Vinod Khosla sits with Katie Rae, CEO and Managing Partner at MIT’s The Engine, to discuss the firm’s investment thesis of societal reinvention and the future of healthcare.
Katie Rae:
Thank you so much for being here.
Vinod Khosla:
Oh, well, it's always exciting to be talking about tough tech, for me.
Katie Rae:
Okay, awesome.
Vinod Khosla:
Call it fun tech more than tough tech.
Katie Rae:
Fun tech? Yeah.
Vinod Khosla:
It is hard sometimes, as many of you know, but it is a lot of fun to be doing this kind of work instead of spreadsheets.
Katie Rae:
So-
Vinod Khosla:
We have a rule against ever using a spreadsheet to calculate a rate of return. I haven't seen one in the last 10 years in our partnership.
Katie Rae:
Interesting. Is somebody in the back office doing it anyway? No.
Vinod Khosla:
No. Well, after the fact we have to report to LPs. Keeps us in business. The fun part, then, is the tech.
Katie Rae:
Yes. So tell me, what are you excited about today in technology?
Vinod Khosla:
If you look at tech and venture capital, until about 10 years ago, mostly it was about technology, so technology innovation for technology. Enterprise apps, mobile, mobile apps, semiconductor chips. It was all about tech. What's really, really exciting, and what, to me, most people have sort of realized but not quite, is now entrepreneurship and tech startups are about every part of GDP. I'm 64, turn 65 soon, and I decided a few years ago if I was going to work until age 80, I'd work for 20 years. I'm actually hoping it's still 25, five years later. I looked at what parts of US GDP could you not reinvent, and none of this crappy 5% of improvements, but really 2X to 10X improvements either in resource utilization, which is okay, or what you could do, or cost for accessibility, and I was shocked. I was completely shocked looking at every part of nongovernmental GDP. I couldn't come up with one area in which I couldn't come up with a technical approach to have that huge multiplier, and that's really exciting.
So my one message when people say, "Well, some areas are not open for technical innovation", there are no such areas. For God's sake, it was eight, nine years ago we invested in the hamburger, now the Impossible Burger. People say, "Well, that's not tech, that's not innovation” other than it took a few hundred man-years of PhD effort to actually get the flavor right. There's like 40 mass specs sitting there doing analysis. Yes. It's more than most tech startups.
Every area of society is open for large tech. So whenever somebody points to me, says, "This area you can't do much about." I just gave you the example of food, and we've done a number of things in food. You can look at an area like construction. I'm so excited. I think in five years or seven years we'll be talking about completely 3D printed buildings with new materials, with one-fifth the carbon footprint, so 80% lower carbon emissions for construction materials, new polymers, new approaches to replace concrete and cement, and I think we can make those low-carbon, too. We're working on both sides of that, and 3D printing them so they're half the cost and accessible, not expensive.
It really doesn't matter where you want to go, there is so much exciting tech, and of course I just finished reading, how many people have read Dave Sinclair's book on lifespan? One or two hands. Well, highly recommended. I'm now convinced I'll be working in 50 years. He thinks we'll have an anti-aging vaccine within 10 years. But, look, you have to be an optimist.
Katie Rae:
Listen, both my grandmothers already lived to 100, so I'm scared. What would that do, 200?
Vinod Khosla:
Yes.
Katie Rae:
No, actually there are a lot of people really talking about living to 150, 180, and they're serious about it.
Vinod Khosla:
They're not just talking about it, he's talking about the mechanisms, the epigenetic programming through which you will solve the problem. This is really fun stuff. Whether it succeeds or fails, it's worth trying.
Katie Rae:
Okay. So what I love-
Vinod Khosla:
And no, that doesn't lead to a population problem, as he very elegantly argues in his book.
Katie Rae:
Because that runs out even if you live to 150?
Vinod Khosla:
Well, actually today, we don't have replenishment rates that are good enough in developed societies, and most societies are trending that way. If you eliminate death, you add about 150,000 people a day, and that's a fraction of 1% population growth which we need to make up, and the dividends from that are large. The economic dividends from that are large.
Katie Rae:
So I have been accused of being an optimist, but now looking at you I actually believe you are the optimist.
