Uncapped #15 | Vinod Khosla from Khosla Ventures

[00:00] We're going to be in an era of abundance that's so large, it's very hard for people to imagine. The simplest way to say it is the need to work will go away. People will work on things because they want to work on things, not because they need to work on things to pay their mortgage. All right. I am super excited to be here with Vinod. Thank you so much for doing this with me today. I really am looking forward to this conversation. Look forward to it. [00:26] When you first walked in here, before we got started, you were saying, well, you know, the world's crazy and you've invested through many periods of time. Obviously, you know, there have been a lot of ups and downs, but you said the world's particularly crazy now. Can you share why you said that or why you're feeling that way? I've been doing venture capital for 40 years, so I've gone through every large innovation cycle. Yeah. [00:52] I've never seen a cycle like this to... [00:56] Put an order of magnitude on it. Almost everything. [01:01] Every job. [01:04] is being reinvented. [01:06] Every material thing is being reinvented differently with AI as a driver. Probably the best way to describe it, if we look 15 years hence, and that's too far to look, we'd have to look back at least 50 years, maybe in the 1960s, to see the delta in change.

[01:36] time. Uh, [01:39] It's almost hard to imagine how society adjusts. [01:44] And if I go even further, [01:47] almost certainly within the next five years. [01:50] Any economically valuable job humans can do, [01:54] AI will be able to do 80% of it, with a few exceptions like [01:59] heart surgery or brain surgery, but by and large, 80% of all jobs can be done by an AI [02:09] And if that's happening in the next five years, [02:14] The world is going to be a very different place. So I see this sort of like invention happening at a frenetic pace that is so large compared to even the dot-com phase when the Internet first came and – [02:31] There was Netscape and TCP/IP and all that. It's almost hard to imagine. No matter where you look, [02:40] Everything's being reinvented in some fundamental way using often the innovations from AI and the capability from AI. [02:53] But also a few other vectors, like stuff happening in biology, stuff happening in fusion and other areas. So everything's up for graph. So I guess that's then leading to a frenzied pace of investment.

[03:12] companies being created, you know, new ideas getting tried all at once. Is your overall read based on, you know, on one hand, it's crazy and frenetic. On the other hand, everything you just said about, yeah, but like maybe a billion jobs are going to be, you know, done by an AI in the next... [03:27] couple years or several years or something like that. Do you net out to being very optimistic about this moment in time for tech? Or do you think there's going to be a lot of carnage? Or how do you think? So my view... [03:40] the next five years to 2030, we look at, [03:44] like productivity improvements, the way economists would say productivity is going up, GDP is going up, accelerating on GDP, things like that. And then if I look to the 15 years hence, 2040 and beyond, we're going to be in an era of abundance that's so large, it's very hard for people to imagine. [04:14] People will work on things because they want to work on things, not because they need to work on things to pay their mortgage. Is that going to require us to change anything about like the social contracts to get there? Yes, I'll come back to that. What I want to say is somewhere starting in 2030, the changes are going to be so disruptive. [04:38] that it's hard to imagine how it sorts out. And it'll sort out differently by country, by region. Different governments will play different roles in allowing it to happen or not allowing it to happen. And so we'll see this very, very disruptive, almost dystopic type of crisis.

[05:00] AI is taking all the jobs kind of thing. But we will have enough productivity, enough goods and services produced to share broadly. Is what you're saying that basically the changes that are already happening will compound to a point where it'll now just be this big disruption? Or are you saying the tech by the year 2030 is then going to be so disruptive that it's going to create all these changes? In other words, are we already on that trajectory or you're [05:27] just saying the tech's going to get there in a way that will be that disruptive? What I'm saying is tech is disruptive in some areas. And it will be everywhere. And it'll be everywhere. But... [05:38] Almost no cooperation. [05:42] Not. [05:42] by 2040. [05:44] These are Fortune 500 companies I'm talking about should run the way it's running, even remotely. I should say most Fortune 500. Yeah. One of my predictions is the 2030s will see a faster rate of demise of Fortune 500 companies than we've ever seen. It's actually pretty high anyway, but we'll see a rate accelerate dramatically. So let me put it differently. [06:14] If you're a healthcare company, and I speak to a lot of healthcare company boards, you look at AI and say it's an opportunity and we'll... [06:23] get more administrative efficiency or something like that. But if I said to you, which I've said to a couple of boards, if all medical expertise is free,

[06:35] You have an unlimited number of primary care doctors, oncologists, gastroenterologists, mental health therapists, you name it. How would you redesign the healthcare system? [06:48] Be crazy. That transition won't happen from existing companies. [06:54] Almost somebody new will reinvent this. Now, there's the issue of regulation. The American Medical Association doesn't like that. I was going to say, there will be a lot of people fighting against this who have a lot of power, who will be good at fighting it for at least some amount of time. For some amount of time. Yeah. But fortunately, we have a small part of our health care system, very small part. [07:16] where [07:17] So payments are capitated, what's called the capitated system or Medicare Advantage, where you get paid a fixed amount to take care of a patient, roughly. That would... [07:28] will be so much more effective because they won't care about reducing costs, which AI can do. Reducing costs for a hospital system is reducing their revenue. Of course. For a physician, reducing costs is cutting their costs down. Now- [07:47] I think that the general model, I think, for the next five years when I talked about productivity improvements, [07:56] you will see... [07:58] Almost every professional [08:01] start to get interns working for you. So every...

[08:06] MD, a practicing physician, will get five fresh MDs as if they are fresh out of Stanford Medical School that will work for them and increases productivity, maybe provide a higher level of service to their patients, all that. But [08:24] That then turns into something much more catastrophic over time. So by and large, next five years, I think we should assume that all professionals start to get interns working for them who are trained in their profession, whether you're a structural engineer or a professional. [08:42] or a salesperson who worked for you [08:46] Good, nice, thin turns behave well. [08:49] But at some point, these interns grow up. They will have way more expertise than the physician. And then it's hard to roll this clock back. It's going to go fast, too. It will go fast, except in some areas where there's regulatory or other contract. Take the example of the Screen Actors Guild last year. [09:11] Anybody who signed up for the Screen Actors Guild thing will go out of business. [09:18] But that change will be slower outside that system. People will be producing movies for $100,000, not $100 million. Yeah, right. [09:32] The dynamics will be different by sector. If this does play out

[09:37] even directionally the way you're saying, and now it's 2040, what do you think is left for people to be doing? Like, there must be some things left. Like... [09:49] Are people going to be [09:51] Still all... [09:52] employed or employable via creating their own businesses and pointing AI around? Or, you know, we have a lot of two-person billion-dollar companies. Does it go that way? Or does it actually just go to like a lot of people are just going to consume a lot of leisure? I'm waiting for the first entrepreneur to call me and say, we can build a billion-dollar revenue company with 10 employees. You can definitely do it with 100 employees. We haven't gotten there yet, [10:22] already being started. I mean, somebody will leverage it tenfold again. There's a question of [10:30] Will there be jobs that need people? [10:33] If you have a billion bipedal robots in the 2040 timeframe, could be five years later, could be five years earlier. But let's say that that was one of my dozen predictions I made when I spoke at TED last year. Almost certainly they're doing more work than human beings do labor today. Yeah. And they'll be smarter. [10:57] And they won't sleep. [10:59] They'll never complain. And you won't have silly notions like a primary care doctor and oncologist. Right. The same robot that like cleans your house is also your doctor. Yeah. Yeah. It's this super intelligence idea. I think people will have lots to do, but it won't be because they need to work for a living. The first thing people forget.

