Preet Bharara:
From CAFE and the Vox Media Podcast Network, welcome to Stay tuned. I’m Preet Bharara.
Vinod Khosla:
I don’t plan on success most of the time. I just plan on asymmetric success once in a while. So I don’t mind a 90% probability of failure if I have a 10% chance of creating an OpenAI and changing the world.
Preet Bharara:
That’s venture capitalist and entrepreneur, Vinod Khosla. Khosla’s career began as a founder himself. In 1982, Khosla started Sun Microsystems, the company that would go on to create the programming language Java. He now heads Khosla Ventures, a venture capital firm with a mission to back disruptive and high impact tech companies. Khosla has made big bets on technology, including a $50 million investment in OpenAI, the company responsible for ChatGPT. Khosla joins me to discuss the trouble with predictions, the future of AI, and the ethos of innovation. That’s coming up, stay tuned.
Q&A
Now let’s get to your questions. This question comes in a tweet from Twitter user @Canhoser1. The question is very simple. “Does Trump have to file for bankruptcy now that he can’t pay the New York fine?” Well, that’s a great question, an interesting one. The quick answer is no, he does not. And as we’ll discuss in a moment, filing for bankruptcy doesn’t get him off the hook. So to remind everyone, there was a judgment by the court in New York State based on the lawsuit brought by New York Attorney General Letitia James. And the total judgment, including pre-judgment interest is about $454 million.
Now, the question keeps getting raised, how if Donald Trump is a multi-billionaire, can he not afford to pay the $454 million? Well, first of all, he doesn’t want to pay it at all and he’s appealed, which is his right. But in order to appeal, according to New York law, he has to post a bond. Now by a bond, that means an outside company, a bond company will allow Donald Trump to put up collateral of some substantial form for the purpose of guaranteeing that the $454 million judgment can be paid. According to Trump’s own lawyers, that amount, the amount of the bond he would need to post is something like $557 million.
The problem for Donald Trump is, again, per his own lawyers, that dozens and dozens of bond companies have declined to be the company to post the bond in this particular case. And as people have speculated, that doesn’t come as a complete surprise given Donald Trump’s overall financial standing and perhaps a concern about being associated with this particular cause.
Now, why can’t Donald Trump pay the judgment or post the bond? Well, in part because he’s not so liquid. Regardless of what you think about the overvaluations and the exaggerations of his properties over time, including as the basis of this court action, he does clearly have enough money. The problem is much of it is tied up in real estate. So he may be property rich but cash poor, at least a half a billion dollars appears to be too much for him.
He has other options though. He can of course sell one of his properties, or he could seek a loan as he has done in the past by putting up some of his properties as collateral. He could also have an even more wealthy and liquid benefactor lent him the money or put up the collateral, but that hasn’t happened so far.
Now, just as with the New York Attorney General case in the E. Jean Carroll matter, you had to post collateral to a bond company in order to post the $91.6 million bond. And the money in collateral connected to the E. Jean Carroll case cannot be used a second time in the attorney general case.
Now, you mentioned bankruptcy in your question. To be sure, remember, in the New York Attorney General case, it wasn’t just Donald Trump individually who was sued by the Attorney General. There were also various Trump entities that were sued. Those entities could presumably file for bankruptcy and get some measure of protection. But Donald Trump as an individual is not allowed to use bankruptcy as a shield against the judgment. So what happens if Donald Trump can’t put up the money himself or can’t get someone else to or can’t get a bond company to help post what he’s required to? Well, starting in a few days, the Attorney General of the State of New York has the ability, even though it’s only a civil case, to freeze bank accounts and sees assets. We’ll see if that happens, but you can expect there’ll be a massive litigation surrounding any efforts to do that as well.
This question comes in an email from Christian who writes, “I listened to Elie’s…” That’s Elie Honig. “I listened to Elie Honig’s book Untouchable, and he talked about how he was one time conflicted off of a case because his dad’s law partner represented a potential witness. How often do prosecutors have conflicts in cases and have to leave the case? Who decides in small and big offices if there’s a conflict?”
Well, those are a series of good questions, Christian. I don’t know what the percentages are, and I think it varies from office to office, but I think the number of times that a prosecutor has to recuse himself or herself is relatively small. And I’ll tell you about a personal experience of mine in a moment. As to who decides. I can speak for the Department of Justice in the federal system. We had, in the Southern District of New York, an ethics officer, someone whose responsibility was to advise on all manner of ethics and integrity, including issues relating to conflicts, and he was a resource for the office for that purpose. There are also ethics officers at the Department of Justice in Washington that people can seek counsel from on whether or not they should recuse and what the best policy would be.