Vinod Khosla:
I'm absolutely the optimist. I always say to all of you, skeptics never did the impossible. Skeptics have never done the impossible. Keep that in mind. It's religion for me. I'd rather try and fail than fail to try. It's just so much more exciting that way. This way I sort of get up every morning, and I got in about two hours ago in a red-eye. I'm excited about every single thing I do.
Katie Rae:
That's awesome. So one of the things that I have loved over the last two-and-a-half years is, we have a company, Commonwealth Fusion, who you are involved with as well, but when I first met them, they said, "Oh, Vinod's been doing a whole bunch of research on us for the last few years to come up with his own opinion on if he believes we'll succeed with our approach." I think the way you looked at that company, the problem was unique, and I wonder if you could share that with the group.
Vinod Khosla:
Sure. First, when I first met Dennis Whyte I said, the fusion's too important a problem to not attempt. By the way, it was the exact same comment I made to Pat Brown when he was starting Impossible Foods. He was one guy, a professor who had never had a job in his life other than research at Stanford, but he had all the passion, and of course, deep technical knowledge. I said to him, too important to not try building a better hamburger and save 30%, 40% of the planet's land area. Think about it. If there's one thing that could change it, you could work on it, wouldn't you? Wouldn't you work on aging? So I told Dennis it was too important to not try, so let's be optimists and say how do we do it, and let's understand the problems.
It took me more than a year to do due diligence. I hired three separate technical teams to do diligence who didn't know about each other. They all did due diligence with Bob Mumgaard and Dennis and each came up with very different risks. In my view, if you ask three experts and all three say, "Yeah, this is not doable", then you have to worry. If they all say it's doable, then you have to worry that too many people will do it. So you have to be in this other space where people doubt it can be done, they say it's possible, some will say it's probable. How you value a venture is really important.
So I got all three teams with very different written analysis. I sent the reports to all of them, and Bob and Dennis had already seen it so we did a nice session. And it came up, there were fundamentally two or three approaches to fusion. One was, you can take a risk on the physics, or you can take a risk on the engineering, and we went through a fairly elaborate process. Before we started, we had figured out at least the knowable risks. Others may still show up, and that was a great process, I loved it. I'm actually looking at ignition fusion now with the same approach.
Katie Rae:
And I think that is one of the things that makes you very different and very important in these big engineering challenges. You're unafraid. It appears that you are unafraid to tackle almost anything.
Vinod Khosla:
Well, of course, everything's tackleable.
Katie Rae:
I agree.
Vinod Khosla:
Wasn't it Carl Sagan who said if a distinguished scientist says something's impossible, they're almost certainly wrong?
Katie Rae:
But I think it's that spirit that is the core of entrepreneurship, right? If you don't meet an entrepreneur who has that same view, you probably should run, because they're not going to tackle the big things. So when you have watched some entrepreneurs go, whether it's Impossible, or many others go through this challenge of taking their bold idea all the way to market and to creating a foundational company, what are some of the risks along the way, and how do we help them?
Vinod Khosla:
Look, it's a minefield out there. If you have a hard technical problem, you think the problem is technical when it isn't. It's just the first step, and you want to take that risk out. But then in the case of fusion, for example, we said, "How do we get investors, even if I believe this is doable, how do we get investors?" And we built a business strategy that said, if all we did was build a 20 Tesla machine and size of the machine goes down by T to the four, Tesla to the fourth power, till you built a 4X Tesla, you had 120th, 256th the size, suddenly your practical fusion reactor could fit on this podium. That felt much more doable, much easier to experiment with. If you made a mistake you could make changes within days. And we took that approach and said, "Let's go pitch to investors. If fusion doesn't work, we have great MRI machines." 20 Tesla magnets are still valuable.
So we sort of built strategies that made it lower risk for other people to put money in, people who wouldn't quite do the technical due diligence, just sort of want to know that it's been done, and that if in failure mode they can get their money back.
Katie Rae:
Right. Right. And then in terms of things that you've seen come all the way to a true commercial success, what are the patterns that you think are positive indicators that things are going well, and when things are going off the rail, what are you seeing?