[11:22] Most of the jobs in this country [11:25] or on this planet. [11:26] are not really jobs. There's what I call servitude. [11:30] If you're a farm worker working for eight hours a day picking ladders, [11:36] And you do that for 40 years. [11:39] or your assembly line worker at GM mounting a tire on a car for eight hours a day for 40 years. Yeah. [11:49] Those aren't jobs. They are servitude because you have to survive. And what a great thing if we could give those people their lives back and those hours of their lives back. What will people do? Humans will be humans. They'll still compete. We love sports. We love entertainment. [12:09] being great at something. Whether you're even a mediocre artist loves painting for themselves, not because it can be recognized by a New York auction house, art auction house. So people will excel in what they do. You know, exporters, skiing, you name it. I love the idea. People have a lot more time to talk, care about their children. Yeah. [12:39] Actually, not be frustrated because children are in conflict with work requirements or career requirements, working a lot more with their elders, parents. [12:51] So there is going to be a lot to do. What's your confidence level that this is the future?

[12:57] High. 80%? [13:00] Hiya. [13:01] I have very high confidence this world happens. [13:06] I think greater than 50% that it happens by 2050. The timing is hard to predict. But like you're 100% sure that happens before like 2200. Oh, no question about it. Yeah. So I'd be shocked if it didn't happen by 2060, where we live in a world of abundance. [13:27] Nobody has to prove they can buy a more expensive car than somebody else. That part, it may still be there. [13:36] including ego. But I do think something fundamental will change. A five-year-old kid going to kindergarten today is being told by their parents, go to school, study hard, get into good college, and then you'll get a good job. That won't be the paradigm. So my kid, my oldest kid's five, and then I have two other younger ones. If you're me, like, what are you thinking right now in terms of like, what's important for like little kids [14:06] when [14:07] I'm in a position where I can do anything I want. I could go play golf, which I don't love, or go sailing, or you pick your favorite. [14:19] I'm mostly driven by curiosity. [14:22] I wanna learn new things. So anytime I can learn something, I'll take a meeting.

[14:28] and I'm fortunate people will teach me but in the future you won't need somebody to be willing to teach you it'll happen. For me it's that. For others it'll be producing the best music they can relative to their own bar. Not [14:46] And clearly people will still want to be like Taylor Swift. Will there be like the last person? One of the places my head was going was somebody still has to. [14:58] build the AI, build the robots, program everything. Will we get to the last person who needed to do that at some point? And then that's actually even that job goes away. First, high degree of... [15:09] uncertainty with respect to the configuration of what people are doing. [15:16] Thank you. [15:17] I don't believe they'll be working 40 years a week for meaning in their life. Yeah. [15:24] unless it's something they really want to do independent of earning a living. Okay. So that we should be clear. And I just recently gave a talk last month on AI dystopia or utopia. And I, by the way, cover all the dystopian scenarios in it first. Oh, God. Before I get to the utopian scenarios. And so it's on our website and it's on the internet. So anybody can find it. What's very clear to me. [15:55] is the dystopian views will be a choice society makes. The utopian view is going to happen anyway. The world of abundance, only a matter of when, not if, and within our lifetimes. It's amazing. And when I say my lifetime, I at least look at 25 years. I'm 70. Easy. More. When you're talking about this dystopia, I think what was implied in what you just said

[16:25] self-inflicted dystopia where we just, you know, humans don't decide to organize correctly to take advantage of this amazing abundance? Absolutely. Do you worry at all about like the AI doomerism version of dystopia at all, where like the AI goes awry? Is that something you think about at all? Yeah. So let me answer the question in two parts. Certain societies will choose not to do it. We [16:55] traditional studios up to a paradigm that makes them order of magnitude less effective in content generation. Because [17:04] We're building an AI videographer. [17:07] In our portfolio, you specify, you don't just generate video, you specify what the lighting angle is and how you pan and zoom the camera and those kinds of things. So the people who... [17:23] Resort to traditional means will not do well. [17:27] people who do do movie production. Today, we can't do a full movie that way, but that's only a matter of time. [17:35] Only a matter of time. Yeah. Today, almost all commercials can be done by an AI. Yeah. And the example you gave of, you know, reproducing a $100 million budget thing with $100,000, of course that does, you know, of course that's going to go that way. And we know Coca-Cola got a lot of flack last summer, last Christmas, because they produced some commercials with only with AI. And people were nitpicking anomalies and stuff.

[18:05] Most people didn't know it was done by an AI. And they did great. And I'm glad they did it. Broke the ice. So there will be a flood of that. So there's that. [18:19] version of dystopia, displacement. Disruption is always fun for the disruptor, very not fun for the disrupted. And so one has to keep that in mind. We have to take care of people who are disrupted in this paradigm. The other end is [18:41] is sentient AI that kills us all. Exactly. I wouldn't say I don't worry about it. I do worry about it. [18:49] But I look at it as a complete basket of risks humanity faces. We've gone through an era where an asteroid hit planet Earth and most life was destroyed. So we know that was a risk and risks still exist. An asteroid could hit the planet. [19:12] More likely a much worse epidemic than pandemic than COVID. [19:20] You know, COVID only killed 7 million people. It's possible it could kill 700 million people. Totally. Right. In fact, I would say that's almost likely. [19:34] So there's a whole basket of risks.

