'Welcome To The Future': Prannoy Roy Speaks To Silicon Valley Innovators
In this three-part series of one hour each brought to you by Prannoy Roy, we look at some actual products being created, some top secret, some soon to be launched - each one has something really special and mind-blowing. We meet many of the actual innovators who are helping design a better world for tomorrow.
Prannoy Roy Speaks To Ajay Madhok, Silicon Valley Startup Builder
NDTV: Well Ajay here we are in Playground. Very appropriately named. This is what you do all your life? Just play. No, this place is amazing. And you have been closely associated with the founders of Playground, and you are closely associated with them now. What makes it tick and what makes it different?
Ajay Madhok: I think they have changed the game and the business of innovation. When we were growing up, capital was very scarce and there was abundant talent. You could do Xerox Park, an AT&T Bell Lab, but to raise money was tough. Money was tough. But today it has flipped. There is plenty of capital, at least in Silicon Valley. But talent has become extremely scarce. And under those circumstances, how do you innovate? What Playground brings to the table is human capital and financial capital under one roof. Expertise, talent and capital working together.
NDTV: So both under one roof?
Ajay Madhok: So they have this idea of start-up studio and VC under one roof.
NDTV: So when you say start-up studio, what does it mean? We know VC is a venture capitalist who provides money, takes risk, has faith in founders etc. What is a start-up studio?
Ajay Madhok: So VCs provide capital and accelerators, they provide everything other than capital, like infrastructure and so on. One way to think about it is to imagine if you are building a spaceship, a rocket, then a start-up studio would be like a launch pad for that rocket. It would be the mission control, ground staff, ground troops, everything that you need to go from 0 to 1. So it is this bunch of people that you see here who have been serial entrepreneurs, who have built products, who have taken many companies from 0 to 1, who can help accelerate a new start-up.
NDTV: When you say taken a company from 0 to 1, what do you mean?
Ajay Madhok: You have an idea, 1 is market outcome. It could be any outcome. It could be taking the product to market. But it is transforming this idea into a market outcome is the journey that we say is 0 to 1.
NDTV: Or you could be bought out or something.
Ajay Madhok: That could be an exit. That's another market outcome.
NDTV: And the reason why this works is what? The four founders here and you are very very closely associated with Playground. You come here almost every day. I don't know how much work you do. I see you pretend to work just like all of us. They, why are they ideal for any start-up?
Ajay Madhok: Start-up has some challenges. So money and open source software has lowered their barriers to entry. But it's really about this journey from entry to exit where there are a lot of these pitfalls. It's like the snake and ladder game. You move and a snake bites you and you come back down. So a start-up goes through that you know. I have a vision and how do I translate that into a product that market would want? So we call it product fit, market fit, what's my go-to market strategy? What's the right insertion point? Pricing, first customer shipment? These are real challenges that every start-up has. So what this expertise brings to the table is that you know they help you avoid these early stage pitfalls. It's like playing chess with a grandmaster.
NDTV: They say they have been there, done that.
Ajay Madhok: Exactly. They can a priori say that you know these five things we don't need to worry about, let's think through these two.
NDTV: So the difference between a VC and Playground is what?
Ajay Madhok: VCs provide capital. They look at an idea, the team, the size of the market and they have certain heuristics to decide. It's a funding decision. Here, they provide expertise as well.
NDTV: So it's a VC plus.
Ajay Madhok: They are a fund and a studio. Okay, you got funded. It's like you got selected to play cricket. So you just bat well. Let me coach you.
NDTV: So they both fund but they look at the team, they look at the ideas, they look at the founders and they fund and they dole out the expertise?
Ajay Madhok: So there are two different business models. The studio's model is that I will take you from 0 to 1 and I will take a stake in the company. I will not charge you any money but my fees are really X %. I take X %. I have real skin in the game. I want you to be successful. The fund has a different business model which is investing in a portfolio of companies and they take an equity stake, but they're putting capital to get equity. Studio is putting talent to get equity. So the Playground's studio is the decision making partner, we call them the general partner or the Playground fund. So they get to decide which investments to make and the studio says come let me take you from 0 to 1. So there's financial capital and human capital together.
NDTV: Got it, got it, that's huge. Traditional VCs must be a little scared of a place like this.
Ajay Madhok: This is the breakthrough innovation model today.
NDTV: Breakthrough innovation model. I was just walking around this morning, fascinated. But one sort of common theme is AI everywhere. How important is AI today and how is it going to change the future of all our work and life?
Ajay Madhok: You saw it. When I was trying to get an internship here, Cheryl said your job is now done by the cart. You know, I could have only brought coffee. So that job is gone and I have to re-skill myself. But jokes aside. When computing became cheaper and more reliable, it found many uses. Be it the military, the government, that is where it started and as it became cheaper, the demand for its applications went up. Now it is everywhere. We carry a computer with us. It's the same thing with AI. If you think about it, it doesn't really have intelligence or judgement just yet. But what it can do is that it can make sense of the data, to predict what the data is not telling you directly. It's like when we see something and we infer. It has many layers. If you unbundle all those layers and look at predictions, AI is very good at predicting. For example, autonomous vehicles. What would a human do? If you can just predict that, you have re-imagined the problem, you have reframed the problem from autonomous driving to a prediction problem. So what AI is doing is lowering the cost of prediction and making it more reliable. So it will fundamentally transform any business that uses decision making. The cost of decision making comes down.
NDTV: All decision making requires some element of prediction, so it uses AI and it would be much easier.
Ajay Madhok: So the textbook use cases are say, trading and you're trying to predict whether the stock will move up or down. That is an easy one. But we all have experienced the transformative power of e-Commerce. Can Amazon ship before you shop? It can predict what you might need tomorrow morning.
NDTV: Or should I start selling this product or not? How much should I produce? AI will look at entire, behavioral aspects, similar products and give a better idea of...
Ajay Madhok: So any workflow can be decomposed into tasks. So certain tasks would fall under the category of predictive.
NDTV: Which is almost the basis of life.
Ajay Madhok: So anywhere you need a prediction, it will be...
NDTV: So is it going to be like, you know when computers came in a long time ago and we thought it is the end of all jobs, but actually computers have created more jobs by improving productivity, by making things that were previously impossible to imagine. Is AI going to, and I know it's difficult to tell, cut jobs or increase jobs?
Ajay Madhok: Think of economics of complements. Anything which can be substituted with cheaper products will get substituted. If you think about decision making as prediction, intuition and judgement, so I can't substitute intuition and judgement, so those jobs will remain. But anything that doesn't really require intuition and judgement will be done better by a machine. So those jobs will get replaced by...
NDTV: So is it correct to say that AI will lose certain kinds of jobs but it will open up a whole new world, of a new kind of jobs with re-skilling, like computers did?
Ajay Madhok: The short answer is that we know how to create new problems. There will always be new problems and machines will keep catching up.
NDTV: Right! That's lovely.
Prannoy Roy Speaks To Peter Barrett, Founder And CTO, Playground Global
NDTV: So why do you call it the Playground? You guys just fool around here, yeah?
Peter Barrett: Well, it's a playground for engineers, the people who created this place have spent their entire career making things, writing software, building hardware. And now is your transition into being able to fund people, creating hardware and software. It really is a playground. We get to work with people who are much smarter and younger than we are. But we can provide the capital and allow them to concentrate on making new mistakes.
NDTV: You have done so much. So why do you come in here every day?
Peter Barrett: The opportunity to create new kinds of companies and new kinds of value for our investors is important. But for me personally, I get to go home every day knowing something I didn't know when I came to work. And that's a fundamental pleasure...
NDTV: That is huge.
Peter Barrett: ...and the scientists and engineers here are domain experts from everything from next generation computing to new classes of hardware and software to new kinds of consumer experiences, all of which I would not get to intimately connect with if I just concentrated on my own career.
NDTV: Yeah, if you were in one organization. Just have a look at it here. If you can just explain what actually, there are kind of groups of start-ups here?
Peter Barrett: Yeah, that is right. So, we have about 10 of our portfolio companies living with us here in the facility. And that's everything from a six person start-up up to, I think, about 140 people in the largest company. So, it's 80,000 sq. feet. Mixed in are our employees who have created literally billions of devices of various kinds and they help the portfolio companies build their organisations and their products.
NDTV: Right, so you have all these start-ups and then you have a team of really high tech, top people who help each of the start-ups. Is that right? Wow.