Generally speaking, recusals happen more often when there’s an appearance of a conflict than when there’s an actual economic conflict of interest. You can imagine the kinds of ways in which that might arise. For example, from time to time I would recruit someone who was in private practice, a partner in private practice, to come back to the Southern district and serve as chief counsel or serve as chief of the criminal division or in some high position at the Southern District. And obviously in the course of their private practice, they may have represented a witness or a target in an investigation that was ongoing at the Southern District of New York. And so reasonably and quite understandably, that person would be completely recused from the ongoing case where they had some interest back when they were in private practice. So that was an obvious and clear recusal example.
I’ll give you another example of something that never came up, but had there ever been during my tenure as US Attorney, a criminal or civil investigation of Senator Charles Schumer for whom I had worked and who recommended me to the position, I obviously would’ve recused myself from that, and it’s possible that the whole office would’ve had to recuse itself from that kind of investigation. That never happened, but that gives you an example of the kind of issue that might require recusal.
Sometimes there are issues where a prosecutor at a particular office may have a spouse who works at a law firm and the spouse is entwined in a criminal investigation, and then you have to work that out, but likely the prosecutor would have to recuse. In my own circumstances, in seven and a half years, I only had the recusal issue come up in a serious way for me one time. It was a case in which a very, very good friend of mine, a longtime friend of mine was on the board of directors of a company that the office had decided to open an investigation into. My friend was not the target. My friend was not the focus of any kind of investigation, but he sat on the board. And so I sought the advice of my own ethics officer in the Southern District.
And also for belt and suspenders purposes, also sought the advice and counsel of the chief ethics officer in Washington, DC. They both in writing confirmed that I did not need to recuse myself, that there was no conflict of interest or even an appearance of a conflict of interest. And my team that was responsible for the investigation also their advice was do not recuse yourself. You know what I did? I recuse myself anyway. To me, and this is the lens through which I have looked at the Fani Willis case and the Nathan Wade matter, if there’s any question at all, any question at all that someone might doubt your neutrality or credibility and think that there might be some conflict of interest, particularly at the end of the day, if you don’t bring an enforcement action or a criminal case, if there’s any doubt that could be left in anyone’s mind, the best course is to recuse. And that’s what I did. And that’s, by the way, what Nathan Wade should have done some time ago.
This question comes in an email from Janie who has clearly been listening to intently to our podcast miniseries, AI on Trial, and Janie asks, “How do we know this is the real Preet speaking and not Bot Bharara?” Well, the short answer is you don’t. That’s the number of the problem that we address in the miniseries and that people around the world are going to need to address in the days, weeks, and months ahead with increasing frequency. AI has gotten so good that it can train on a person’s voice, someone like me, it can train on my podcast, and with the right tools and the right inputs can generate a voice that can sound remarkably like mine, and if not now, at least in the near future might be indistinguishable from my actual voice.
If you’re concerned about these issues about AI like I am and like Janie appears to be and you haven’t listened yet, check out our miniseries AI on Trial. In the third and final episode, we address what could happen if an AI chatbot impersonated me as the host of a copycat podcast called Stay Tuned With Bot Bharara. I’m joined, as I am in every episode, by Duke Law professor Nita Farahany as we break down different hypotheticals on what the future holds for AI regulation. For now, take it on my word. It’s me.
I will be right back with my conversation with Vinod Khosla.
THE INTERVIEW
Vinod Khosla has invested in some of the most successful technology companies on the planet. He shares his thoughts on major tech breakthroughs and an AI-enabled future. Vinod Khosla, welcome to the show. It’s great to have you.
Vinod Khosla:
It’s great to be here.
Preet Bharara:
As I said before we hit the record button, I have about seven hours of things I want to ask you about because your interests and expertise extend far and wide, but we’ll try to do it in an hour. Right off the bat, I have two things I want to say for the record. Number one, and for full disclosure, number one, you are a client of the law firm for which I work where I practice. That’s point one. Point two is I will not be billing you for this hour.
Vinod Khosla:
Well, that’s good to know.
Preet Bharara:
This is a pro bono interview as they say. So before we get to technology and AI and innovation and all those other interesting things that you have so much insight into, I want to ask a general question about the economy. I know you’ve made comments about the so-called expertise of experts. So lots of people in the last year, including people who have come on this podcast who were very well credentialed, very smart, very well-meaning, said we were certain to have a recession, and they were wrong. How did that happen?