[19:37] Do I worry about sentient AI killing everybody on the planet? [19:42] Yes, but more than I worry about President Xi doing nasty things? No. Yeah. So when it comes to AI... [19:52] I want us to have the Western world to have the best AI so that we are not subject to President Xi or President Putin. [20:01] that is by far the greatest risk. That makes total sense. And this is what the people who call [20:08] to slow down AI research, miss. [20:11] Putin isn't slowing down, not that he's very good at it, but he's [20:16] Chi in China definitely can... [20:20] do this. You know, it's a world in which you can't argue with it. They sort of say the end justifies the means. We in the Western world care more about the means. Marc Andreessen made this point on the podcast too, which was basically so much of culture and production is going to live in how the AI's [20:42] you know, programmed. And if, if it's our doctors and our teachers and our lawyers and everything like that, you really want it to be. [20:49] Well, we've seen it with TikTok. [20:51] We did. Right. If you're on TikTok, you're more likely to be China inclined, favorably inclined towards China. That's culture. And President Xi had said in one of his speeches not very long ago,

[21:08] The next battle... [21:10] will be fought on the internet, [21:12] It'll be a battle for people's minds. He was clearly thinking TikTok. So there is, you know, [21:21] dystopic AI killing humans and all that, or AI in warfare. [21:27] weapons and robot soldiers and all that. That seems very important too. Yeah, that's very, very important. But... [21:36] It's even more important that we win the race to have social influence. If I'm giving all the entertainment to the rest of the world, if I'm giving all the free doctors, if I'm giving all the few... [21:51] tutors, my philosophy, political philosophy prevails. So I do think by 2040, the biggest risk we might face that I worry about is China using both good AI, cyber AI, warfare AI, but also socially good AI, like free doctors to everybody on the planet and to embed their political [22:21] It's probably the single biggest thing I worry about. In some ways, it seems like politics, both sort of the way we think about it normally, but also just sort of broadly socially, is going to get even more important than it is right now over time. Yeah. Look, we live in a capitalist system, but people forget capitalism is mostly by permission of democracy. Yes. In our system.

[22:51] Vivo. [22:52] adjusts easier. [22:54] Now, China has the advantage of using what I call Tiananmen Square techniques to enforce whatever they want. [23:04] So, it's going to be an interesting 15 years. And, you know, it's going to be interesting, but also critically important 15 years to what philosophy starts to dominate the planet. I want to spend most of our time continuing to talk about the future, but I have one historical event I want to ask you about. The tee up to this question is going to sound like, you know, just an opportunity to... [23:28] talk about being a great investor. But I specifically want to ask you about your OpenAI investment. And the angle I want to ask you about it from is to try to figure out what, if anything, we can learn from how you did that and how you decided to do it. I know that you were sort of like the first venture investor to do it. And you did it in real size and something that probably at the time, a lot of people didn't. [23:52] didn't want to do, maybe thought it was a science project. There was a lot of [23:56] lack of clarity on the whole thing. I think that now it is what it is, but at the time, it was probably an extremely hard bet to make. And so what I was hoping was to try to, as much as possible, play back the tape inside your brain to see what we can learn from that. So it was a very large investment, more than twice the largest initial investment I'd made

[24:26] But I want to go back to a different aspect. [24:30] era and talk about [24:32] this idea of people like to be in herds. The herd mentality was crypto and a couple of other areas. And by the way, I've never called myself an investor. In 40 years, I always say I'm a venture assistant, helping entrepreneurs. It's a much better mental model for me. [24:54] Not the investing is a side effect, so I can keep doing it. [25:02] while people follow herd instinct. [25:07] I like to think about what are the fundamentals and can they happen? Fundamentals don't always happen, but they can. And in 1996, I started a company called Juniper. We worked on a business plan in my office for six months with the founder, Pradeep Sindhu. Why? Because the world was saying the Internet would be a protocol called ATM, asynchronous transfer mode. [25:37] Every major company. [25:39] had given up on TCP/IP for the public internet [25:43] And we're only going to use ATM because the telcos were specifying ATM. I talked to the senior management of almost every telco and ATMs. [25:55] Every single one said to me they wouldn't use TCP IP. I talked to the CTO of Cisco, who said they have no plans to ever do TCP IP above 155 megabits, which is my home service. It's far below my home service for a public router.

[26:17] But I looked at the data. TCP IP usage had been climbing, [26:23] And I said, this will scale. [26:25] and [26:27] Every expert is wrong in the field. We should come back to that if you remember this notion of experts. And the data said TCPIP wasn't exponential and we were near the knee of the curve. [26:40] so [26:42] Fast forward to today, and by the way, I'd invested in TCP/IP, and it was a core part of what Sun did in 1982. [26:52] And this was 14 years later. You see the exponential in TCP IP as a communications medium and [27:00] We bet heavily. By the way, we made $7 billion on a $3 million investment when almost no venture investment ever made even a billion. That's crazy. It sounds good. 2,500 extra money. Yeah, I'd like one of those. Yeah, you only need one. Hopefully, OpenAI for us will do better to go back to your question. But fast forward, in the year 2000, I gave an interview which is in the New York Times, so it's on the record. [27:30] That said, AI will have us redefine what it means to be human, which is this utopia thing we were just talking about. Will people need jobs? [27:41] That was the year 2000 or 2001. 2012, I wrote. [27:46] Two blogs.

[27:48] One was called Do We Need Doctors? Under the premise, AI will do... [27:53] all the doctoring expertise that we need. This is 2012. 2012. It's in TechCrunch. You can look up. And the second was called, Do We Need Teachers? Idea that all tutoring will be done by an AI too. So those were 2012. What gave you confidence at that point that that was going to happen? [28:14] Because this was before the breakthroughs was transformed, way before all that stuff. So, and this is where I was getting to. [28:21] What I looked at was the best talent was going into AI. [28:26] You went to MIT or [28:28] Stanford or Caltech or wherever, people wanted to work on AI. Toronto, of course, a pretty important place in this history. Best talent was going. [28:41] In 2016, I gave a talk at the National Bureau of Economic Research. I should probably release those slides. My first slide was the top 20 occupations in the U.S. by a number of people. And I said, I can't see which ones of these can't be displaced. Now, this was to a bunch of economists. That was 2016. I was convinced. But in that talk, I had... [29:08] human performance and current AI performance. And most people were making fun of AI, because you tell and ask an 83 year old man if they were pregnant,

[29:22] It made silly mistakes. I looked at the rate of progress. [29:27] and the rate of talent influx. [29:29] Those two things convinced me the moment for AI was soon. Now, this was 2016. [29:37] So by 2018, the moment for AI was soon. [29:39] when [29:40] Elon Musk [29:42] backed off funding OpenAI.com. [29:45] He'd committed a lot of money that he didn't meet his commitments for, but [29:50] And when Sam called me, I looked at that and said, [29:53] You have the right team. [29:55] I was very worried about China. I thought there was only two other teams. One was inside Google, which was moving very slowly, and another in Baidu, which was stealing Google's people by setting up an office down here and stealing their expertise. I'm very hawkish on China. And OpenAI needed to exist. [30:15] as a separate vector of AI progress. [30:18] And it had the right team with Sam running it when Sam called me. It was almost a no-brainer. Because you were such a prepared mind for this by that time. Yeah. I was convinced... [30:31] the impact would be large. Now, I had no idea when there would be breakthroughs, [30:37] Obviously, I couldn't [30:39] predict transformer models, even though the paper was out, was going to be that large an impact. [30:46] In fact, [30:47] Transformers wasn't what OpenAI was working on. So how was this technique works? It was more like talent and the rate of progress. And I actually thought it'd be multiple techniques. I remember talking to Sam as I have a Lego block model. The more different types of Lego blocks you do, the more the commentatorial possibilities explode.