Peter Barrett: Yeah that is right. They are software engineers and electrical engineers and mechanical engineers, industrial designers, product people. All the people you need to build a modern technology company and they act as river guides and keep our portfolios on track and help them build their organisations.
NDTV: It just is amazing the atmosphere and they don't need supervision to work hard. This would not work in India because we need supervision.
Peter Barrett: Well, I think...
NDTV: They are motivated.
Peter Barrett: ...the nature of the start-ups, they tend to be driven and they tend to be deeply curious. You know, they are drivers of that desire to create, desire to make a company that has a real identity, to make stuff that has not existed before, it keeps people pretty busy.
NDTV: And we will go into some of the amazing things that are happening here. But let us just go downstairs and walk around a bit. You don't have stairs built in there do you?
Peter Barrett: Should we take the slide?
NDTV: Beginning of the end of this session.
Peter Barrett: After you... Alright.
NDTV: Let me see how it's done.
Peter Barrett: Just be careful, it is just a little faster than it looks.
NDTV: Wow. But I like the way it slows down at the end of the slide.
Peter Barrett: Well we noticed that some clothes slide faster than others and every now and then, somebody would do one bounce and slide across the floor.
NDTV: Let us just walk around here. First of all, the wonderful cafe you have here. It's all, everything is free?
Peter Barrett: Yeah, so we have you know, some of the companies virtually live here, so giving them three hot meals a day keeps them happy and healthy. It also provides a community where the companies can share each other's experiences. And we're also recruiting against Apple and Google and Facebook, who are down the street, so providing some creature comforts really helps these companies hire the best and the brightest.
NDTV: And I must say what I am also impressed by here is the music. It is all my generation. You have got a good choice of music. Even your parents have, you're well brought up.
Peter Barrett: Yeah, it switches back and forth. But you know there are multi generations here, there is a mix of old masters and young geniuses, so we are trying to teach the young geniuses the correct music so...
NDTV: And you, this is also a place where people can meet. I noticed there are a couple of tables where you can actually use them as a whiteboard.
Peter Barrett: Yup that is correct. Yeah. And you often see people from different companies sharing ideas. Right. Collaborating on, because they are noncompetitive, they do tend to help each other as part of a broader, as if it is part of a broader whole.
NDTV: You don't take on two competitive companies?
Peter Barrett: That's right, that is right!
NDTV: That makes a lot of sense actually. This looks really intimidating to a lay person but we will come to all that later. All kind of animals and vehicles and stuff. But you, the kind of areas that you are working on, the borderline of improbability and impossibility. What are those?
Peter Barrett: Well, so we are looking at things like next generation computing, things like robotics, things like the replacement for classical computing, the advent of quantum computers, which will be revolutionary. So, all of those technologies require a real deep investment of capital and great engineering, but have huge consequences when they succeed.
NDTV: So, what kind of model, do you say a huge amount of capital. Do you put in the capital and you put in the training as well? It is a new kind of model actually, that you seem to be working here.
Peter Barrett: Right, so we do put in conventional capital, but it is the human capital and the capital of experience that makes the difference.
NDTV: So which is totally different from a VC for example?
Peter Barrett: Right. We're the only VC in the world that have 50 engineers who have made the mistakes, who have built the products before, who know where the bodies are buried, that can allow our companies to focus on the new mistakes, to be able to build without suffering the infant mortality or the basic simple problems that kill a lot of these companies and let them put their intellectual capital and their actual capital into building their core differentiation.
NDTV: Wow. So, your experience and your team of 50 you said, wow. And they help people not make the conventional mistakes again.
Peter Barrett: Right, and we can connect them with, you know, with technology, with ideas, with recruiting help, with architectural oversight and just good advice to keep them out of trouble.
NDTV: Wonderful. But before we, I would like you to explain to the average person what quantum computing you are doing and what it is. But before that, I would like to know, which is the most aggressive, because I am you know, I need a bit of aggression.
Peter Barrett: Well I think they are all fairly aggressive. Some of these have been in movies, in Transformers.
NDTV: In movies?
Peter Barrett: Yup, so they don't encourage good behaviour, but they are quite beautiful pieces of sculpture.
NDTV: Wow, they are just beautiful.
Peter Barrett: They are a little too loud to ride indoors but...
NDTV: You don't allow that? My God. They are from different movies basically.
Peter Barrett: Yeah, and they are all one of a kind and they are all not particularly fuel efficient.
NDTV: Right. So, tell me, back to quantum computing. Give me an example of what you are trying to achieve here with quantum computing.
Peter Barrett: Well, when we have quantum computers, large-scale quantum computers, we can solve a number of problems which are long standing problems in material science and in medicine. A lot of things we do today with, say catalysts for making agriculture, right. The ability to make fertilizer is a high temperature, high pressure process and we have no idea mechanically how it works. We produce fertilizer with an extraordinary amount of energy, and bacteria, do it at room temperature and we have no idea how that works. We simply don't understand that mechanism, but with a quantum computer you can. You can go and actually do the engineering to create, firstly to understand the catalyst we have, but also to create new ones that are obviously more efficient. So with a quantum machine, an engineer can create fertilizer at room temperature the way Biology does.
NDTV: We can't recreate Biology, which is one of the most complex natural forms of transformation.
Peter Barrett: And we will also be able to engineer catalysts and materials that go well beyond what nature does. We don't understand how photosynthesis works. We don't understand the mechanism at the core of photosynthesis. And we can design a process which does the same thing, which is enormously more efficient. We could have buildings that literally sublime out of thin air. Trees do it. Trees are made of air. Right, there is no reason why we can't grow buildings the way nature grows trees.
NDTV: Grow a building?
Peter Barrett: So, there is carbon in the air. There's all the material in the air required to build a building. You can imagine with the right catalyst and the right material science, having a building grow out of literally thin air and in the process take carbon out of the air which is something that has a key value.
NDTV: That is mind blowing. So, one day, when you have done your research, well it's about to happen. You will have a seed for a building. You plant a seed and you will have a building. Wow!
Peter Barrett: Well, there is no laws of Physics to be broken. It is just engineering and having the right computation to do that engineering.
NDTV: Wow. That is mind blowing. And you also mentioned you are working on optical computing. What do you mean by that?
Peter Barrett: That is correct. So, there is, one of the limitations of computing today is to move electrons around inside silicon takes enormous amount of power, generates enormous amount of heat. And there are limits as to how small you can make a device that uses electrons to do the computing. Photons are much easy to move around. They are potentially much, much lower power and you could do computation on a huge scale, which physically isn't possible with electrons and silicon. So we have a number of efforts on the way to use optics instead of electronics to build next generation computers, to build things that can accelerate machine learning and be faster than conventional electronics.
NDTV: So give me an example of something you are working on, or a problem you are trying to solve, very specific. If you would like to.
Peter Barrett: Well, yeah. There are some things I will definitely love to be more explicit about...
NDTV: No no, of course you can't.
Peter Barrett: ...but for example, we had a company here that created the first silicon based neural network accelerator. And you can tell a private company, and as machine learning becomes more and more central workload in the Cloud, we recognise that there is a limit to how good it can be with transistance. So, we have a company that is building a neural network accelerator, which is entirely doing the heavy lifting in optics. And it can be many, many, many times faster than its electronic equivalent.
NDTV: So, when you talk about neural network and machine learning and deep learning, in lay terms what would that mean?
Peter Barrett: So, there has been a revolution in the way we construct software, that allows machines to do complicated tasks, which up until now have been thought of as really impossible to do with human intelligence. And also, algorithms that replace a lot of handmade code with behaviour, which is learned directly by the machine. So, for example, I think we have seen the technologies that looking at the photo of a cat recognises the cat or dog and its breed, very difficult to design in our algorithm to do that, but now teaching a machine to do that is very straight-forward.
NDTV: This is visual recognition?
Peter Barrett: Right, so perception tasks, and really what you are doing is taking a high dimensional space, which is very abstract or a complicated idea of a cat, and then training a machine to produce that to the dimension of cat or dog or a person or a tree. And that is kind of eerie, right. It looks like the machine is intelligent, but it doesn't broadly apply to the deep intelligence that we have, to the general intelligence that we have yet, and won't for a long time.
NDTV: And you get a lot of Type 1 and Type 2 errors, false positives, as you get better that gets reduced, does it become more and more accurate?
Peter Barrett: Yeah, so we are still in, actually, early days for those kinds of technologies, and they do improve and they are improving very, very quickly. We are getting good at training with things like this optical technology, that can really broaden the kinds of things it can recognise and kinds of perception tasks you can build. But it will absolutely change everything about the way we write software to solve those kinds of problems.