Vinod Khosla:
I think the single biggest indicator of whether things will go well or not is the team you build. I always say the team you build is the company you build, not the plan you make. The business plan is largely irrelevant. It's really almost never even remotely valuable. It is valuable for one thing and one thing only, because to judge the quality of thinking of an entrepreneur, but I can do that in other ways without a business plan. But it is useful for that, other than as a plan to follow, no, it's not useful. So it depends on the team they assemble, and that becomes the single most important indicator.
The second most important one is how open or closed they are to ideas. And some are very close, they believe they have the answer, and those usually will fail. The probability of success is fairly low. And then there's people who keep an open mind, take all the input, all the critique, and then make their own mind. And then there's people who take all the input and do whatever somebody tells them, and they don't usually work, either. So taking as much input as possible as a culture, discussing it, and then deciding what you want to do for yourself, not based on what somebody told you, those are two big indicators, the team and how you think about the process.
Katie Rae:
Love that. Seems very real to me. Okay. So we sit amongst, I don't know, are there 30 universities in this region, some very important ones from a technology standpoint. If you were the president of one of those universities and you were inspiring the next generation, what would you do?
Vinod Khosla:
Here's what I would say, and I would say this to all of you. The world is a much simpler problem if you consider the problem to solve than it appears, way simpler. Let me tell you what I mean. When I turned 60, I framed my mission the following way, and it's a lot of fun. I said, 700 million people of seven billion people on the planet have a rich lifestyle, rich in energy, rich in transportation, rich in healthcare, rich in building space, rich in education. No matter how you look at it, they have access to cars, rich in transportation. Now if we can have enough resources, and we will come back and talk about the tools available to us to do this. If seven billion people can have the richness of lifestyle that the 700 million have, without destroying the planet in terms of resources, that's pretty good magic.
And I found it was very, very doable. You know, every space there's between a 2X and 10X easy, easy, available to us through these tough techs. Not only that, there's not 1,000 problems to solve. I think if you look at somewhere between 10 to 12 problems that we solve, we take care of so much of the planet. So climate change frustrates me because people keep talking about incremental solutions. There's only two things that matter, good policy and tough tech. Nothing else in between. You know, Sheryl Crow or somebody said you should use one sheet of toilet paper, not two. Those things don't matter, right?
Turn off the lights when you leave the room. 2%, 3% energy savings in a building doesn't matter. One year of GDP growth will make up for that. But if you look at 10 big areas, and I admire Elon Musk for this. Five years ago nobody believed in 2025 or 2030 the fastest growing area would be electric cars, 25% to 50% of cars will be electric. Five years ago not a single person in the auto industry believed it. I'll come back to this point about the auto industry. Elon proved to the world in five years he can go from near-bankruptcy to more valuable than BMW, and now everybody, even the most conservative automakers like VW have declared 25% by 2025.
So he's instigated a change that's very large in a very large area. Pat Brown's done that with plant proteins. Now it doesn't matter whether Pat succeeds or Elon succeeds, the world already believes plant proteins can happen. I couldn't come up with more than a dozen areas like that that would change at least sustainability completely. Construction materials, use far less construction materials for housing. Housing for seven million people is a linear problem. It doesn't get exponentially harder. If you look at fusion we talked about, it's one of the reasons I want to work on fusion, so there's four there. There's transportation. We're well on our way to solving that problem. Public transit driven by AVs should eliminate the need for cars on this planet, might also eliminate most of steel mining and lots of other side industries.
Think about it this way. The car I drove in was probably 3,000 pounds, carrying 300 or 200 pounds of passenger. We need 10 times the weight of steel to carry a person? And we drive it 12,000 miles a year instead of 120? I just defined a factor of 100 in how more efficient. A factor of 100 on how much efficient steel can be for automotive. What makes that happen? Autonomous vehicle technology, and using it for public transit. There's no reason there's public transit we need an Uber. You could reserve streets for public transit, and turns out if you do the modeling, throughput capacity per foot of lane way is maximized for these kinds of approaches, and that's the only, the same roads in Boston could carry five times the capacity, five times or 10 times the capacity. You could run the simulation models for this. So suddenly I changed roads so there's no more than a dozen. I could keep going.
Katie Rae:
Well, let's dive on healthcare. What are you excited about in healthcare, and then I want to get back to policy.