[31:13] I think it was following the data and not the herd. So was there a couple year period where you were basically very primed that, [31:22] you were just waiting for the right team at some point. [31:24] Like, it seems like between 2016 and 2018, you probably believed this pitch from just, you just needed the right team. Well, I was convinced... [31:33] of the following. [31:35] AI would have a huge impact because the rate of progress I was seeing on these benchmarks against human performance, even though they were way down there and almost silly performance. But the rate of change was high and the rate of talent and flux was high. The long backing off funding OpenAI was fortunate. I can't say I engineered it. I wish I had. But it is following the data and not the herd instinct. Yeah. [32:05] And frankly, 2018, when we made the decision, it closed in early 2019, uh, [32:12] Everybody thought AI was not a real thing. People thought it was a science project. It was a science project. OpenAI was not using transformer models. It was just my belief. You put the right resources and they'll figure it out. I told my partner, I have no idea whether it happens in technology. [32:31] Three years or 10 years? But... [32:34] The IRR would work out. It was too important to not try was the phrase I used. Some things are too important to not try. And I often try those. At the same time, be invested in fusion at Commonwealth Fusion. Again, I don't know.

[32:51] started it from practically scratch, because Bob was a senior fellow at the MIT Fusion Lab. This idea that some things in society are too important to not try, and I'd rather try and fail than fail to try. And most people, most of the time, just fail to try. And so this combination of factors and knowing AI would have a large impact once we got it, [33:17] was what got me convinced. Okay, so that's a perfect setup into now looking to the future. That construction got you to a big bet in AI. What are the other areas right now that you feel maybe like, you know, the same way you felt in 2015 about AI? What do you feel that way about now? You mentioned fusion, but like what are a few of the areas that you feel this way about the future now? Well, first, related to AI, robotics industry, [33:46] will take... [33:48] A little longer, but I think we'll have the chat GP moment in the next two to three years. What could that look like? A robot that doesn't need to be programmed. It learns. It isn't programmed to do tasks. And it's a generalized, you know, maybe physical robot in our space. Well, the world is organized around the humanoid form factor. And so it's probably a humanoid. Humanoid is the only common enough form factor to get to high enough manufacturing volumes to get the cost down. [34:17] Almost everybody in the 2030s will have a humanoid robot at home.

[34:23] probably start with something narrow, like do your cooking for you. It can chop vegetables, cook food, clean dishes, but stays within the kitchen environment. And, you know, if you have a Prius, it costs three, four hundred dollars a month. The robot will cost three, four hundred dollars a month. It won't matter. And, you know, for anybody making more than a certain amount of [34:53] I think that's how it'll start in the home. But factories, farms, farm workers is a massive area that's mostly unaddressed. What part of the humanoid intelligent robot are we most limited on right now? Like, is it about the robotics? Is it like the intelligence? Is it? It's the intelligence. [35:14] Clearly, we've seen out of China many, many hardware form factors. They're pretty damn amazing. If you see them boxing or running a race, it's pretty cool. [35:27] Cool. But they're not learning robots. You change the environment and they don't do as well. Yeah. I think this intelligence that learns and adapts and doesn't need to be programmed to do a task. [35:42] Yeah, if you walk a human in here and say, clean up, they'll know what to do. A robot needs to do that. Why is this not going to be like Apple? And I know that, I'm not saying I think it is, but why do you think that we don't already have the sort of prototypes of this out from the biggest, most well-equipped hardware companies in the world?

[36:05] So I've been doing innovation for 40 years. [36:08] and innovation only. I can't think of very many large examples [36:16] that we had large innovation [36:19] came from somebody who was large. [36:22] or in the business. Yeah. Can you go through some of them? Yeah. [36:28] Retailing. You know, when we invested in Amazon, I was a client in 96, and [36:34] Nobody thought you could compete with Walmart. You'd think Walmart would in a way. Look at media. Was it... [36:42] NBC or CBS, or was it [36:45] Little things like Netflix and Twitter and YouTube and Facebook, and nobody even knew they were in the media business. Uber, did that come from Hertz or Avis, car rental or taxi companies? No. Airbnb, did it come from Hilton or Hyatt? Yeah. They have more rooms than Hilton and Hyatt now. [37:15] space launches. Yeah. Yeah. You know, my own history, I got enamored with the fact that when I was a Kleiner, they talked about Genentech started by an associate at Kleiner when nobody believed biotechnology was a field. No pharma company was doing it. Yeah. Yeah.

[37:35] Autopartic, yeah. We saw the cloud. We saw self-driving electric cars. [37:40] Everybody had given up on it. How about cloud, actually? Because that was Amazon. So you kind of give them. Yeah. Cloud is interesting. I think I should say founder-driven companies. [37:54] Have permission to try crazy things. Like Apple when Steve was there with the iPhone. Yeah, when Steve was there, they tried the iPhone. Yeah. And if you go back to the press in 2007, everybody said a phone at $600 or $700 that has no keyboard to dial. Right. Like, what a silly idea. Yeah. [38:15] So it's founder-led companies that innovate. [38:19] Self-driving would be 15 years away if Larry and Sergey hadn't said, we're going to do self-driving at Google. [38:29] Or Elon had said, well, I'll do electric vehicles. Yeah. Right. So basically, Tesla should be much more likely on humanoids than Apple, for example. It is founder driven and Elon is paying attention to it. So they're definitely a good candidate. But so are a number of other companies that have a lot of faith in the startup ecosystem of founders with very creative ideas. Yeah. [38:59] But that's an example of a large innovation. So no matter where I look,

[39:06] So the cloud definitely happened. When I was starting Sun, [39:11] Everybody said that these DEC WAX computers and IBM mainframes, why don't you build a graphics terminal for them? I said, no, I'd rather just replace those companies. Neither of those companies are in the computer business anymore. And frankly, it took five years to get DEC out of the computer business into mostly a customer support business after we started. Yeah. Which is pretty stunning to me as a 20-some-year-old kid. Totally. [39:41] So, [39:42] repeating that pattern. [39:45] There's zero chance a company like Siemens or GE could do fusion. [39:51] Right. Zero chance. What's the mechanic for what? I, of course, agree. But what's the mechanic that's stopping them? Is it risk? Is it talent? Here's the first thing I would say. [40:01] In these four years, in all the examples I've just cited for you, and I could give you more, [40:07] None of them were done by somebody who knew the area. [40:12] Experts are terrible at predicting the [40:16] future, they extrapolate the past. Entrepreneurs invent the future they want. It's a bastardization of an Alan Kay quote, but [40:28] So you imagine the world and you try and make it happen. [40:32] and you're not biased by what is experience. Experience is a set of biases that prevent you from making mistakes. And that's what experts do. And that's why