NDTV: Amazing, amazing. It is really cutting edge stuff and what, something you mentioned about for health technology? Give us some examples of that.
Peter Barrett: Yeah, so there is a number of different domains you can apply machine onto. The way we discover drugs for therapeutics is pretty artisanal. At the moment, there is just lots of guessing and testing and experimentation. And we are starting to seek companies applying machine learning to that challenge. Machines are already better at certain oncology and pathology tasks than humans at recognising cancer and recognising what is a cancer cell. Largely because machines can do, a doctor can only do a couple of hundreds of those a year, machines can do millions.
NDTV: That's very useful for India because we have, I mean our doctors are most hard working, they see the most patients, but they just have a little bit of time for each patient, otherwise it is a few versus ignoring the rest. And this could be transformational for a developing country, third world countries.
Peter Barrett: And there is another thing that people don't really appreciate yet and that is, an extraordinary doctor can create this knowledge during the course of his career, but when he leaves or when he dies, that knowledge goes with him.
NDTV: If we could impart only so much, 10%, yeah that is true.
Peter Barrett: But machines can share each other's experiences, right. So if we could read each other's minds and share each other's experiences and knowledge, we would organise a civilisation differently. Machines can do that. So once they learn how to solve a particular problem, once they learn how to recognize a particular class of drugs, all machines can do that.
NDTV: And then share it with each other.
Peter Barrett: Yeah, so it is a huge exponential amplification of the utility of this consistency.
NDTV: I don't know whether it's confidential although there is a website, one of your teams you are building up reverses Type 2 diabetes?
Peter Barrett: That is correct, yeah so that is, I am actually Type 1 diabetic myself, is a way of using a different kind of continuous care, where coaches and doctors can work with patients to really change the outcome of Type 2 diabetes. And the treatment is, it's a mix of coaching and I had other interventions, which for the first time really dramatically change the way it progresses.
NDTV: And it is already working?
Peter Barrett: Yeah, there are patients who are currently under treatment.
NDTV: And is there a website that people can go to?
Peter Barrett: Yeah, the company is called Virta Health.
NDTV: So, Virta Health, virtahealth.com?
Peter Barrett: That's correct. And currently the treatments are available in North America, but it's inspiring CEO, and I have never seen patient testimonials like that, it is really an extraordinary programme.
NDTV: God bless you. Before we kind of come to the final point, driverless cars, it is like the 'in' thing these days. What is your take on that?
Peter Barrett: Well, the driverless cars are interesting, it speaks to you what you are talking about with machine learning. That we have seen them do things which seem like human capabilities, to recognise a cat, to be able to do these near perception tasks. And I think we have been seduced into thinking that applies to broader forms and deeper forms of human intelligence, by the human priors, to know how to get onto the hole in a tunnel, or navigate to an intersection in Mumbai. It is a very complicated human task, which requires all human priors to get right, and we have been trying to make these driverless vehicles like independent sentient creatures, to work the way we do to solve their problem. But they don't need to, because they can read each other's minds. They can see around corners...
NDTV: Oops, I'm sorry this is pure coincidence.
Peter Barrett: This is an autonomous vehicle who has been delivering donuts.
NDTV: Hey, come here, come here. Hello. Wow, delivering donuts.
Peter Barrett: And there is one donut left.
NDTV: Well actually that is me, because that is how I started my career. I promise you. I used to go around the office delivering bread and cakes and coffee.
Peter Barrett: That is great. The Canvas. This is a very simple autonomous vehicle and it does not have to deal with traffic and it does not have to stare down somebody at intersections to cross traffic.
NDTV: So, do they like to be patted, do they have...
Peter Barrett: No, they have buttons but not emotions.
NDTV: So, tell me, back to the driverless car. There are different levels, right, of how good they are. What level are we at and when will the driverless car, when will it become actually implementable?
Peter Barrett: So currently there are a series of affordances in a car for keeping you in your lane, giving you warnings if you are approaching another car. So these level 1 and level 2 features are available today. So increased control and there will be increasing features that make it more difficult for a car to run into another car, for a car to drive into...
NDTV: Level 3 or level 4.
Peter Barrett: Right, now, when you get to level 4 and level 5, there is increasing of autonomy of the vehicle that doesn't require the human to do the driving. Now at level 5, the idea is you can drive anywhere, under all circumstances and there is no human interaction.
NDTV: At all circumstances, zero human interaction.
Peter Barrett: And I think that is many decades away.
NDTV: Many decades, that is interesting.
Peter Barrett: One step down, the vehicle almost always drives itself at level 4. And I think the challenge there is, you can make a car that takes over when humans make a mistake. Very difficult to make a car that makes the human take over when it makes a mistake. And I think we have seen recent tragic examples of that, I don't think you can do that. Now we have had a robot in vehicles for many years which is really, really good at helping humans to understand circumstances. So, analog braking is a robot that is really, really good at applying brakes, far better than any human can. I mean it does it by modulating the brake.
NDTV: So, he is off you know, you're not taking any cookies, nothing.
Peter Barrett: Yes, I guess he is going back to the kitchen for more donuts.
NDTV: So, analog braking is working well.
Peter Barrett: And I think the path to autonomy lies incrementally adding more robots like that, the drive assistance, make sure that cars don't run into each other. But I think there is another general way of approaching the problem, which is make these machines the mind readers that they can be. Let them see around the corner by seeing through each other's eyes and collaborate with the cities that they're driving in, I think that's the way to scale...
NDTV: That is level 5, but we're level 4 now. Level 4 is a little dangerous.
Peter Barrett: There aren't many level 4 yet.
NDTV: But even at level 4, even 1% of the time you need a human to take over. Should that be allowed?
Peter Barrett: So, there is a body folk who thinks that level 4 should be illegal. And I think it's an intractable problem.
NDTV: I thought you thought that there are no problems that are intractable. That's why we are going to get into this.
Peter Barrett: There are always other ways of solving the problem. The promise of autonomy, the problem is freeing people up from mundane tasks. I think it is the implementation, there is always an implementation of this and I'm not sure this action of autonomy is the right way of approaching at it.
NDTV: So, you think, either level 5, you certainly don't have level 4 running around the roads?
Peter Barrett: I think an autonomous system that almost always works, is not an autonomous system at all.
NDTV: An autonomous system that almost always works is not autonomous. I think that is a really good point and I hope everybody hears that, because it is worrying and the distinction is very clear you made. Last question in the context of, what next? Because there is lots of big ideas being cracked every day and new learning as you said.
Peter Barrett: Yeah, I think that within these walls, there are some absolutely revolutionary companies. When they succeed, that will enable new companies and new industries to be formed. I have heard one of the CEOs here say that he believes that the computing revolution has not happened yet. I believe that. I also believe that industrial revolution has not really happened yet either. And the advent of next generation computing technologies and the possibilities for engineering that follow that, really will change the way we work and live, create an extraordinary amount of wealth, create enormous opportunities for investment and really change the way we live.
NDTV: You are saying there is a whole revolution to come both in computing and just in technologies. Just give us a glimpse of what the world is going to be like 15 years from now. How the change, 15 may be too long, I don't know.
Peter Barrett: Hopefully, we are in an environment where we can reverse the damage we have done over the last couple of centuries. To be able to have more and more of wealth population living in fair and equitable environment.
NDTV: In what way?
Peter Barrett: Well, to be able to create wealth, to have high quality, inexpensive food, to be able to have, you know, to be able to grow buildings...
NDTV: High quality, inexpensive food is like huge for a country like India, huge...
Peter Barrett: And then we are in a position to address the limitations of existing agriculture, with things like, understanding how to create a fertilizer without room temperature, without the Haber process for example, to be able to apply practices for automation, to be able to inexpensively address limits in labour to create the food, engineering different kinds of plants that are adaptable to different climates and different environments to the soil.
NDTV: Or productive, cheaper cost. I mean that's a revolution in agriculture in itself.
Peter Barrett: I think there are huge opportunities in technology to address those kinds of things. And that the more people we put in the position to spend time being scientists and engineers, creating this kind of wealth, is a positive feedback that addresses more and more of those problems, right. We are, I think India produces in the order of magnitude more engineers than the US does. And science is, for some reason, under siege in the US. But there is great positive feedback though, with the right capital and right engineers and people being able to concentrate on those kinds of problems, the opportunity to create wealth and make it better for everybody is virtually unlimited. And as I say we are just at the infancy of these new computing technologies, which unlock material science and chemistry and medicine in a fascinating way. So I think most of the good ideas have not been had yet, most of the interesting companies have not been invested yet and we want to make a contribution to that as much as we can.
