better way to develop robotic technology, and all the different technologies, than the way that things are being done right now, which is mostly in very siloed projects, very little collaboration across the project. 

So, we’ve been thinking about that for a while, and that’s when the blockchain space happened last year, and then my game background kind of kicked in. Maybe because people don’t have the right incentives to collaborate, if you think of collaboration in business as a game, people just playing a game, it’s like a one-shot game for example, bit of dilemma and mass equilibrium, everyone is going to cheat in a way, non-collaborate. But if we remodel it and repeat the game so that people will play the game over again, now most of them there’s some very interesting game theoretical approaches, which is we can enforce the behavior out of this game, so that we can not only incentivize but force collaboration between the players in the game. So, that’s how we designed the Kambria platform.  

 

The game theory is an interesting angle because I think you hit really on one of the critical aspects of designing a networked system that provides incentives for the participants. Could you first talk about your background in game theory, and what drew you to the discipline; and I have to say by the way that in business school 20-years ago I had a terrific professor who taught micro-economic as game theory, he participated in some of the spectrum auctions. A lot of game theory is very counterintuitive I think to people who are not initiated in understanding. But I think its just fascinating, and the fact that you are applying that to token economics, or blockchain system design, I think is so very relevant, and I’d love to get a little bit of your background and perspective how you apply what you’ve learned from your work in game theory, to Kambria. 

When I was an undergrad at Carnegie Mellon, I did a bunch of research on Robo Soccer, so basically programmed the Sony AIBO dog, I don’t know if you remember it, but it’s a little robotic dog program for a team of them to play soccer against another team. So that’s a lot of machine-learning, but also a lot of coordination between these robots. That’s what really got me interested in this space, like a multi-agency system with game theoretical angles, and machine learning angles. So, I wanted to continue that, so when I started my PhD at Stanford, I met with my adviser Professor Shohan at Stanford, and one area that he was working on which really got me interested, how you learn in game. How do you come up with a machine learning algorithm for this software programme? We call it agent autonomous software programme that they can learn how to play in different game settings. 

So again, some auto-openings that might be learning as well. So, now it opens a whole new area of thinking, because as you are learning, the openings are also learning, so in a sense you can even teach the openings to do certain things that you want them to do, in favor of your outcome, and so it’s a fascinating area. That’s the research and the background of game theory.  

When it comes to Kambria, one of the applications is, we’d been thinking about, okay, so incentivize someone to contribute to the platform is not hard, you can use tokens or even free ads to incentivize them. But how do you ensure that they will continue to collaborate and contribute in the long-term to the platform, and protect their IT so that they are not ripped off by contributing and sharing these technologies? So, we came up with something we called rule violation which is a decentralized protocol, in the sense that allows crowd protection, allows people to make sure that anyone who is using the technology for commercial purposes will have to pay a licensing fee back to the platform. But if they violate the licensing fee, they will incur severe penalties, in which that will make whatever gain they have in the short run with cheating, it’s not going to be worth that because of the steep penalties. So, in the technical terms we call it grim trigger strategy, that’s one of the strategies that we deploy in designing the Kambria token model. 

 

That’s interesting, so I guess with open source software, you always have the risk that people don’t pay for it if they don’t have to? 

Yes, they just take the technology and run away with it, and never contribute back. 

 

Right, so the incentives, how do you incentivize people to participate in your ecosystem? Do you use the tokens, or is there a system of rewards and penalties that accrue? How does that work? 

We see all of this like the open platform for people to collaborate and develop new technology. But an important angle here is that it has to be driven by something very concrete in terms of market demand, or potential. Technology just for the sake of technology is not something that we are too excited about. So, we created Kambria with different compliments, and one of the core compliments is like a marketplace for new technology, so companies, or entrepreneurs can put down these prices called bounties for some technology that they need for their market, so they can see like a huge application potential. Once we put that on Kambria, developers around the world can collaborate and compete for these bounties, which by the way it can code it in small contracts with winning criteria and all that. 

So, this way when a team of people who might not even know each other, so one guy from Russia can work with one guy from China, can work with another guy in the US, as long as they put their sharing reward in smart contracts. Once they win that bounty the bounty will then get split between the other people in this team. The technology will then get incorporated onto the platform, and open to other people so other people can get on top of it, so they don’t have to be against their will. This way we can allow technology to develop at a much faster speed than just normally one project at a time. 

That’s the incentive that it offers. Then the next step is, we’ll allow anyone to do R&D with the technology from Kambria to be free, but anyone who wants to commercialize it will have to pay a licensing fee back to the platform. This would then get split between the developers, the original person/team who put out the bounties, the token holders who back/invested into these projects, and so this way it’s like a win-win-win for everyone. 

