Sridhar Iyengar: The CEO of Elemental Machines on Life After a $260 Million Exit – EP. 170

5:10 – “My parents are of Indian origin, and my co-founder, Sonny, his parents are of Vietnamese origin. And having parents of Asian descent, the advice we got was: ‘Stay in school, don’t skip class, study hard, go work for a big company.’ The absolute worst advice you could be giving somebody at that time, at that university. So we often joke about that, that if we hadn’t listened to our parents, we would have maybe ended up in Silicon Valley and done all that much earlier. But then I was really fascinated by academia. I wanted to be a researcher, I wanted to be a professor, I thought. So I had this wonderful opportunity to go to graduate school and I went off to Cambridge [University, U.K.] and did a PhD in bioengineering, and specifically in biological and chemical sensors. That way I could learn about life sciences and biology and still leverage my engineering double D background. And the work we did there, that was sort of the inspiration and basis for my first startup.”

6:26 – “During the late nineties, when I was out of the U.S. and out of the software and tech sector, in academia, in the UK, in the life sciences, I saw a lot of my peers, my former classmates, all going up to Silicon Valley, doing startups. And I remember this distinctly, watching how difficult, and how laborious it was for my PhD advisor to raise funding through grants, getting a half a million dollars or half a million pound grant took so much effort. And yet I’m looking at folks that I went to college with that are raising 5–10 million bucks.”

7:38 – “I wanted to come back to the U.S., after living in the U.K. for a number of years. I contacted my old roommate, Sonny, who was actually doing a PhD. Program at MI T.…he emails me back within three minutes and says, ‘Hey, I just started a software company. Come join me.’ And I’m sitting here going, I’ve been roommates with you. I’ve known you for many years. So when you say you started a software company, I take it with a grain of salt. So come on, let’s have a talk. So we talked about it and I’m like, ‘All right. Let me see what this is like.’ So we had a gentleman’s agreement, which was if he had raised enough money to pay me for a year, then he would have a year of my time.”

10:12 – “About six–eight months into my first stint at Sonny’s first company, I said, ‘Hey, I think when my year’s up, I’m going to resurrect my PhD work and continue that. I think there’s more to it. And my PhD was in glucose monitoring, glucose sensing healthcare. And, that’s kind of what I did. At nights and weekends, I started resurrecting some of the stuff and doing experiments in the kitchen, and then after my year was up, I tried it full time and soon after that my business partner, Sonny, left his own startup…and he was going to go back to MIT to finish his PhD, at which point I said, ‘Uh uh, you owe me a year of your life.’ He did a little soul searching and said, ‘All right, let’s do it.’ And so that’s how our first company together, AgaMatrix, was founded in 2001.”

13:41 – “I do think AI is overhyped but it’s overhyped because I think it’s also slightly misunderstood because a lot of the value in the AI approach is not fancy AI algorithms and techniques – those are fairly open source and off the shelf – the value comes in, in what is a training data that you have? And do you have access to data that nobody else does that gives you competitive advantage? And as you progress your platform or your product, are you gathering more unique data that makes your training model that much more competitive and special? So the focus should really be on what special unique data do you have? Not that you’re an AI company.”

15:06 – (Ross) “We’re coming into a legal minefield with that stuff [AI] because people are saying, ‘Hey, we can make these images.’ And somebody is saying, ‘Yeah, but it’s all trained on Getty images or a protected source,’ or it’s based on, ‘Hey, my artwork is showing up in an AI rendering.’ So that seems to be something that’s only going to increase as we start solidifying the law around the input source of these AI models.”
(Sridhar) “Yeah, well, legislation always lags technology. You see that with gig workers like Uber and DoorDash and all this, and the legislation around protecting gig workers is what, a decade behind the innovation? So that’s not surprising. Legislation should come in but what’s going to end up happening, I think, is that that legislation is going to increase the cost burden for doing these types of things and that’s going to put certain people out of the market and give others a competitive edge because they’re able to navigate around that.”