[40:44] Very rarely do experts innovate in their area. Yeah. Look, even the COVID vaccine came from two startups. [40:53] Moderna and the BioNTech. You'd think the pharma was well-equipped to mobilize. [41:02] Not a chance. So [41:04] You know, it's a... [41:07] It's why the US has done so well. People have permission to try. [41:12] And if you try and fail, [41:14] You don't ruin your career. In a big company, you'll ruin your career. Right. To your question. Right. You can't afford to fail. Yeah. So you start with I can't fail, which means you can't take a large risk. My view is very simple. Most people. [41:31] Reduce risk. [41:33] to increase the probability of success. I do the opposite. Start with, I want high consequences of success. I don't care about the probability of failure. I think that is like one of the best things about tech as an ecosystem. I think not everybody thinks quite like what you're saying, but in general, in tech, you can fail and it's okay. And I think even in other, I worked in France and New York for a hot minute. And even there, it's just like, it's just how it works. [42:03] But when we invested in Square in 2010, [42:07] The following was the situation. [42:11] He had just been fired from Twitter. [42:14] Tutu was a tiny company. He'd just been fired. And he had like four failed startups. His history on Wikipedia. Like he had four failed startups. So, you know, traditional wisdom would say, don't back him. Was your mindset around risk always just increase the consequence of success? Or did that evolve over time? It was very much that from, you know, my first startup was go on, take on the computer business. Everybody big is dumb.

[42:43] which is a little bit of an arrogant viewpoint. But if we hadn't had that kind of hubris, [42:52] We probably wouldn't have attempted what we attempted. So I think hubris is true. [42:58] an important ingredient of great entrepreneurship. - Yes. - We'll build an AGI, right? Like that's hubris. It's arrogance, only I can do it. [43:11] So these are entrepreneurial characteristics that are very important, but also, [43:17] thinking from first principles, which is why experts do poorly in their areas, because they have too many biases of what works and what doesn't or how to do things. While somebody new in the area reinvents the area, they figure it out from first principles in the new environment. That's an essential characteristic. So, you know, one thing I always get asked is, what do you look for in entrepreneurs? [43:46] You know, [43:46] I seldom look for deep expertise in their area. [43:51] I do look for people who are thinking from first principles and learning rapidly. Even in YC batch, if I'm talking to a YC partner about some startup, the question that's most important to me is how much have they changed their plan over the last three months? Yeah. Rapid evolution, rapid thinking, rapid learning is what matters. So this went into it. I'm all over the map.

[44:21] getting sort of the other areas you're very excited about. Um, that was robotics, um, which, [44:27] I agree, humongous. In some ways, that's like the final thing to figure out. You mentioned fusion. I'm very bullish about energy. Well, many formats of energy. So, I think there's... [44:40] two large formats. Fusion is one, and there's multiple approaches to fusion. But I'm also very excited about super hot geothermal. [44:53] So, [44:54] Again, this ties to this risk idea. [44:59] Ingeotermal. [45:00] Completely uneconomic business. [45:03] You buy power because it's green, not because it's cheap. And people like cheap stuff. Yep. Turns out almost all geothermal in the U.S. is at 200 degrees or 250 degrees max. If you increase the temperature to 450 degrees, you get 10 times, 6 to 10 times the power, depending upon conditions, from the same well. That's surprising. For two reasons. [45:33] So the energy content of the steam goes up and the efficiency percentage. So you get a multiplier. Don't need to get into technical details, but. [45:44] Now it's cheaper than natural gas if you can get to 450 degrees. Nobody has to worry about super hot geothermal is green. It's just cheap. Wow. Right. And so why don't people attempt it?

[46:00] because people have decided you can't drill at 450 degrees because drill bits collapse. Got it. So I sort of approached it differently. I said, what would it take to drill at 450 degrees? [46:15] If I can solve that problem, everything else takes care of it. And we know there's 450 degrees in most parts of the earth. It's only a matter of how deep you go. And so I set off on that problem. I think we are making great progress. So I think it's as powerful as fusion in supplying most of Western United States. Yeah. [46:38] You can get that. [46:45] watts of power. And I hope this year I can prove we can do all seven gigahertz competitive with natural gas. Wow. Without worrying. It works in the Trump era where he doesn't care to believe in climate change. It still works. And I love that kind of project. I'm also very bullish. Commonwealth Fusion will do well. I don't know enough about Sam's project in Helion, but I'm [47:15] The science is going to work. The impact will be enormous. Yeah. Yeah. It's worth doing. Solar also. Is that on your radar? We are doing solar. We probably have in the Western world, China has some great solar technologies. The next generation of solar called perovskite solar cells. We're working on that. But there, the impact isn't as large. Yeah. Yeah.

[47:39] And that power is already so economic that there's other uses for that power. So we are doing that. But... [47:51] But look. [47:52] I'm relatively optimistic. I see our path clear to getting to zero emissions by 2050. I think most of the technologies do that will be proven in the early 2030s. Is there any sort of countervailing force with the fact that we're going to be consuming a lot more power via AI? People always ask me that. I think if you get fusion right, [48:22] I think very early 2030s, we'll get our first power plants. And that power curve. That power curve just catches up. Yeah. So there'll be this period between now and 2035 when these things aren't scalable yet. Yeah. But beyond that, I don't think we should. Powers will be cheap and plentiful and zero emissions and zero resources. By the way, if you're creative enough. [48:48] You can do this in every area. [48:52] We're producing cement today. [48:55] Thank you. [48:56] We not only don't say like environmentalists do, let's go shut down cement plants every time. There's one in Cupertino that, [49:07] the environmentalists have been trying to shut down. My approach is very simple. They are emitting carbon dioxide. Let's capture the carbon dioxide and turn it into carbonate.

[49:17] We get more cement out of the same amount of limestone materials [49:22] mind at a lower cost per ton. Now, what's there not to like? [49:28] Do you have to care about climate to have that work? No. And we are in production on this now. So I'm sure cement is sold. Steel is the same way. [49:40] I'm pretty certain. [49:42] will be able to produce steel with zero emissions [49:47] or low emissions initially at a lower cost or similar cost to today's scale. I'm pretty damn excited. I think this is mostly solved. So, you know, climate is a very solvable problem, even without AI, and AI will change. [50:10] UNI TÂBĂCEARGI [50:12] climate. So that's energy, generally speaking. Energy, resources, steel, cement. Yep. Robotics we got. Robotics we got. What else? Well, I'm super excited. So I was mentioning this talk I did. It's called Plausible Tomorrows. So you can Google that too. I have a dozen predictions that look [50:36] Really implausible. [50:38] Unless somebody sits down with me and I walk them through each and they say, I see no reason this won't happen. I think by 2050, we can replace most cars in most cities. Self-driving or just don't have cars?