NDTV: God bless you, and best of luck. Love the world you are looking at. The world we are looking at.
Peter Barrett: It is a delight to be a part of it, thank you.
Prannoy Roy Speaks To Kunal Ghosh, Founder And CEO, Inscopix
NDTV: Well Kunal Ghosh, really good of you to spare your time.
Kunal Ghosh: Likewise, thank you.
NDTV: Kids like you, you make us proud. And you have a lot of work to do and taking time off like that is wonderful.
Kunal Ghosh: Thank you sir.
NDTV: Tell me a little bit about your company. It is really inspiring, what is the name of it, firstly?
Kunal Ghosh: Inscopix.
Prannoy Roy: What's the big picture that it's trying to do.
Kunal Ghosh: So Inscopix is a neurotech company. We are based right out here in Palo Alto, spun out of Stanford several years back. We were founded with the mission to help decode the brain, this black box. We have very little understanding of the brain.
NDTV: The brain is a black box. I love that.
Kunal Ghosh: The brain is a black box and it has been a black box for decades. It is precisely because we know so little of about how it works, let alone how it doesn't work, that we are where we are today with respect to brain diseases and mental illnesses. The statistics are just mind boggling. You know, something like 1 in 10 adults worldwide suffers from mental illness.
NDTV: Wow, wow.
Kunal Ghosh: It is hundreds of millions of folks around the world, whether it is diagnosed or undiagnosed. The global cost is, I think, something like three trillion on society. It's what it is today because we don't understand how the brain works. We don't have good treatments, let alone cures. So our thesis is to say that let us empower researchers around the world to address this knowledge gap. Let us demystify this black box. You know, enlightened researchers with our platform, our analytics.
NDTV: So you are studying, understanding the brain, which is a huge breakthrough. Is that what you are doing?
Kunal Ghosh: I like to say we are empowering the heroes out there, who are going to be making the breakthroughs, with our technology, with our analytics, with our personnel. We certainly are investigating some specific brain functions and some particular processes, but it is the heroes out there at universities, research institutions, pharmaceutical companies that are using our tools, our products and are trying to understand how we encode memories. How do we get addicted to substances? How do we eat? You know, what are the behaviours that result in binge eating that results in obesity, which is a big problem in the US. These are all brain circuit problems. Our platform is, for the very first time, literally shedding light. It is an imaging based device, so it is literally shedding light on the circuits that no one understood, how they gave rise to basic functions that we take for granted. In the case of diseases, like Parkinson's, together with drug companies, we're for the first time discovering how circuits are malfunctioning.
NDTV: Circuits in the brain? Wow.
Kunal Ghosh: Brain circuits, how they malfunctioning and giving rise to the symptoms. And the hope here is that if we can understand how the circuits malfunction, we might be able to develop much better treatments that re-tune the circuit, in some ways analogous to controlling blood pressure. If we can control blood pressure, the person can lead a long healthy life, hopefully, not succumb to cardiovascular diseases.
NDTV: So you can do a similar kind of thing with different aspects of the brain?
Kunal Ghosh: Exactly, that's the hope. That once we demystify this black box, we will understand better what to do to help people.
NDTV: So what you are doing could, apart from understanding all these, the binge eating, also could help with Parkinson's and Alzheimer's?
Kunal Ghosh: Absolutely. In fact that's the big opportunity that we have here at Inscopix and you know, for humanity, that if we can start understanding in a disease like Parkinson's - how the Parkinsonian circuit in a brain is malfunctioning, we can distinguish normal from the abnormal, then we might be able to empower drug developers, pharmaceutical companies, device developers to do much better than sticking crude invasive electrodes in the brain. Very crude, sledgehammer like treatments is what we have today. This is the state today.
NDTV: Let me just bring your eyes on this. Inscopix, this is one of your products, isn't it? It looks very well developed actually, already! Explain a bit of it, just for a layperson.
Kunal Ghosh: Sure. So as I was mentioning, the technology actually spawned out of work we did back at Stanford. And this is the core of the platform. It is a miniature, wearable microscope. It is a microscope-based technology, with an integrated HD video camera. In many ways it is similar to the chips in the camera that is currently photographing, filming us. So this device will be head mounted, not on human subjects yet, since right now we are empowering researchers to understand basic circuits that are malfunctioning, and disease models with the hope of then translating later to humans. But researchers will be able to watch, with this device, hundreds of thousands of neurons firing in real time during behaviour. So you can really start studying the brain action and visualise how brain circuits, neuronal processes and patterns give rise to behaviours.
NDTV: Interesting that you say that these are cameras, it helps visualise. So like this would be fitted where, eventually?
Kunal Ghosh: This would be head mounted
NDTV: Like, here or here? Or inside?
Kunal Ghosh: On the cranium. Or it could also be implanted, in a minimally invasive way, on the structure that is being studied. Right now it is for research purposes. So it is on research subjects and is used to understand, as I was saying, how specific neurons are encoding for memories. In the Parkinson's study that I was mentioning, this platform, this technology showed how in a particular animal model of Parkinson's, we knew exactly what was happening in the circuit that gave rise to the behavioural abnormalities. And this is the kicker, that most drug companies today rely only on the behavioural read outs, if you will, that are very crude, very coarse. And then most of the drugs end up failing in humans because the animal behavioural model is in no way a predictor of how well the drug will work in the clinic. With our technology, by looking at the brain circuit, in vivo, we can now show literally how the compound, the drug is working on the circuit.
NDTV: I see.
Kunal Ghosh: And it's way more sophisticated than the behavioural read out. Meaning that we can tell whether a drug will work or not, predict its clinical efficacy based on the circuit biomarker. So it is changing the paradigm in neuroscience, neurology, neuropsychiatry. From looking at the brain just as a collection of chemicals, that is how we have looked at it, inject some more dopamine, or pull back some dopamine, that is how we treat patients, into looking at the brain as a big electrical network. And if we can decode this network, then we can understand how to re-tune the network with precision therapeutics and hopefully improve the quality of life, if not curing some of these diseases in the long term.
NDTV: And when do we think, like say Parkinson's or Alzheimer's could benefit from this?
Kunal Ghosh: Well, in Parkinson's we already are in the process of conducting larger scale research trials, where we are starting to develop AI-driven methods to predict the accuracy or clinical efficacy of a compound on a device. And the hope is that in five to 10 years, some of these results will start translating into therapeutics, that could be used in humans and hopefully do much better than anything that's out there today.
NDTV: Anything about Alzheimer's?
Kunal Ghosh: Alzheimer's, mental illnesses, um, mental illnesses are extremely poorly understood, so at least a disease like Parkinson's, at least there are some therapies in the clinic. You know we have deep brain stimulators, it might be crude, we have L-DOPA, might not be optimal, but at least there's something for early stage patients. You look at depression, schizophrenia, bipolar, there are no good therapies out there. Diagnosis is also very primitive. So this kind of a technology is extremely well suited to look at mental illnesses, understand how the circuits are being disrupted and eventually...
NDTV: So what do you actually see?
Kunal Ghosh: You will see literally hundreds and thousands of neurons that are flashing, like stars in a night sky, the brain lights up and you can see a bunch of neurons that are firing and the patterns of firing result in the function or the behaviour that we take for granted.
NDTV: Wow. That is a huge breakthrough.
Kunal Ghosh: Well that's endoscopics and that's our promise, that if we can start decoding these patterns' effectivity, let us first provide these patterns' effectivity. Let all our researchers, pharmaceutical companies see for the very first time the brain in action. President (Barack) Obama, in the State of the Union, said this, probably more eloquent than I am, and I think in 2013, that this three pound mass between our ears is a mystery. But if there is a way to start unlocking this mass, if there is a way to literally study the brain in action, to visualise the brain in action, then we will be, hopefully, at the next phase of human evolution, where we are finally starting to demystify this organ, conquer brain diseases and just improve the state of the world. Inscopix today is one of the few companies at least, that is literally empowering researchers to do that, to study the brain in action.
NDTV: And you are a part of the Playground?
Kunal Ghosh: We are a part of the Playground. We love the fact that Playground invests in platform companies and that's what Inscopix says, it is a platform for mapping the brain and eventually, of course, for enabling the development of next generation therapeutics.