19:33 – (Ross) “We’re at a point where all of these advancements are just going to be far beyond the comprehension because ‘Hey, here’s a black box. We put a whole bunch of input and then this stuff gets spit out and it’s either useful or harmful. But nobody really knows what’s going on in the black box.’”
(Sridhar) “It’s interesting that has very big implications for healthcare because if you’re going to use AI techniques, or AI tools, to diagnose a patient or prescribe a therapy, then how do you validate that that’s correct? The only way to really do that is just to do loads and loads and loads of clinical trials to get that through. A) that’s extraordinarily expensive, but also here’s the interesting thing – the more data you feed into a model, with every new training set, the model changes. So if you came up with a diagnostic device or software or a therapeutic, if you ran that on a model that was done when the FDA cleared your product versus a new model today, those results could be different. Now, they could be better, but then that means you have another regulatory burden to go through. So, the FDA is starting to get their head around how to regulate that. I think they’ll get there. It’s just going to take some time, but there’s huge regulatory and legal implications for using tools that humans don’t fully understand yet.”

21:05 – (Ross) “I heard a study not too long ago that AI was trained on x-ray data, or I think it was x-rays, and the upshot was from an extremely narrow section of an x-ray, AI could tell whether the patient was male or female and no doctor understood why or what it was. It was identifying something, and it was correct, but nobody understood what that difference was. So the decisions that it’s making, we don’t know what it’s deciding.”
(Sridhar) “It’s picking up information and data that is not perceptible to the human, might be a point here, a point here, and a point here.”
(Ross) “It’s a ratio, right.”
(Sridhar) “Yeah. And it’s connecting three dots that we as a human don’t see the connection to because they’re very disparate. But if those three dots appear all the time, consistently, then it picks that up as a signature. And so therein is the flaw that if you feed it with biased input training data, then you’re going to get biased output. And that’s one of the big issues with a lot of these telemedicine apps or even skin cancer apps, for example is, are you training it on the right skin tones? Are you getting the full diversity of inputs? And the answer is not yet because that’s costly. It takes money. It takes time. So eventually the industry will get there, but those are some of the very valid concerns.”
(Ross) “And there was also that story off…they were trying to detect skin cancer or something like that – and every single picture that had skin cancer had a ruler in it because they were measuring a dot. And so AI learned that if you see a ruler, it’s skin cancer, because the ones that didn’t, didn’t have the ruler or something like that.”
(Sridhar) “Sounds about right.”

23:49 – “Don’t take money until you need it. And if you can build a business without taking outside capital then do that. And the reason is, as soon as you take in outside capital, you have other stakeholders that you need to be responsible to. And convincing them and telling them your vision becomes a large part of your role. And when you have the right investors that share that, then that’s great. If you don’t have the right investors…it occupies a large part of your brain. But fundamentally, not all businesses need outside capital.”

28:49 – “I think to me, the biggest fear is regret. And that was just something that was instilled into me by my parents at a very, very early age. They came from India to the US. And they always had a very adventurous spirit. You have to, to make a journey like that, especially back in the day. And they were very much into travel. We traveled a ton. They exposed me to all walks of people, all walks of life, all different cultures and atmospheres and all that. And my dad’s biggest philosophy was just, ‘Give it a try. Just go do it. Don’t ever say you can’t. If you don’t want to, that’s fine. But don’t ever say you don’t think you can. Just try it. What’s the worst that could happen?’ And so I had this feeling instilled into me that I would regret not trying far more than trying and failing.”

42:33 – “At my last company, Misfit, we actually had three founders. So myself, Sonny, who’s always been my co-founder, and then the third co-founder was John Scully, former CEO of Apple…And I’m pretty sure John won’t even remember this conversation because it was kind of a throwaway conversation, but he said something that fundamentally changed how I thought about building a company, and was the foundation for how we built Elemental Machines…I asked him, so, ‘How do you make that proverbial dent in the universe and how do you really impact an industry? How do you change an entire industry?’…And what he said was, ‘Look, if you really want to impact an industry, like an entire industry, you need to build infrastructure…you need to build infrastructure that entire industry can be built upon.’ And of course, the example was Apple, iTunes, iOS, iPhone, obviously, but also Facebook – it’s infrastructure if you think about it, but then what he said was but the problem with infrastructure is nobody buys infrastructure, it’s very, very hard to sell infrastructure. You become a consultant, a contractor…but people buy products. So what you need to do is to build a product, to build a business around it, but when no one’s looking, build infrastructure underneath it. Or in other words, as you build, build product on top of your own infrastructure. And if you look at what iPhone did, it is infrastructure.”