Vinod Khosla:
Well, one of my favorite all time books is a book called Expert Political Judgment, which is by a guy called Phil Tetlock. He was a professor at UC Berkeley, did 20 years of research on expert opinion. His general question over 20 years following 28,000 forecasts with 84,000 scenarios, which he classified when the forecast was made, not retroactively, his general conclusion was experts have about the accuracy of that throwing monkeys. There’s a consumer version of this book called The Future Babble that Dan Gardner wrote, but the original book is worth reading.
To understand when experts are right or wrong, he has a second book that talks about that. But generally, experts are good at extrapolating the past, not predicting a future when it changes. And it’s a little bit akin to, I have a pretty good weather forecasting model without any expertise in that, which is tomorrow will be like today. And I’m generally 70% right when I use that simple rule. And experts do that naturally. It doesn’t matter if it’s forecast like Russia will walk over Ukraine in a week, which happened a few years ago, they’re still not walked over them. Now the conventional wisdom is nobody can win and everybody has the same forecast. So I’m very skeptical of forecasts and look to see who is inventing a future they want with real passion and drive, and they have a shot at changing the future.
So I’ve lived with this my whole life. In 1985, I was told there won’t be a PC in every home. 1990, everybody saying Grandma will never use email. In 1995, nobody believed in the internet. In year 2000, nobody believed in pervasive mobile. I can keep going on and on and on. Nobody would believe that you can have a phone without a keyboard. That was the big thing before the iPhone was brought on. So-
Preet Bharara:
Yeah. Tell the folks at Blackberry who didn’t predict that.
Vinod Khosla:
Yeah. So my view is, in this more serious versions of this, the AT&T asked McKinsey to make a forecast on the number of cell phones in the year 2000 because they make 15-year plans. And in 1986, and this is a really important illustrative example, they predicted less than a million cell phones in the United States in the year 2000. They were 109 million, so they were off by more than 10000%.
Here’s the question to ask. Could they have predicted, McKinsey, the nature of a processor in the year 2000, the nature of a display, the nature of batteries? You can predict all those component technologies? You can’t make a forecast. And it’s important to know when you can’t make a forecast versus you make a set of assumption based on which you make a forecast. By the way, same is true of the number of electric cars. The Department of Energy Consulting, all the experts at General Motors and Ford and Volkswagen, made a forecast for a number of electric cars. The forecast was made in 2011 for the year 2035, 25 year forecast that Elon Musk exceeded in 2016. So they were wronged by 20 years in five years. So these are all illustrative examples of why it’s much more important to have entrepreneurs like Elon Musk say, “I’m going to make electric cars happen,” not to rely on experts and academic forecasts.
Preet Bharara:
It’s one of the problems particularly in the area of tech that a long-term forecast is virtually impossible, a 15-year forecast with respect to phones or a 25 year forecast with respect to electronic or electric vehicles. Is that inherently a silly enterprise?
Vinod Khosla:
It is a silly enterprise. I mean, sit here today, everybody’s jumping on AI. In 2018, when we invested in OpenAI, people thought… And I won’t name names, this was silly, and that was only five years ago that we invest in OpenAI. Now of course, five year’s a long time ago, and today it’s surprising and shocking. And now we are worried about the inverse, which is sentient AI destroying humanity. It is hard to predict where AI will be in 10 or 15 or 20 years. I’m pretty optimistic. Actually I think that’s an easier forecast to make because we will underestimate what will happen.
Preet Bharara:
I have a lot of questions about artificial intelligence that I want to ask you about, but just another question about this predicting the future and how difficult it is. You are essentially in the business of investing, which is by definition a predictive enterprise. How do you go about making good decisions about the future when you’ve just described how difficult it is for even experts to get it right?
Vinod Khosla:
I think we approach it a little bit differently. We are not making a forecast. We are attempting to develop technology breakthroughs that will cause a different future to happen. See the difference? I’m predicting where we can drive with our own companies the future because of technology breakthrough.
So let’s take a simple example. In 2018, I talked about we invested in OpenAI as AI, we’ll come back to that. We also invested in Commonwealth Fusion systems. When Fusion was on nobody’s list, it was a no touch zone for even the Department of Energy. We counted on technology breakthroughs that we were going to make happen, not others. The biggest was there was already a breakthrough in high temperature superconductors, and we said, “What if we could build a 20 Tesla magnet with this, a really, really powerful magnet?” Fusion would become more tractable. And we made that bet.