NDTV: Kunal, you will be successful. You are a Bengali, so there is no question. Good luck and God bless you. You are working in such an important area.
Kunal Ghosh: Thank you. I appreciate it, my pleasure.
Prannoy Roy Speaks To Ajay Madhok, Silicon Valley Startup Builder
NDTV: Well Ajay here we are in Playground. Very appropriately named. This is what you do all your life? Just play. No, this place is amazing. And you have been closely associated with the founders of Playground, and you are closely associated with them now. What makes it tick and what makes it different?
Ajay Madhok: I think they have changed the game and the business of innovation. When we were growing up, capital was very scarce and there was abundant talent. You could do Xerox Park, an AT&T Bell Lab, but to raise money was tough. Money was tough. But today it has flipped. There is plenty of capital, at least in Silicon Valley. But talent has become extremely scarce. And under those circumstances, how do you innovate? What Playground brings to the table is human capital and financial capital under one roof. Expertise, talent and capital working together.
NDTV: So both under one roof?
Ajay Madhok: So they have this idea of start-up studio and VC under one roof.
NDTV: So when you say start-up studio, what does it mean? We know VC is a venture capitalist who provides money, takes risk, has faith in founders etc. What is a start-up studio?
Ajay Madhok: So VCs provide capital and accelerators, they provide everything other than capital, like infrastructure and so on. One way to think about it is to imagine if you are building a spaceship, a rocket, then a start-up studio would be like a launch pad for that rocket. It would be the mission control, ground staff, ground troops, everything that you need to go from 0 to 1. So it is this bunch of people that you see here who have been serial entrepreneurs, who have built products, who have taken many companies from 0 to 1, who can help accelerate a new start-up.
NDTV: When you say taken a company from 0 to 1, what do you mean?
Ajay Madhok: You have an idea, 1 is market outcome. It could be any outcome. It could be taking the product to market. But it is transforming this idea into a market outcome is the journey that we say is 0 to 1.
NDTV: Or you could be bought out or something.
Ajay Madhok: That could be an exit. That's another market outcome.
NDTV: And the reason why this works is what? The four founders here and you are very very closely associated with Playground. You come here almost every day. I don't know how much work you do. I see you pretend to work just like all of us. They, why are they ideal for any start-up?
Ajay Madhok: Start-up has some challenges. So money and open source software has lowered their barriers to entry. But it's really about this journey from entry to exit where there are a lot of these pitfalls. It's like the snake and ladder game. You move and a snake bites you and you come back down. So a start-up goes through that you know. I have a vision and how do I translate that into a product that market would want? So we call it product fit, market fit, what's my go-to market strategy? What's the right insertion point? Pricing, first customer shipment? These are real challenges that every start-up has. So what this expertise brings to the table is that you know they help you avoid these early stage pitfalls. It's like playing chess with a grandmaster.
NDTV: They say they have been there, done that.
Ajay Madhok: Exactly. They can a priori say that you know these five things we don't need to worry about, let's think through these two.
NDTV: So the difference between a VC and Playground is what?
Ajay Madhok: VCs provide capital. They look at an idea, the team, the size of the market and they have certain heuristics to decide. It's a funding decision. Here, they provide expertise as well.
NDTV: So it's a VC plus.
Ajay Madhok: They are a fund and a studio. Okay, you got funded. It's like you got selected to play cricket. So you just bat well. Let me coach you.
NDTV: So they both fund but they look at the team, they look at the ideas, they look at the founders and they fund and they dole out the expertise?
Ajay Madhok: So there are two different business models. The studio's model is that I will take you from 0 to 1 and I will take a stake in the company. I will not charge you any money but my fees are really X %. I take X %. I have real skin in the game. I want you to be successful. The fund has a different business model which is investing in a portfolio of companies and they take an equity stake, but they're putting capital to get equity. Studio is putting talent to get equity. So the Playground's studio is the decision making partner, we call them the general partner or the Playground fund. So they get to decide which investments to make and the studio says come let me take you from 0 to 1. So there's financial capital and human capital together.
NDTV: Got it, got it, that's huge. Traditional VCs must be a little scared of a place like this.
Ajay Madhok: This is the breakthrough innovation model today.
NDTV: Breakthrough innovation model. I was just walking around this morning, fascinated. But one sort of common theme is AI everywhere. How important is AI today and how is it going to change the future of all our work and life?
Ajay Madhok: You saw it. When I was trying to get an internship here, Cheryl said your job is now done by the cart. You know, I could have only brought coffee. So that job is gone and I have to re-skill myself. But jokes aside. When computing became cheaper and more reliable, it found many uses. Be it the military, the government, that is where it started and as it became cheaper, the demand for its applications went up. Now it is everywhere. We carry a computer with us. It's the same thing with AI. If you think about it, it doesn't really have intelligence or judgement just yet. But what it can do is that it can make sense of the data, to predict what the data is not telling you directly. It's like when we see something and we infer. It has many layers. If you unbundle all those layers and look at predictions, AI is very good at predicting. For example, autonomous vehicles. What would a human do? If you can just predict that, you have re-imagined the problem, you have reframed the problem from autonomous driving to a prediction problem. So what AI is doing is lowering the cost of prediction and making it more reliable. So it will fundamentally transform any business that uses decision making. The cost of decision making comes down.
NDTV: All decision making requires some element of prediction, so it uses AI and it would be much easier.
Ajay Madhok: So the textbook use cases are say, trading and you're trying to predict whether the stock will move up or down. That is an easy one. But we all have experienced the transformative power of e-Commerce. Can Amazon ship before you shop? It can predict what you might need tomorrow morning.
NDTV: Or should I start selling this product or not? How much should I produce? AI will look at entire, behavioral aspects, similar products and give a better idea of...
Ajay Madhok: So any workflow can be decomposed into tasks. So certain tasks would fall under the category of predictive.
NDTV: Which is almost the basis of life.
Ajay Madhok: So anywhere you need a prediction, it will be...
NDTV: So is it going to be like, you know when computers came in a long time ago and we thought it is the end of all jobs, but actually computers have created more jobs by improving productivity, by making things that were previously impossible to imagine. Is AI going to, and I know it's difficult to tell, cut jobs or increase jobs?
Ajay Madhok: Think of economics of complements. Anything which can be substituted with cheaper products will get substituted. If you think about decision making as prediction, intuition and judgement, so I can't substitute intuition and judgement, so those jobs will remain. But anything that doesn't really require intuition and judgement will be done better by a machine. So those jobs will get replaced by...
NDTV: So is it correct to say that AI will lose certain kinds of jobs but it will open up a whole new world, of a new kind of jobs with re-skilling, like computers did?
Ajay Madhok: The short answer is that we know how to create new problems. There will always be new problems and machines will keep catching up.
NDTV: Right! That's lovely.
Prannoy Roy Speaks To Peter Barrett, Founder And CTO, Playground Global
NDTV: So why do you call it the Playground? You guys just fool around here, yeah?
Peter Barrett: Well, it's a playground for engineers, the people who created this place have spent their entire career making things, writing software, building hardware. And now is your transition into being able to fund people, creating hardware and software. It really is a playground. We get to work with people who are much smarter and younger than we are. But we can provide the capital and allow them to concentrate on making new mistakes.
NDTV: You have done so much. So why do you come in here every day?
Peter Barrett: The opportunity to create new kinds of companies and new kinds of value for our investors is important. But for me personally, I get to go home every day knowing something I didn't know when I came to work. And that's a fundamental pleasure...
NDTV: That is huge.
Peter Barrett: ...and the scientists and engineers here are domain experts from everything from next generation computing to new classes of hardware and software to new kinds of consumer experiences, all of which I would not get to intimately connect with if I just concentrated on my own career.
NDTV: Yeah, if you were in one organization. Just have a look at it here. If you can just explain what actually, there are kind of groups of start-ups here?
Peter Barrett: Yeah, that is right. So, we have about 10 of our portfolio companies living with us here in the facility. And that's everything from a six person start-up up to, I think, about 140 people in the largest company. So, it's 80,000 sq. feet. Mixed in are our employees who have created literally billions of devices of various kinds and they help the portfolio companies build their organisations and their products.
NDTV: Right, so you have all these start-ups and then you have a team of really high tech, top people who help each of the start-ups. Is that right? Wow.
Peter Barrett: Yeah that is right. They are software engineers and electrical engineers and mechanical engineers, industrial designers, product people. All the people you need to build a modern technology company and they act as river guides and keep our portfolios on track and help them build their organisations.