Vinod Khosla:
Okay. Healthcare's really, really exciting. First let me start by saying healthcare's better today than it has ever been, better than it was 10 years ago, 10 years ago was better than 20 years ago, so it's much better than it ever has been. But it is not a science. The treatment you get depends on the physician you get. Now that's silly. Even in Boston, every health system has their own care protocols. The care protocols, there should be one optimal one if I walk in with a torn Achilles tendon or something like that, right? I don't get one care protocol. Every single one thinks they have experts who have different care protocols. It's just patently silly. Every med student starts with the Hippocratic Oath, first do no harm.
I can prove to you it's mathematically incorrect. It's a stupid thing to do. You should never take that oath, because if you could save 1,000 lives and kill 100, you shouldn't do it because of the Hippocratic Oath. But physicians aren't mathematicians. If you had enough time I'd tell some fun stories of objections I've gotten from physicians. One actually argued with me in a talk that the median physician wasn't below average.
Katie Rae:
Stats wasn't required.
Vinod Khosla:
And they don't understand probability at all, which is the fundamental core of disease management. But there's two things that are happening. First, the medicine we practice today is practice and it can prove that people observe through RCT’s and others what has worked and what hasn't worked but they're really small, incremental improvements. I think the practice of medicine can be changed into the science of medicine.
Katie Rae:
And what will it take to get there? Is it AI, is it humans working together? How does that evolve?
Vinod Khosla:
So let me describe it at two levels, right? One of those really important global problems is physicians. So when my son graduated from Stanford at 26... sorry, 20-whatever it was, 24.
Katie Rae:
One or two, yeah.
Vinod Khosla:
He graduated with his graduate degree. I said, "Work on something really, really important." So they're building an AI family care doctor. Turns out none of the medical data, and the license they explained from MGH, and they got all the medical records from Stanford, and the whole decision support system from Vanderbilt, none of it is actually as useful as building it from scratch with AI. Now imagine if you could make, and they've just announced they're introducing free primary care starting in California, but the goal is, if you could offer free primary care for the planet, what more fun thing could you work on? What is more fun than that, right? So I'm glad they assembled a great AI team to do that. Obviously, radiologists, same way. Anything image-based, it's a done deal.
In fact, I love the fact, it was eight years ago I tore my ACL skiing on the second day of my vacation so I had a lot of spare time on my vacation. I took an MRI, took it to three different surgeons, three very different prescriptions. This was 2011, Christmas time. So I wrote a blog called Do We Need Doctors? I did also write a blog the following week because I was still laid up, Do We Need Teachers? But if for the same MRI I get three radically different prescriptions, from do surgery immediately, to do something else, to don't do anything. There's something wrong, really, very wrong. So anything image-based can now be done by an AI better. I got a lot of critique for that blog eight years ago. Four years ago I wrote a 100-page thesis called 20% Doctor Included.
By the way, the first version had 80% Doctor Not Needed, but the doctors objected to that. When I changed it to 20% Included they were so much happier. Great math skills. Honest, really. The first version was called Replacing 80% of Doctor Time, Functions, and 20% Included is better than 80% not needed. And this is a 100-page PDF in which I defined all the detailed steps. But I loved it because earlier this year, the editor of JAMA, the Journal of the American Medical Association said they are no longer accepting papers saying AI's better than humans at reading images of any sort.
Katie Rae:
Really?
Vinod Khosla:
Think about it. It's just, this can happen very easily. Think of the dozen things, each of you should pick one of those, I talked about them in other places, and work on it. By the way, water's one of the hard problems, but if you solve fusion, you'll solve water. I don't think transportation's hard. I don't think healthcare is hard. I don't think, like air conditioning, HVAC is probably one of the harder problems. If you look at the top dozen, other than thermoelectrics and non-cadmium telluride thermoelectrics, it's a hard problem. But back to your question in healthcare, the first phase is AI physicians.
With so many papers coming out on mutations. If you have, take breast cancer, your oncologist, even at MGH let alone in rural Iowa, your community oncologist has not read and heard about the last 1,000 papers and the last 100 mutations. Pretty unlikely. Trivially easy for an AI oncologist to keep track of it. So radiologists, oncologists, so I think the first phase of transforming medicine will be replacing with AI and more data the kinds of things that we as humans do. But what's way more exciting to me is, humans do what they've been doing because we've designed these systems to serve humans. So your blood test, your little Quest sheet has 30 tests on it, or however many. The doctor checks off three.