[50:51] Public transit. [50:53] that meets the following requirements. It's always cheaper than cars. [50:57] and cheaper than today's public transit, cheaper than today's public transit, without losses for the transit agency. That's always faster. [51:07] than public transit. [51:09] because it's like Uber, it's hailed not [51:12] scheduled. It's all personal. It's two and four person vehicles, and it fits in a bicycling. Not only [51:23] By the way, public transit, Fusion, OpenAI, I invested in the same business. [51:29] time frame, roughly, around 2018. So I'm very proud of that vintage year. That's a good year. But people said, public transit startup, you must be totally crazy. And I said, Google Swing Waymo is the wrong way to do transit. Because the vehicles are too big or what? No, it just increased congestion. [51:50] Okay. So what should it be? So, and I think Waymo will have a great business for delivering people from transit routes to the last mile. There's only one question in a city. [52:03] when it comes to traffic. [52:06] Can you increase throughput 10x for the same street width without redoing the street width? So we designed a public transit system of self-driving vehicles in bicycle lane width. [52:20] So it's easy to insert in a city, has 10 times the capacity of a light rail system.

[52:27] or car system, so more capacity than light rail system in a bicycle laneway. And it's on demand because the cars just, you hail them like an Uber. So turns out, [52:40] Every single project we have bid so far, [52:43] We bid a lot. Four have been decided. We won every single one for a public transit system from a startup. And the startup wasn't invited to bid on any of these projects because nobody knew about us. Nobody invites a startup. San Jose. [53:03] invited 32 bidders to build a transit system from the airport to the Google campus and then to the Apple campus in San Jose. [53:14] We bid over the transom because most of these bids are open. And we won it outright. Think about it. So what's the vehicle? It's a little pod. Think Disney World. It's like a little narrow car. Yeah. [53:27] Yeah, that makes sense. It fits in a bicycle lane. Yeah. By the way, it can fit above a bicycle lane too. Yeah. So you can go twice as many as cars on a street or something like that. So it turns out it has way more capacity than cars in the street. Or even if everybody took buses, this would have more capacity in passengers per hour for a given street width. Are they also coordinating with each other so the cars are closer together? Is that part of it? Absolutely. You can coordinate. Yes. Yeah.

[53:57] raised infrastructure, [53:59] It's so simple, you can do prefab. Yeah. You don't have to do the Boston tunnel dig for 20 years. Right. Wow. You just do a prefab, put it in place. Yeah. So. That's a good one. It is thinking from first principles. Yeah. That matters. And it leads to, and being ambitious. Okay. And how about medicine? Because that's like 20% or more of GDP. That's a huge one. Yeah. Obviously, AI is a big part of this, but you spend a lot of time in medicine, separate from AI, with AI. What about medicine? [54:29] of, which is a big part of our GDP, a quarter of that spend is medical expertise. [54:37] is pink doctors. [54:40] Almost certainly that goes to zero. [54:44] Can go to zero. So what are we doing? We're building AI primary care doctors, AI physical, mental health therapists, AI physical therapists, AI oncologists, AI, you name it. Actually, maybe just to get specific on what you mean about that going to zero. If we have for every 100 doctors per capita we have today in like 2050, do you think we have like zero or 10 or like what do you think this is? I don't know the answer. [55:13] But there are some. You will always have some to learn from. [55:19] For the AI to learn from. For the AI to learn from, though the AI can boot itself to be much better. [55:26] And I'll give you today's data of a multicenter study at Stanford. They took the best academic institutions and judged the best.

[55:37] Complex AI diagnosis. [55:40] Physicians only. [55:43] And these are academic learning centers with the best physicians, 73% accuracy. So 27% of the patients were getting the wrong diagnosis. That's like a lot. That's like one in, yeah, that's a lot. It's wonderful. [55:58] Think about it. And this complex means it was a serious thing. It wasn't just the flu. AI alone, 88%. Accurate. Accurate. [56:07] But then they gave the AI to the doctors. The doctors improved from 73% to 76%. The AI got degraded from 88% to 76%. That's funny. And this is a serious multicenter study. I love medicine. What will happen is everybody on the planet will [56:26] For less than a dollar a month, we'll get a free primary care, a free physician, initially called primary care, but they'll do everything. My son's company, QRI, is called Multispecialty Primary Care because the AI knows enough gastroenterology to provide that. Now, the hard problem is the American Medical Association won't let the AI write a prescription. Yeah. [56:56] the diagnosis and the prescription or testing recommendation or whatever. Yeah, I was going to say, I mean, if the humanoid at home, [57:05] plays out too, then it's probably also doing some amount. I mean, it's not going to do surgery, but it might be doing some amount of giving you a physical or giving you an injection or whatever else you need. Watch your gate and say, hey. Yeah. Something's up. Yeah. Something's up. Yeah. So that's a quarter of medicine. What about the other three quarters? Oh,

[57:24] Thanks. [57:25] A quarter is pharmaceuticals. Mm-hmm. [57:29] That we are seeing an explosion in AI in drug design. Yes. Both small molecule and... Are you pretty bullish on that? I'm very bullish on that. Do you think it will speed up the rate of drug discovery slash make it cheaper? Or do you think it's more about it will help us discover undiscoverable things? Probably a little bit of both. The hard part in that is the regulatory cycle. [57:59] against lots of human genetics and lots of human organs. So [58:04] We have an effort to do 10,000 tests, and it could be 1,000 different tests. [58:11] genetics, so people, people, and 10 organs each in one week, in one untouched robot. So it just does the test, screens the drug and looks for effects and growth and all that. [58:26] Maybe something like that would replace human trials. Most likely, at some point soon, it'll reduce the size and length of these trials. So I think drug discovery will see not only better molecules. The bigger problem in drug discovery is not to do what's called lead compounds for a disease, for a drug, but to increase the probability of success.

[58:56] have a lead drug candidate to actual FDA approval is so low. Like out of 100, less than five will make it. So it's increasing that probability because you're starting with a better candidate and you're managing the process better. So that's a quarter. I do think there is diagnostics imaging. And by the way, we had the most... [59:26] coolest self-driving for MRI machines, so you don't need MRI machines. Now, you need today technicians to run MRI machines, but if you have a self-driver, you don't need that. Cardiac MRI is so specialized because the heart is moving that you can't staff centers with cardiac MRI technicians. [59:48] We turn an hour and 15 minute exam into 20 minutes, single button press. You can even do it remotely. [59:56] Because so we did the same for ultrasound, self-driving for ultrasound machines. An ultrasound machine might cost $25,000, but a technician to do a cardiac MRI will cost you $150,000. So why care about the machine? It's the person. If you have self-driving, we had this company that GE actually acquired from us. [1:00:18] So all I'm saying is diagnostics and imaging is a quarter of the spend. It'll get cheaper. I'm shocked it's a quarter of the spend. I didn't realize that. Testing of all sorts.