NDTV: It just is amazing the atmosphere and they don't need supervision to work hard. This would not work in India because we need supervision.
Peter Barrett: Well, I think...
NDTV: They are motivated.
Peter Barrett: ...the nature of the start-ups, they tend to be driven and they tend to be deeply curious. You know, they are drivers of that desire to create, desire to make a company that has a real identity, to make stuff that has not existed before, it keeps people pretty busy.
NDTV: And we will go into some of the amazing things that are happening here. But let us just go downstairs and walk around a bit. You don't have stairs built in there do you?
Peter Barrett: Should we take the slide?
NDTV: Beginning of the end of this session.
Peter Barrett: After you... Alright.
NDTV: Let me see how it's done.
Peter Barrett: Just be careful, it is just a little faster than it looks.
NDTV: Wow. But I like the way it slows down at the end of the slide.
Peter Barrett: Well we noticed that some clothes slide faster than others and every now and then, somebody would do one bounce and slide across the floor.
NDTV: Let us just walk around here. First of all, the wonderful cafe you have here. It's all, everything is free?
Peter Barrett: Yeah, so we have you know, some of the companies virtually live here, so giving them three hot meals a day keeps them happy and healthy. It also provides a community where the companies can share each other's experiences. And we're also recruiting against Apple and Google and Facebook, who are down the street, so providing some creature comforts really helps these companies hire the best and the brightest.
NDTV: And I must say what I am also impressed by here is the music. It is all my generation. You have got a good choice of music. Even your parents have, you're well brought up.
Peter Barrett: Yeah, it switches back and forth. But you know there are multi generations here, there is a mix of old masters and young geniuses, so we are trying to teach the young geniuses the correct music so...
NDTV: And you, this is also a place where people can meet. I noticed there are a couple of tables where you can actually use them as a whiteboard.
Peter Barrett: Yup that is correct. Yeah. And you often see people from different companies sharing ideas. Right. Collaborating on, because they are noncompetitive, they do tend to help each other as part of a broader, as if it is part of a broader whole.
NDTV: You don't take on two competitive companies?
Peter Barrett: That's right, that is right!
NDTV: That makes a lot of sense actually. This looks really intimidating to a lay person but we will come to all that later. All kind of animals and vehicles and stuff. But you, the kind of areas that you are working on, the borderline of improbability and impossibility. What are those?
Peter Barrett: Well, so we are looking at things like next generation computing, things like robotics, things like the replacement for classical computing, the advent of quantum computers, which will be revolutionary. So, all of those technologies require a real deep investment of capital and great engineering, but have huge consequences when they succeed.
NDTV: So, what kind of model, do you say a huge amount of capital. Do you put in the capital and you put in the training as well? It is a new kind of model actually, that you seem to be working here.
Peter Barrett: Right, so we do put in conventional capital, but it is the human capital and the capital of experience that makes the difference.
NDTV: So which is totally different from a VC for example?
Peter Barrett: Right. We're the only VC in the world that have 50 engineers who have made the mistakes, who have built the products before, who know where the bodies are buried, that can allow our companies to focus on the new mistakes, to be able to build without suffering the infant mortality or the basic simple problems that kill a lot of these companies and let them put their intellectual capital and their actual capital into building their core differentiation.
NDTV: Wow. So, your experience and your team of 50 you said, wow. And they help people not make the conventional mistakes again.
Peter Barrett: Right, and we can connect them with, you know, with technology, with ideas, with recruiting help, with architectural oversight and just good advice to keep them out of trouble.
NDTV: Wonderful. But before we, I would like you to explain to the average person what quantum computing you are doing and what it is. But before that, I would like to know, which is the most aggressive, because I am you know, I need a bit of aggression.
Peter Barrett: Well I think they are all fairly aggressive. Some of these have been in movies, in Transformers.
NDTV: In movies?
Peter Barrett: Yup, so they don't encourage good behaviour, but they are quite beautiful pieces of sculpture.
NDTV: Wow, they are just beautiful.
Peter Barrett: They are a little too loud to ride indoors but...
NDTV: You don't allow that? My God. They are from different movies basically.
Peter Barrett: Yeah, and they are all one of a kind and they are all not particularly fuel efficient.
NDTV: Right. So, tell me, back to quantum computing. Give me an example of what you are trying to achieve here with quantum computing.
Peter Barrett: Well, when we have quantum computers, large-scale quantum computers, we can solve a number of problems which are long standing problems in material science and in medicine. A lot of things we do today with, say catalysts for making agriculture, right. The ability to make fertilizer is a high temperature, high pressure process and we have no idea mechanically how it works. We produce fertilizer with an extraordinary amount of energy, and bacteria, do it at room temperature and we have no idea how that works. We simply don't understand that mechanism, but with a quantum computer you can. You can go and actually do the engineering to create, firstly to understand the catalyst we have, but also to create new ones that are obviously more efficient. So with a quantum machine, an engineer can create fertilizer at room temperature the way Biology does.
NDTV: We can't recreate Biology, which is one of the most complex natural forms of transformation.
Peter Barrett: And we will also be able to engineer catalysts and materials that go well beyond what nature does. We don't understand how photosynthesis works. We don't understand the mechanism at the core of photosynthesis. And we can design a process which does the same thing, which is enormously more efficient. We could have buildings that literally sublime out of thin air. Trees do it. Trees are made of air. Right, there is no reason why we can't grow buildings the way nature grows trees.
NDTV: Grow a building?
Peter Barrett: So, there is carbon in the air. There's all the material in the air required to build a building. You can imagine with the right catalyst and the right material science, having a building grow out of literally thin air and in the process take carbon out of the air which is something that has a key value.
NDTV: That is mind blowing. So, one day, when you have done your research, well it's about to happen. You will have a seed for a building. You plant a seed and you will have a building. Wow!
Peter Barrett: Well, there is no laws of Physics to be broken. It is just engineering and having the right computation to do that engineering.
NDTV: Wow. That is mind blowing. And you also mentioned you are working on optical computing. What do you mean by that?
Peter Barrett: That is correct. So, there is, one of the limitations of computing today is to move electrons around inside silicon takes enormous amount of power, generates enormous amount of heat. And there are limits as to how small you can make a device that uses electrons to do the computing. Photons are much easy to move around. They are potentially much, much lower power and you could do computation on a huge scale, which physically isn't possible with electrons and silicon. So we have a number of efforts on the way to use optics instead of electronics to build next generation computers, to build things that can accelerate machine learning and be faster than conventional electronics.
NDTV: So give me an example of something you are working on, or a problem you are trying to solve, very specific. If you would like to.
Peter Barrett: Well, yeah. There are some things I will definitely love to be more explicit about...
NDTV: No no, of course you can't.
Peter Barrett: ...but for example, we had a company here that created the first silicon based neural network accelerator. And you can tell a private company, and as machine learning becomes more and more central workload in the Cloud, we recognise that there is a limit to how good it can be with transistance. So, we have a company that is building a neural network accelerator, which is entirely doing the heavy lifting in optics. And it can be many, many, many times faster than its electronic equivalent.
NDTV: So, when you talk about neural network and machine learning and deep learning, in lay terms what would that mean?
Peter Barrett: So, there has been a revolution in the way we construct software, that allows machines to do complicated tasks, which up until now have been thought of as really impossible to do with human intelligence. And also, algorithms that replace a lot of handmade code with behaviour, which is learned directly by the machine. So, for example, I think we have seen the technologies that looking at the photo of a cat recognises the cat or dog and its breed, very difficult to design in our algorithm to do that, but now teaching a machine to do that is very straight-forward.
NDTV: This is visual recognition?
Peter Barrett: Right, so perception tasks, and really what you are doing is taking a high dimensional space, which is very abstract or a complicated idea of a cat, and then training a machine to produce that to the dimension of cat or dog or a person or a tree. And that is kind of eerie, right. It looks like the machine is intelligent, but it doesn't broadly apply to the deep intelligence that we have, to the general intelligence that we have yet, and won't for a long time.
NDTV: And you get a lot of Type 1 and Type 2 errors, false positives, as you get better that gets reduced, does it become more and more accurate?
Peter Barrett: Yeah, so we are still in, actually, early days for those kinds of technologies, and they do improve and they are improving very, very quickly. We are getting good at training with things like this optical technology, that can really broaden the kinds of things it can recognise and kinds of perception tasks you can build. But it will absolutely change everything about the way we write software to solve those kinds of problems.