Now why does he check off three? Because he can't look at more. Now a single mass spec dried blood spot can give you 10,000 biomarker data points. If you gave that to a physician, what would he do? If you look at an ECG. As an undergrad in India at age 17 I worked on thermal printing of ECGs as my project. It worked out pretty well, but have you ever asked what frequency band the ECG should have? It's a historical continuation because you're still eyeballing this. Down in the frequency domain there is data we just don't even see as humans that we should be collecting. We've now proven in a joint venture between one of our companies, AliveCor and Mayo, that you can tell blood potassium level, a chemistry test, just from the ECG, from a $99 ECG you can buy on Amazon.
So when we start collecting all this data, every blood test is 10,000 biomarkers, and every image x-ray is not meant for viewing by humans, so you might do a hyperspectral read of an image, for example. There's so many possibilities, I'm sort of dying to try them all. We'll start to see the next age, which is you diagnose disease differently. I'd say in healthcare my goal, people don't think about it, the idea of the Hippocratic Oath is wrong, but the idea of symptom-based medicine is mostly wrong, because when you have a symptom, you already have a disease most of the time. Now acute conditions like breaking a bone, a little bit different. Getting to the burn unit with burns because your house burned down, very different. But mostly symptom-based medicine should disappear. Alzheimer's disease starts 20, 30, 40 years before its symptoms show up. If you wait for the symptoms, it's too late. Diabetes, it should be 20 years. They told me when I was 40 I'd be diabetic within five years. But the symptom-based medicine can be eliminated with the kinds of fine biomarker analysis.
By the way, there's really, really exciting possibilities in cross-disciplinary functions. So, how many bio people here? Quite a few. So you know anti-TNF inhibitors like Humira, very large drugs. That drug itself is probably, I don't know, $15 billion dollars. Half of it is wasted because it doesn't work. Now there's no way, at least in biology we found a way to determine if it'll work on a patient or not. Turns out Barabasi at Northeastern working in network theory, not biology, has figured out the test. We now have a startup here called Scipher that uses network theory, which I first ran into when I did my sabbatical at the Santa Fe Institute in complex systems theory, can determine a simple blood test, save $50,000 a year of a drug that's not going to help a patient and costs the healthcare system a lot.
So across disciplines there's so much room, and most of the innovations, especially in medicine, will come from across disciplines, not just, I don't think an oncologist will solve oncology or cancer. Almost certainly a physicist will play a larger role, most likely, probabilistically speaking.
Katie Rae:
Okay, so I want to dive into two areas. One is policy, and you mentioned it, I think, policy can drive these industries to do the right thing and adopt the right things. But I'm curious how you frame policy, and what we can do with that in order to bring some of these kinds of top ideas or top innovations to the world.
Vinod Khosla:
Well, sustainability is being held back because people never believe policy would force them to reduce carbon footprints. We know we need that driver to increase the amount of R&D for very hard tech projects. But it applies in every area. Even today, in all of the United States except nine states, the American Medical Association has successfully passed legislation to prevent a registered nurse practitioner with four years of medical training from prescribing a simple Z-Pak. Why? Because it costs them a $300 office visit. It's sad, but that's what the AMA is doing. The dentists have passed a regulation, you can't get a dental hygienist to clean your teeth except under the purview of a dentist. Well, we wouldn't need dentists if cleaning was easy.
Katie Rae:
Maybe acutely.
Vinod Khosla:
Or at least 80% of them. But sadly, that's legislation in many, many states. So embedded interests, when we talk about healthcare reduction, the healthcare industry sees as healthcare revenue reduction. So politics and policy, in most areas, play a big role.
Katie Rae:
So we live in a global world. That seems redundant to say, but we're competing, just take China for instance, with a much more top-down, policy-driven innovation economy. How should we think about that? How should we frame that in our minds about the world that we live in, and how do we compete?