[1:00:30] And then a quarter is in hospital care. These are very rough numbers, but in hospital care. And that will improve too. So the net of all of this is like a multi-trillion dollar retrofit. Yes. Yeah. Yeah. [1:00:45] But. [1:00:46] If expertise is free and you can give [1:00:50] somebody 10 times the amount of care. The average person [1:00:57] US citizen gets one primary care visit a year, roughly. In Australia, it's four or five times that. So we are poor anyway. If you provided way more coupons, [1:01:11] Thank you. [1:01:12] Cure. [1:01:13] Up front, you reduce disease burden downstream. Fewer people would have heart surgeries, which are super expensive, or emergency room visits because you took care of them at the beginnings of the detection of their disease. So every expensive disease, whether it's cardiac, Alzheimer's, Parkinson's, [1:01:36] We are well on our way to discovering them 10 years before the symptoms manifest. So I have... [1:01:42] argued a piece I did called 20% doctor included in 2016. I said, symptom-based medicine will disappear. You don't detect disease from symptoms. You detect disease based on what's going on in the body. So, uh, [1:01:59] That's medicine. Yeah. I'm pretty excited about it. The level of precision that's possible. I think in the next five years or 10 years, biological intervention will approach computer programming in being able to precisely program things, design molecules. So we are starting to, we have a couple of startups looking to do drugs for one patient at a time.

[1:02:29] Wow. Anomaly. And that's because that includes like the patient's DNA is factored in or something? Yeah. Their disease, their specific mutation can be addressed for them. Wow. There's also broader categories like a single shot treatment for sickle cell disease, right? [1:02:52] Massive problem. Yeah. If you can do a genetic treatment, one shot, I think that'll happen in the next 10 years. As I'm listening to you talk about all of these different areas and like you think on such a grand societal scale, but you probably also have a bunch of sort of maybe seemingly boring, but like good ideas kind of come through Coastal Ventures all the time, too. Yes. Do you still enjoy making those kinds of investments as well? Is that more where you have [1:03:22] people on your team who are more interested? Well, one, there are other people in the partnership who... [1:03:27] enjoy just successful startups. I do too. Look, I enjoy winning. When I do a GitLab, that's a win. When we do Cognition, it's a win. When we do Replit, it's a win. I mean, the growth rates for some of these companies are stunning. Shocking, yeah. So yeah, it's still fun to be disruptive. Yeah, it's just sort of a different bucket of [1:03:57] societally impactful technologies,

[1:04:02] I will go through actually starting the company. But Fusion, Bob was a senior fellow, Commonwealth Fusion. Bob was a senior fellow at the MIT Plasma Fusion Lab when I met him. [1:04:15] Now, he had ideas, and I got him – [1:04:19] sort of like instigated, so to speak. Yeah. He's done all the work. And then he instigated the field of fusion. So whether he succeeds or not, there's so many different startups now. One will succeed. Yeah. So Conwell Fusion or Helion don't have to succeed for fusion to succeed. So I like instigating things. The same thing we did. It was very profitable to make 2,500 times your money. [1:04:50] having the world be TCP IP instead of ATM. It's interesting to look back at the press from 1996, just assumed it was going to be the internet was going to be ATM. So I'm actually, you know, I always tell Pradeep, the founder of that, if we hadn't done that, [1:05:09] The internet wouldn't be TCP IP, which is horrific to think about. It'd be in the hands of AT&T engineering the protocols, which they had done. And this is why Lucent went out of business. Practically, they're still around. Alcatel, Newbridge, Nortel, all the telecom players sort of gradually disappeared because they didn't win place or show in the race for the internet. And the same will happen in other areas.

[1:05:39] You used the word instigating when you could have maybe used a word like incubating. I also heard earlier in the conversation, I think you talked about not being an investor. I think also the way that you work with founders, and I've heard you talk about not being founder-friendly or that there's a different way to think about being founder-friendly. What I'm curious about is there's a lot of sort of... [1:06:04] uh, [1:06:05] new ways to describe what [1:06:07] VCs do and how they describe themselves and mindsets around it. Do you think that in general, the role that venture investors see themselves playing, do you think we're trending in a direction where it's getting better all the time? Or do you think that in any ways, the industry has lost its footing on some of these concepts? Look, there's [1:06:26] All kinds of VCs with all kinds of approaches. [1:06:30] Some are marketing labels. [1:06:32] Founders [1:06:34] Friendly is a real disservice to founders. [1:06:37] It's like if you said to your kids... [1:06:41] Mm. [1:06:42] If you told your five-year-old... Do whatever you want. Do whatever you want. I will always say yes. Yeah. That's... [1:06:49] You want to go swimming without me? Yeah, like whatever. Or eat as much candy or eat as much Coke. And by the way, my kids, when they were growing up, [1:07:01] At age five, they could eat as much candy and as much Coke as they wanted, as much junk food. [1:07:07] We always had it laying around.

[1:07:10] They didn't have to ask us. [1:07:13] Bye. [1:07:14] We taught them how to measure themselves on how much Coke they drink, whether it's good or not. So I would say we taught them control, not gave them yes, no permission. Yeah, but the most important thing was that you set those parameters for a kid. Yeah. Yeah. Right? Yeah. [1:07:33] Literally, at age 12, I thought of they can make 10% of the important decisions. By age 18, they have to make 90. And my job is to teach them how to make the leash longer. But the same is true of entrepreneurs. So a brilliant engineer from DeepMind starts a company. [1:07:55] What do they know about finance? What do they know about marketing? What do they know about even hiring or managing? Our job is to teach them. [1:08:07] just like I taught my kids. By 18, my kids were pretty independent. I didn't worry about what they did in college. All four went to college, [1:08:15] five miles from our house at Stanford. None of them had this, I want to get away, because they felt they had enough leash, enough permission. They didn't need to, quote, unquote, get away. Can I put up like a steel man argument of the other side to hear your reaction to it? Basically, it would be something like, [1:08:35] The sort of the strongest entrepreneurs, which is where all the returns are, don't really need much help or much management. And so better to just invest in the people who didn't need help in the first place. I disagree. That is very much the founders fund argument. Yeah, they do very well with that. Great founders do well, whether you like it or not. Yeah, whatever you do.