NDTV: Amazing, amazing. It is really cutting edge stuff and what, something you mentioned about for health technology? Give us some examples of that.
Peter Barrett: Yeah, so there is a number of different domains you can apply machine onto. The way we discover drugs for therapeutics is pretty artisanal. At the moment, there is just lots of guessing and testing and experimentation. And we are starting to seek companies applying machine learning to that challenge. Machines are already better at certain oncology and pathology tasks than humans at recognising cancer and recognising what is a cancer cell. Largely because machines can do, a doctor can only do a couple of hundreds of those a year, machines can do millions.
NDTV: That's very useful for India because we have, I mean our doctors are most hard working, they see the most patients, but they just have a little bit of time for each patient, otherwise it is a few versus ignoring the rest. And this could be transformational for a developing country, third world countries.
Peter Barrett: And there is another thing that people don't really appreciate yet and that is, an extraordinary doctor can create this knowledge during the course of his career, but when he leaves or when he dies, that knowledge goes with him.
NDTV: If we could impart only so much, 10%, yeah that is true.
Peter Barrett: But machines can share each other's experiences, right. So if we could read each other's minds and share each other's experiences and knowledge, we would organise a civilisation differently. Machines can do that. So once they learn how to solve a particular problem, once they learn how to recognize a particular class of drugs, all machines can do that.
NDTV: And then share it with each other.
Peter Barrett: Yeah, so it is a huge exponential amplification of the utility of this consistency.
NDTV: I don't know whether it's confidential although there is a website, one of your teams you are building up reverses Type 2 diabetes?
Peter Barrett: That is correct, yeah so that is, I am actually Type 1 diabetic myself, is a way of using a different kind of continuous care, where coaches and doctors can work with patients to really change the outcome of Type 2 diabetes. And the treatment is, it's a mix of coaching and I had other interventions, which for the first time really dramatically change the way it progresses.
NDTV: And it is already working?
Peter Barrett: Yeah, there are patients who are currently under treatment.
NDTV: And is there a website that people can go to?
Peter Barrett: Yeah, the company is called Virta Health.
NDTV: So, Virta Health, virtahealth.com?
Peter Barrett: That's correct. And currently the treatments are available in North America, but it's inspiring CEO, and I have never seen patient testimonials like that, it is really an extraordinary programme.
NDTV: God bless you. Before we kind of come to the final point, driverless cars, it is like the 'in' thing these days. What is your take on that?
Peter Barrett: Well, the driverless cars are interesting, it speaks to you what you are talking about with machine learning. That we have seen them do things which seem like human capabilities, to recognise a cat, to be able to do these near perception tasks. And I think we have been seduced into thinking that applies to broader forms and deeper forms of human intelligence, by the human priors, to know how to get onto the hole in a tunnel, or navigate to an intersection in Mumbai. It is a very complicated human task, which requires all human priors to get right, and we have been trying to make these driverless vehicles like independent sentient creatures, to work the way we do to solve their problem. But they don't need to, because they can read each other's minds. They can see around corners...
NDTV: Oops, I'm sorry this is pure coincidence.
Peter Barrett: This is an autonomous vehicle who has been delivering donuts.
NDTV: Hey, come here, come here. Hello. Wow, delivering donuts.
Peter Barrett: And there is one donut left.
NDTV: Well actually that is me, because that is how I started my career. I promise you. I used to go around the office delivering bread and cakes and coffee.
Peter Barrett: That is great. The Canvas. This is a very simple autonomous vehicle and it does not have to deal with traffic and it does not have to stare down somebody at intersections to cross traffic.
NDTV: So, do they like to be patted, do they have...
Peter Barrett: No, they have buttons but not emotions.
NDTV: So, tell me, back to the driverless car. There are different levels, right, of how good they are. What level are we at and when will the driverless car, when will it become actually implementable?
Peter Barrett: So currently there are a series of affordances in a car for keeping you in your lane, giving you warnings if you are approaching another car. So these level 1 and level 2 features are available today. So increased control and there will be increasing features that make it more difficult for a car to run into another car, for a car to drive into...
NDTV: Level 3 or level 4.
Peter Barrett: Right, now, when you get to level 4 and level 5, there is increasing of autonomy of the vehicle that doesn't require the human to do the driving. Now at level 5, the idea is you can drive anywhere, under all circumstances and there is no human interaction.
NDTV: At all circumstances, zero human interaction.
Peter Barrett: And I think that is many decades away.
NDTV: Many decades, that is interesting.
Peter Barrett: One step down, the vehicle almost always drives itself at level 4. And I think the challenge there is, you can make a car that takes over when humans make a mistake. Very difficult to make a car that makes the human take over when it makes a mistake. And I think we have seen recent tragic examples of that, I don't think you can do that. Now we have had a robot in vehicles for many years which is really, really good at helping humans to understand circumstances. So, analog braking is a robot that is really, really good at applying brakes, far better than any human can. I mean it does it by modulating the brake.
NDTV: So, he is off you know, you're not taking any cookies, nothing.
Peter Barrett: Yes, I guess he is going back to the kitchen for more donuts.
NDTV: So, analog braking is working well.
Peter Barrett: And I think the path to autonomy lies incrementally adding more robots like that, the drive assistance, make sure that cars don't run into each other. But I think there is another general way of approaching the problem, which is make these machines the mind readers that they can be. Let them see around the corner by seeing through each other's eyes and collaborate with the cities that they're driving in, I think that's the way to scale...
NDTV: That is level 5, but we're level 4 now. Level 4 is a little dangerous.
Peter Barrett: There aren't many level 4 yet.
NDTV: But even at level 4, even 1% of the time you need a human to take over. Should that be allowed?
Peter Barrett: So, there is a body folk who thinks that level 4 should be illegal. And I think it's an intractable problem.
NDTV: I thought you thought that there are no problems that are intractable. That's why we are going to get into this.
Peter Barrett: There are always other ways of solving the problem. The promise of autonomy, the problem is freeing people up from mundane tasks. I think it is the implementation, there is always an implementation of this and I'm not sure this action of autonomy is the right way of approaching at it.
NDTV: So, you think, either level 5, you certainly don't have level 4 running around the roads?
Peter Barrett: I think an autonomous system that almost always works, is not an autonomous system at all.
NDTV: An autonomous system that almost always works is not autonomous. I think that is a really good point and I hope everybody hears that, because it is worrying and the distinction is very clear you made. Last question in the context of, what next? Because there is lots of big ideas being cracked every day and new learning as you said.
Peter Barrett: Yeah, I think that within these walls, there are some absolutely revolutionary companies. When they succeed, that will enable new companies and new industries to be formed. I have heard one of the CEOs here say that he believes that the computing revolution has not happened yet. I believe that. I also believe that industrial revolution has not really happened yet either. And the advent of next generation computing technologies and the possibilities for engineering that follow that, really will change the way we work and live, create an extraordinary amount of wealth, create enormous opportunities for investment and really change the way we live.
NDTV: You are saying there is a whole revolution to come both in computing and just in technologies. Just give us a glimpse of what the world is going to be like 15 years from now. How the change, 15 may be too long, I don't know.
Peter Barrett: Hopefully, we are in an environment where we can reverse the damage we have done over the last couple of centuries. To be able to have more and more of wealth population living in fair and equitable environment.
NDTV: In what way?
Peter Barrett: Well, to be able to create wealth, to have high quality, inexpensive food, to be able to have, you know, to be able to grow buildings...
NDTV: High quality, inexpensive food is like huge for a country like India, huge...
Peter Barrett: And then we are in a position to address the limitations of existing agriculture, with things like, understanding how to create a fertilizer without room temperature, without the Haber process for example, to be able to apply practices for automation, to be able to inexpensively address limits in labour to create the food, engineering different kinds of plants that are adaptable to different climates and different environments to the soil.
NDTV: Or productive, cheaper cost. I mean that's a revolution in agriculture in itself.
Peter Barrett: I think there are huge opportunities in technology to address those kinds of things. And that the more people we put in the position to spend time being scientists and engineers, creating this kind of wealth, is a positive feedback that addresses more and more of those problems, right. We are, I think India produces in the order of magnitude more engineers than the US does. And science is, for some reason, under siege in the US. But there is great positive feedback though, with the right capital and right engineers and people being able to concentrate on those kinds of problems, the opportunity to create wealth and make it better for everybody is virtually unlimited. And as I say we are just at the infancy of these new computing technologies, which unlock material science and chemistry and medicine in a fascinating way. So I think most of the good ideas have not been had yet, most of the interesting companies have not been invested yet and we want to make a contribution to that as much as we can.