Vinod Khosla:
I would say the following. I think it's a real concern, but top-down, policy-driven approaches have their disadvantages. If they didn't, Russia as a planned economy, as a communist economy would be better than the US capitalistic, chaotic economy. There's value to chaos as a system-level behavior. I think it's something we should worry about. I think embedded interest is what I most worry about slowing innovation. You know, the data pane, for example, very good reasons to worry about it, but it was mostly instigated by traditional media in Europe because they were seeing their business disappear, and there were no tech giants in Europe. That's not to say they didn't have good causes to get behind. I actually invited the lady who's sort of well-known, gave a TED Talk on how the 1% work change changed Brexit.
I invited her to our CEO Summit, and I asked her, "You know, what about, if you took Google away, what would you miss?" And she said, "Not much." And I said, "I would have no access to information sitting in India as a student, like most students do today. Have you thought about it?" So, look, any technology, there's good uses and bad uses, and we as a society have to play Whack-a-Mole with that. We have to keep improving our laws and be clear about what's permitted or not, and the law-making process, policy-making process will be subject to influence by self-interested parties. I mean, the taxi industries tried to stop Uber. You see it at both micro and macro levels.
Katie Rae:
Okay. So one of the things that we're passionate about is bringing more potential funders into tough tech and helping people understand the different risk levels and how they should dig into the different risks that they're interested in, right? When you think about what needs to surround companies and ideas like this, the capital that needs to surround them, what do you think we need more of, or how should we think about kind of inviting more capital into this sector?
Vinod Khosla:
Well, that's a hard question. Those of you who have done tough tech, hard tech, whatever you want to call it, financing is really, really hard, and in some domains, finding partners is really hard, and I want to finish on the note of partners. In other domains, the biotechnology, the pharmas, all set up to partner. In the energy industry, nobody will partner because they don't see a threat, so it depends on the industry. The large companies in Google, Facebook, always looking to partner or buy or acquire. In the automotive business, they never do that. So each industry's a little bit different, but the most powerful thing you can do is role models. Food tech was a dead science. With plant proteins and Impossible and Beyond Meat, now it's a hot area. It's not hard getting funding in that area now.
So we have to do what humans love best, which is role model some success, and I think that's really important. Let me just finally say on partners, mostly there are disadvantages to startups other than for funding. The notion of partner expertise is really silly, it's biased. So let me say the following. What is experience? Experience is a bias that's mostly positive if nothing changes. So if you're extrapolating the past, experience is a good bias to have. If you're trying to invent the future, experience is a terrible bias to have. And so, partners bring a lot of experience and traditional bias. There was an autonomous vehicle company and they got like three people from the auto industry on their board, and I literally said to them, "This will fail because of the board you put together. Too much experience and the wrong kind of experience."
But let me end on this sort of hopeful note. I looked at the last 40 years that I've been in tech. Yes, it's been that long. That's all I've ever done. When I looked at the major innovations, I couldn't think of one major innovation that came from an established, large company. Not one. I should say formed company that wasn't run by its founder. No matter where you look, autonomous cars didn't come from the auto industry, electric cars didn't come from the auto industry. SpaceX and Rocket Lab didn't come from Boeing or Airbus or Lockheed. Pharmaceutical industry didn't do biotechnology, Genentech did. Media, it wasn't Fox and NBC, it was Twitter and Facebook and YouTube. Transportation wasn't a transportation company, it was Uber and Lyft. So I couldn't think of a single area, with two small exceptions.
In the early '70s, Bank of America took their credit and put it on a plastic card. That was the best physical example I could come up of what you might marginally call a large enough innovation in the financial industry, because all they do is hack risk. They did CDOs and those kinds of financial instruments, but they're just playing with themselves, not adding any value to society, mostly.
Katie Rae:
Okay, we're going to end there. I will just say, what a pleasure to have you, and thank you for, not just your optimism, but I think there are really big important problems to solve, and you're sitting in a room of people trying to do that.
Vinod Khosla:
Look, we have the ability to solve these problems. I'm absolutely convinced they will be solved, and this is why I mentioned David Sinclair's thing. Seems like an impossible problem. Any expert you ask will most likely, 99% probability give you the answer, "It can't be done." But it can be done. We have to try, and try, and then fail if we must till we try again. But there's a lot to be done. It's very rewarding, too, to work on these problems, and it's very, very fun. Thank you all very much.
Katie Rae:
Thank you so much.