In the first three years, we spent just building this one magnet. And we were successful, by the way, about 10 times faster than the ITAR project, which is the multi government based project to build fusion. They could build a magnet in 30 years that we did in three. And so entrepreneurial energy is the drive behind large changes.
Preet Bharara:
How do you measure that?
Vinod Khosla:
You can’t measure it, but let me retrospectively ask you the following. Did Amazon disrupt retail or did Walmart disrupt retail? Was media done by NBC and CBS and Fox? Or YouTube and Twitter and Facebook and Netflix? People who didn’t even know they were in the media business when they started. Did Boeing or Lockheed or Airbus do space or was it companies like SpaceX and Rocket Lab?
Preet Bharara:
Isn’t the answer both?
Vinod Khosla:
No. No, it’s only the startups. It’s only the startups that change access to space in a dramatic way.
Preet Bharara:
This is maybe a dumb observation, but at one point in the past, all those other companies were startups too or not?
Vinod Khosla:
Well, at some point they were. And I think that-
Preet Bharara:
And at their entry points, they were disruptors themselves, right?
Vinod Khosla:
Yeah. And when they lose their founders and the founder drive, they lose their ability to innovate is the point I’m making.
Preet Bharara:
Well then how do you feel about Apple?
Vinod Khosla:
Well, it’s interesting to see what Apple will innovate on. Steve was clearly the driver of much of the changes. He had the courage to say a phone is a different device, it’s not the device you principally talk on. It’s the device with the moniker, there’s an app for that, whatever app you want, whether it’s bird watching or recognizing plants or ordering an Uber. So entrepreneurs drive that. You look at the Uber, Hertz and Avis could have done that or a taxi company. Who did that? Uber. Did Hilton and Hyatt innovate room rentals or Airbnb? Every area I look at in the last 40 years that I’ve been doing innovation, I can’t think of a single example of a large innovation coming from an existing player. Intel can grow from one silicon process to the next generation because it’s so predictable. Still hard. It’s incremental innovation. But large innovation in 40 years, I haven’t seen one of them come from an entity. There’s no innovation in energy coming from General Electric or Siemens, which you’d expect-
Preet Bharara:
Right. But that doesn’t solve the problem of investment. So for example, there was a company that was an innovator called Facebook. There was another company that was an innovator called MySpace. MySpace is nowhere and Facebook is dominant, Meta is dominant. As an investor, when you’re looking at these startups and thinking about what you’ve just been describing, making the case for massive innovation that will be sustainable and expansive, how do you choose between the winners and the losers at the outset?
Vinod Khosla:
So two things to remember. This is a really good example. MySpace was an innovator. I was pretty early in conversations with Rupert Murdoch when he decided to buy MySpace. In fact, we had a long conversation when we were at a News Corp offsite in Carmel, California. On MySpace, he bought the company, but what did they do? Apply media’s traditional vision to MySpace and it lost to Facebook, another startup, despite all the resources of News Corp. That’s a very good example. So what do we-
Preet Bharara:
But my question is that, but at the time that that decision was made, was it foreseeable that that would be the losing proposition?
Vinod Khosla:
No, it wasn’t.
Preet Bharara:
Right. That’s the dilemma, right?
Vinod Khosla:
Here’s the thing to realize. There’s a Harvard Business School case with a quote from me on the first line of case. “My willingness to fail allows me to succeed.” So I don’t plan on success most of the time. I just plan on asymmetric success once in a while. So I don’t mind a 90% probability of failure if I have a 10% chance of creating an OpenAI and changing the world.
Preet Bharara:
Right. But that’s in part because you are an investor and you can diversify your investments. But if you’re the startup and you’re the founder, then you’re not diversified and it’s hard, right?
Vinod Khosla:
You’re not. But a founder doesn’t need the same scale of outcome that we do. So, many founders do really well when we consider something an inconsequential return. Founders also do multiple startups in a row. If one fails, they move to the next. And that’s the characteristic of within a startup, they will iterate and pivot and find the right product market fit. Good founders do that. And if one fails, they’ll do the next. And the great thing about Silicon Valley is that failure is considered good learning not a bad thing, unlike most parts of-
Preet Bharara:
Yeah. Is that universally? So when you’re looking at somebody who is on their fourth startup and the first three fizzled away or failed, intuitively, people who are listening to this would think, “Well, you don’t want to bet on someone who’s failed three times.” And maybe sometimes that’s true. Is it universally true or should it be universally true outside of Silicon Valley as well that those failures were learning experiences which build a path for success on the fourth try or the fifth try? How should investors think about that?