 

That’s an interesting use of blockchain technology for building incentive around the open source development. I’d like to shift now and ask you a bit about the tech community in Vietnam, you’ve been in the US for 20-years as an entrepreneur and as a student, and I hear from several of my good friends who mentor and travel in Vietnam, that there’s just an amazing amount of talent and entrepreneurial energy. Could you talk a little bit about what the state of technology, and the entrepreneurial community in Vietnam, for Americans or anybody else thinking of looking for talent, or looking for good projects; what is a good way to engage to try to harness some of that talent? 

I think Vietnam has one of the big advantages in terms of the workforce. We have a large population of younger professionals, and traditionally Vietnam has been quite strong in terms of foundational science and programming, computer science, we have a lot of good talent there as programmers, software developers, and that’s still growing. There’s a lot of incentives put together by the government with some big companies there trying to increase further the quality and the quantity of software developers in Vietnam. So, I think the talent pool there is probably one of the biggest advantages that Vietnam has. That’s something I would recommend entrepreneurs or software founders, or even big corporation in the US pay attention to, because you can really make good use of that access. The cost is much lower than the software engineer in the US, but I think the quality of the top talent there are very comparable to the talents here. So, that’s something quite exciting. For us in Kambria, we set up a best profit in Vietnam, to try to harness all of that.  

In terms of the entrepreneur space startup, it’s been a trend for the last 10-years in Vietnam. A lot of people get excited about becoming an entrepreneur, and they’re putting together all sorts of different applications, all different products. Vietnam has a good mobile penetration ratio, coupled with the younger population people are very open to ecommerce, and buying new products, trying out new services online, things like that. So, I think there are a lot of good potentials there. Several mixed funds are coming into Vietnam, setting up investment vehicles there as well. I think for the last 10-years the GDP of Vietnam is one of the fastest growing in the world as well, so you can see a lot of opportunities there. 

 

It’s great, I think not that many people have paid attention to the progress there. People hear a lot about what’s going on, certainly in China, Korea, and Singapore, but I think Vietnam I hear much more about it, and it’s great to see this talent getting connected in the global marketplace. 

I want to ask looking forward, what are some of the concerns that you have? You do have an enormous amount of great technology opportunities in markets, but in your view, what are some of the big challenges to realizing your vision of much more affordable, and universally available robotics for consumers? 

There are a couple of challenges, one is like the legal and personal adoption of this technology. A lot of good advanced technology coming out which might be brought by another value, not just because the legal framework is not set up, as well as because it’s too new, it might take a long time to educate the consumer and get them used to this technology and product. I think that could be quite challenging. 

One more challenge, or more of a concern I have is, we tend to focus on different very specific areas where we see the immediate profit, but we tend to forget other areas that are as important, but not immediately profitable. So, for example I believe that pollution is a huge issue that we are already dealing with right now, but things like ocean pollution with plastic and all that stuff, it’s going to become a much bigger issue soon. But right now, not much investment is going into this space because people don’t see a way to make profit right away, so how do we mitigate this so that we can put investment into these sorts of projects that have a longer term, longer return, but would be crucial for sustainability of our society? 

 

When you look at where you’re most optimistic on the flip-side, what are some of the areas where you may expect some of the most impactful change to come sooner, than later? 

That’s a good question. I think a couple of different things that have been quite exciting; agriculture is one area that I’m very optimistic, in the US we probably don’t see it as much, but I think like all these advanced technologies in agriculture could really push up the productivity side, and this is important because we are still growing very quickly our world population. Without significantly increasing the productivity of agriculture activities, we’re going to run out of food! So, I think this is the way with AI robotics, and some other biotech’s, hopefully we can mitigate the shortage of food issue. 

Another area that I’m quite excited about is healthcare, I believe AI and robotics have a lot of good applications in this area, and really bringing down the cost of healthcare services, and increasing the quality of the services. I don’t think technology is going to be able to replace doctors and nurses right away, but at this technology can augment whatever they’re doing to reduce error rates, and increase the quality of diagnosis, increase the quality of treatment, and the quality of the services as well. So, those are the two areas that I’m most excited about. 

 

That’s great. One final question I like to ask on all the podcasts is, a book recommendation or a resource recommendation that you could share with our listeners. 

Since we’re talking about innovation technology and all that, one of the books that came immediately to mind is the book called ‘Originals’, by Adam Grant. That is a fascinating book talking about how we can really push new ideas and innovation at the right time and right place, to make the most impact out of it. It’s quite an interesting book. 

 

That’s a great recommendation and thank you very much for that. This has been a fascinating conversation and I’m so glad we were able to connect. 

Again, this is Ed Maguire with Momenta Partners, and we’ve been speaking with Dr Thuc Vu. We’ll include the recommendations in the shown notes, but thank you so much for taking the time, and very much looking forward to seeing your progress in the future. 

Keep in touch and thank you for having me on the show, and thanks to the listener. 

 

[End] 

Subscribe to Our Podcasts