NDTV: God bless you, and best of luck. Love the world you are looking at. The world we are looking at.
Peter Barrett: It is a delight to be a part of it, thank you.
Prannoy Roy Speaks To Kunal Ghosh, Founder And CEO, Inscopix
NDTV: Well Kunal Ghosh, really good of you to spare your time.
Kunal Ghosh: Likewise, thank you.
NDTV: Kids like you, you make us proud. And you have a lot of work to do and taking time off like that is wonderful.
Kunal Ghosh: Thank you sir.
NDTV: Tell me a little bit about your company. It is really inspiring, what is the name of it, firstly?
Kunal Ghosh: Inscopix.
Prannoy Roy: What's the big picture that it's trying to do.
Kunal Ghosh: So Inscopix is a neurotech company. We are based right out here in Palo Alto, spun out of Stanford several years back. We were founded with the mission to help decode the brain, this black box. We have very little understanding of the brain.
NDTV: The brain is a black box. I love that.
Kunal Ghosh: The brain is a black box and it has been a black box for decades. It is precisely because we know so little of about how it works, let alone how it doesn't work, that we are where we are today with respect to brain diseases and mental illnesses. The statistics are just mind boggling. You know, something like 1 in 10 adults worldwide suffers from mental illness.
NDTV: Wow, wow.
Kunal Ghosh: It is hundreds of millions of folks around the world, whether it is diagnosed or undiagnosed. The global cost is, I think, something like three trillion on society. It's what it is today because we don't understand how the brain works. We don't have good treatments, let alone cures. So our thesis is to say that let us empower researchers around the world to address this knowledge gap. Let us demystify this black box. You know, enlightened researchers with our platform, our analytics.
NDTV: So you are studying, understanding the brain, which is a huge breakthrough. Is that what you are doing?
Kunal Ghosh: I like to say we are empowering the heroes out there, who are going to be making the breakthroughs, with our technology, with our analytics, with our personnel. We certainly are investigating some specific brain functions and some particular processes, but it is the heroes out there at universities, research institutions, pharmaceutical companies that are using our tools, our products and are trying to understand how we encode memories. How do we get addicted to substances? How do we eat? You know, what are the behaviours that result in binge eating that results in obesity, which is a big problem in the US. These are all brain circuit problems. Our platform is, for the very first time, literally shedding light. It is an imaging based device, so it is literally shedding light on the circuits that no one understood, how they gave rise to basic functions that we take for granted. In the case of diseases, like Parkinson's, together with drug companies, we're for the first time discovering how circuits are malfunctioning.
NDTV: Circuits in the brain? Wow.
Kunal Ghosh: Brain circuits, how they malfunctioning and giving rise to the symptoms. And the hope here is that if we can understand how the circuits malfunction, we might be able to develop much better treatments that re-tune the circuit, in some ways analogous to controlling blood pressure. If we can control blood pressure, the person can lead a long healthy life, hopefully, not succumb to cardiovascular diseases.
NDTV: So you can do a similar kind of thing with different aspects of the brain?
Kunal Ghosh: Exactly, that's the hope. That once we demystify this black box, we will understand better what to do to help people.
NDTV: So what you are doing could, apart from understanding all these, the binge eating, also could help with Parkinson's and Alzheimer's?
Kunal Ghosh: Absolutely. In fact that's the big opportunity that we have here at Inscopix and you know, for humanity, that if we can start understanding in a disease like Parkinson's - how the Parkinsonian circuit in a brain is malfunctioning, we can distinguish normal from the abnormal, then we might be able to empower drug developers, pharmaceutical companies, device developers to do much better than sticking crude invasive electrodes in the brain. Very crude, sledgehammer like treatments is what we have today. This is the state today.
NDTV: Let me just bring your eyes on this. Inscopix, this is one of your products, isn't it? It looks very well developed actually, already! Explain a bit of it, just for a layperson.
Kunal Ghosh: Sure. So as I was mentioning, the technology actually spawned out of work we did back at Stanford. And this is the core of the platform. It is a miniature, wearable microscope. It is a microscope-based technology, with an integrated HD video camera. In many ways it is similar to the chips in the camera that is currently photographing, filming us. So this device will be head mounted, not on human subjects yet, since right now we are empowering researchers to understand basic circuits that are malfunctioning, and disease models with the hope of then translating later to humans. But researchers will be able to watch, with this device, hundreds of thousands of neurons firing in real time during behaviour. So you can really start studying the brain action and visualise how brain circuits, neuronal processes and patterns give rise to behaviours.
NDTV: Interesting that you say that these are cameras, it helps visualise. So like this would be fitted where, eventually?
Kunal Ghosh: This would be head mounted
NDTV: Like, here or here? Or inside?
Kunal Ghosh: On the cranium. Or it could also be implanted, in a minimally invasive way, on the structure that is being studied. Right now it is for research purposes. So it is on research subjects and is used to understand, as I was saying, how specific neurons are encoding for memories. In the Parkinson's study that I was mentioning, this platform, this technology showed how in a particular animal model of Parkinson's, we knew exactly what was happening in the circuit that gave rise to the behavioural abnormalities. And this is the kicker, that most drug companies today rely only on the behavioural read outs, if you will, that are very crude, very coarse. And then most of the drugs end up failing in humans because the animal behavioural model is in no way a predictor of how well the drug will work in the clinic. With our technology, by looking at the brain circuit, in vivo, we can now show literally how the compound, the drug is working on the circuit.
NDTV: I see.
Kunal Ghosh: And it's way more sophisticated than the behavioural read out. Meaning that we can tell whether a drug will work or not, predict its clinical efficacy based on the circuit biomarker. So it is changing the paradigm in neuroscience, neurology, neuropsychiatry. From looking at the brain just as a collection of chemicals, that is how we have looked at it, inject some more dopamine, or pull back some dopamine, that is how we treat patients, into looking at the brain as a big electrical network. And if we can decode this network, then we can understand how to re-tune the network with precision therapeutics and hopefully improve the quality of life, if not curing some of these diseases in the long term.
NDTV: And when do we think, like say Parkinson's or Alzheimer's could benefit from this?
Kunal Ghosh: Well, in Parkinson's we already are in the process of conducting larger scale research trials, where we are starting to develop AI-driven methods to predict the accuracy or clinical efficacy of a compound on a device. And the hope is that in five to 10 years, some of these results will start translating into therapeutics, that could be used in humans and hopefully do much better than anything that's out there today.
NDTV: Anything about Alzheimer's?
Kunal Ghosh: Alzheimer's, mental illnesses, um, mental illnesses are extremely poorly understood, so at least a disease like Parkinson's, at least there are some therapies in the clinic. You know we have deep brain stimulators, it might be crude, we have L-DOPA, might not be optimal, but at least there's something for early stage patients. You look at depression, schizophrenia, bipolar, there are no good therapies out there. Diagnosis is also very primitive. So this kind of a technology is extremely well suited to look at mental illnesses, understand how the circuits are being disrupted and eventually...
NDTV: So what do you actually see?
Kunal Ghosh: You will see literally hundreds and thousands of neurons that are flashing, like stars in a night sky, the brain lights up and you can see a bunch of neurons that are firing and the patterns of firing result in the function or the behaviour that we take for granted.
NDTV: Wow. That is a huge breakthrough.
Kunal Ghosh: Well that's endoscopics and that's our promise, that if we can start decoding these patterns' effectivity, let us first provide these patterns' effectivity. Let all our researchers, pharmaceutical companies see for the very first time the brain in action. President (Barack) Obama, in the State of the Union, said this, probably more eloquent than I am, and I think in 2013, that this three pound mass between our ears is a mystery. But if there is a way to start unlocking this mass, if there is a way to literally study the brain in action, to visualise the brain in action, then we will be, hopefully, at the next phase of human evolution, where we are finally starting to demystify this organ, conquer brain diseases and just improve the state of the world. Inscopix today is one of the few companies at least, that is literally empowering researchers to do that, to study the brain in action.
NDTV: And you are a part of the Playground?
Kunal Ghosh: We are a part of the Playground. We love the fact that Playground invests in platform companies and that's what Inscopix says, it is a platform for mapping the brain and eventually, of course, for enabling the development of next generation therapeutics.
NDTV: Kunal, you will be successful. You are a Bengali, so there is no question. Good luck and God bless you. You are working in such an important area.
Kunal Ghosh: Thank you. I appreciate it, my pleasure.