[1:08:59] Where I will disagree. [1:09:02] is they can do even better [1:09:05] if they have a debating partner. So my view is very simple. I almost never... [1:09:11] work at a board. So I don't even go to boards now. You won't show up to a board, do you mean? I don't get on boards. And Square is a great example. I got off the board the week before the IPO and everybody was like, hey, this is sexy IPO. Why are you getting off the board? I don't like the burden of- The governance. [1:09:41] better. [1:09:42] to governance for the public market. It's not very fun. It sounds glamorous, but it's not. You know, Jack and I would do regular monthly dinners, and I'd always challenge him and always push him. And he loved that, so he took the time to do these dinners. You talk to Max Lefkin at a firm. It's a public company. We long ago distributed stock. [1:10:12] for an hour every quarter. [1:10:14] Right. Even the strongest founders like Jack and Max and need this. It's like even a pro sports player has a coach. Yes. You want people to challenge you. [1:10:27] and debate you

[1:10:29] but not make decisions. I find it silly that boards want to take a walk [1:10:36] I won't even take a vote on the acquisition. I'll just say I'll support whatever management wants to do, even if I disagree strongly. But I do want to have the opportunity to debate it with the management team. So Jigdeep Singh is a founder, one of our robotics companies. I've done six companies with him. There was a company... [1:10:59] I'm forgetting the name called light something optical company. The team wanted to sell. I only asked to give me two hours with the whole team to make. And I did over the weekend the presentation I gave to them on why they should not sell. But before that presentation, as long as that, the two hours, I signed the papers and said, you have my word anyway. You can choose to use it or tear it up. I love that. Yeah. [1:11:29] Yeah. [1:11:30] Give the final say to the management team. Yeah. [1:11:34] But I do want to debate it. And I think every strong founder loves business. [1:11:40] feedback that challenges them. And it's the only thing really valuable to a strong founder. Great founders love that. Not so great founders are so uncertain of themselves, they don't like that. And there, I prefer to teach them how to be a great founder. But almost every strong founder loves feedback. I had Keith on this podcast and

[1:12:07] what you're describing is a big part of, I think, what drew him back to Kosa. Yeah. By the way, on that topic, I'm curious. I read that as one of the sort of generationally great [1:12:19] sort of recruiting or re-recruiting, um, [1:12:23] you know, moves that has happened. And I think he's an amazing investor, obviously. Did you know for a while that you wanted that to happen? We're a different kind of firm. And I said earlier, we are not, I've not called myself an investor ever. Right. I always say I'm a venture assistant. If you look at our website, since we started the firm in 2004, the first tagline is venture assistance. By the way, [1:12:53] I prefer brutal honesty to hypocritical politeness. In vain, most VCs give... [1:13:01] Polite... [1:13:01] feedback to an entrepreneur, they're doing a disservice. They're basically saying you're doing great. Don't examine anything. It makes you feel good as an entrepreneur, but it doesn't help you build a better company. And the only things that make you feel better are things that change your thinking or challenge your thinking. So it's an important characteristic for founders to keep in mind. What are they looking for? The best help or just people who will go away? So it's a style [1:13:31] It's different. [1:13:32] We don't think of investments. We've never calculated, never means not in the last 10 years. I can't think of a single instance where we calculated an IRR.

[1:13:44] Not one. An investment firm that never calculates IRRs because we want to build something significant and then the economics will work out. So that's our approach. [1:13:57] But we do want to make [1:13:59] increase the probability of success of the founder, [1:14:02] And we want to increase the magnitude of the potential success also by thinking about should you do X or Y. Founders underestimate that. So let's say a founder is selling 10% of a company. [1:14:19] 10 million at 100 million post. [1:14:24] The founders forget they're keeping 90%. And what happens to the price per share of that 90%, whether they give up 10% or 12%, keep 90 or 88, trivial compared to the 500% swing you can get by taking one path or a different path. That's where the opportunity for maximization is. That's where it's not about investing, but getting the best help building a company. [1:14:54] And that's what I sort of spend all my time doing. Just yesterday, I was talking to this founder. It's a very cool company, self-driving for these heavy machines, bulldozers and loaders. That's cool. Right. And we had an interesting session. [1:15:13] And I reframed his business from automating big Caterpillar machines.

[1:15:19] to [1:15:21] I'll just rent you the driver. [1:15:23] You have trouble staffing drivers. Most of these are in mines and remote places. Equipment, [1:15:30] A multimillion dollar piece of equipment sits idle because you don't have an operator. I said, just do rental by the hour and guarantee you will always have a driver available. [1:15:43] Always have a job. [1:15:45] This reframing, the entrepreneur is pretty excited about. It's the same thing, but it's not. Yeah. It's a much easier sell. It's not saying change your equipment. It's saying, hey, you can't keep your equipment running all the time because you don't have operators. I'll run it for you. You're paying operators. I'll do it better, faster, more hours. You get more of your equipment. It's a simple framing change. But... [1:16:14] You could have a similar technology discussion, too. I know you're going to have to go in a minute. So I just want to ask you, there's one, and there's a million more questions I want to ask. But there's one I really wanted to ask, which is you've been not just like – [1:16:27] doing venture, but you've been doing it like really successfully and in a super relevant way for a long time through a lot of cycles. Before we started, you said, you know, you're turning 70 or you turn 70 this year. You feel 25. This is kind of like the dream. And do you know what you attribute all of this to? Yeah. Very importantly, I give a talk at Stanford in 2015, the Stanford Business School 2015. I said, I'm internally driven.

[1:16:57] I don't care what others think of me. So it feels obnoxious sometimes because I'll do my thing independent of what others think. I'm doing this because it's fun. [1:17:07] I get somebody teaching me about mining and [1:17:13] constraints in construction or self-driving an MRI machine or a big loader, they're the same thing. Clever ideas like public transit or can we make fusion happen? So, [1:17:26] I'm in it because it's fun for me. [1:17:29] And most of the people at Khostla [1:17:32] love what they're doing and would do the same thing if they didn't get paid. [1:17:39] And so we do it for a different reason than making the most money. [1:17:45] Because we focus on larger impact, [1:17:49] The returns really do well or take care of itself. You know, it's sort of like what's motivating you. In fact, this was a conversation I had Sam and me investing in Open AI. He knows I care about the impact. Something happened, making something happen. Fusion is too important to not try it. I want the impact of fusion. There's nothing I could do in making money that would impact me more saying, [1:18:19] part in getting fusion to start and hopefully happen. Even if Commonwealth Fusion doesn't do it, almost all the companies have been enabled and investing in the field has been enabled because people looked at Commonwealth Fusion. So this idea of instigating change is very motivational. It's way more impactful and also fun and way more learning than almost anything else.

[1:18:49] I have one addiction, which is to learning. I can't learn enough new things, whether it's about biology or an AI algorithm or hypersonic flight. We are doing Mach 5 flight, too. And we haven't talked about national defense. We are spending a lot of time there. Would love to. But all these areas. [1:19:10] Just fun. Yes. And that's why I don't mind. I still work 80 hours a week, and I hope 25 years from now I'm working 80 hours a week, health permitting, touch wood. But- [1:19:22] I think it's a very different motivation of why we do things. I tell our LPs, I'm probably the only one foolish enough to tell my LPs, if I can have more impact and less return, I'll pick more impact every single time. As long as they get a minimum return that they're comfortable with. Well, you seem to be getting both. Vinod, thank you so much. This was amazing. I am super appreciative of your time and all that you've done for the industry. Great. Thank you. It's a lot of fun. [1:19:52] Bye.