Vinod Khosla:
So here’s the way I think about it. I am not an investor. I’m mostly in the business of backing great entrepreneurs, then helping them succeed.
I’ll give you a very specific example. We backed Jack Dorsey at Square in 2010. I believe it was four or five people in Square. Then called Square, now called Block. If you look at Wikipedia, Jack had failed about four times and just being fired from Twitter as CEO. What you judge is the entrepreneur, not what happened because startups have a low probability of being large successes. And we just learned to judge really good entrepreneurs and really good markets to go after, and in technical startups like Fusion or OpenAI, judge how likely a technical breakthrough is going to be. That’s a judgment call. There’s no science to it. We often get egg in our face, and I don’t mind getting egg in my face if it’s a smart failure. I thrive on the fact that I’m not embarrassed to fail. And I sometimes say my great skill and learning, it comes from the fact that I’ve failed more often than most people I ever meet because I’ve tried more things than most people.
Preet Bharara:
Right. You’ve been up at bat more often.
Vinod Khosla:
Right. The thing to remember is it’s better to try and fail than fail to try. And the vast majority of the people don’t like risks, so they failed to try. From my point of view, they failed before they start instead of trying something and failing at it.
Preet Bharara:
Why your focus on technology? Is it because it’s interesting? Is it because it’s potentially lucrative? Is it because you think it can transform society and make it better or all of the above?
Vinod Khosla:
Frankly, I think technology is the only resource multiplier that’s essential for all 7 billion people on the planet to have the kind of lifestyle, maybe the 10% of the richest or 700 million people have, mostly in the western world but the other parts of the world too. Those of us who live a rich lifestyle are rich in education, rich in transportation, rich in housing, rich in healthcare, rich in entertainment, all kinds of stuff. You can’t have 7 billion people get there the same way with 10 times the number of cars on the planet, 10 times the amount of steel and cement. So we have to invent new solutions so everybody can have a great lifestyle. Loo, for me, it’s a great game to play. I preferred it to golf or chess.
Preet Bharara:
It’s more expensive even than golf.
Vinod Khosla:
It is more expensive, it is more rewarding, and you make a difference. And I don’t like golf and I don’t like sailing. I’ll go for a fun sail once in a while, but I couldn’t do it as my principal activity. Think these things, when you set up a challenge, a hard challenge, a Mount Everest that’s hard to climb and people will fail at attempting it, it’s very fun to be the underdog. I don’t know if you’ve seen 11 large predictions I made back-
Preet Bharara:
Oh, I’m going to ask you about several of them.
Vinod Khosla:
Okay.
Preet Bharara:
I’m definitely asking you about some of them, especially after you said that people who predict are generally fail. So we’re going to hold you to these.
Vinod Khosla:
You can hold me to them. And I may fail at them, but I’m working on every single one of them. That’s the key. I’m not letting it happen in the world. I’m backing entrepreneurs who are trying to make these things happen with a very long term persistent vision and a reasonable technical risk plan. So it’s very much like Tesla. The world wouldn’t have electric cars if Elon Musk had not tried to make it happen. There was no other forecast. Six, seven years ago, Volkswagen had no plans to do electric cars. Today they have no plans to do traditional ice engine or internal combustion engine plants.
Preet Bharara:
Probably screwed up their 15-year plan enormously.
Vinod Khosla:
Yes. But why did that happen? Because Elon Musk illustrated what could be done.
Preet Bharara:
Well, how do you build into, and this is maybe a dumb question, if the only constant is change, and you know that in any 15 year period, for sure, given the modern rate of technological advancement, that there will be unforeseen advancements and twists and turns, some of which you’ve described? How do you build that into your 15-year plan? Or should 15 year plans be abandoned altogether?
Vinod Khosla:
Look, directionally, there’s things you can say. Specifically, you can’t predict anything. So I’ll give you a funny example. In 2004, I gave a talk called the device that used to be a phone. Right? This was three years before the iPhone. I had one speculation that there was enough compute power in other things to build a computer into your phone. And so if that was true, then you wouldn’t use it mostly for talking. I was abysmally wrong on how it might be used. But directionally, the fact that… And thanks to Steve Jobs, he conceived all kinds of apps running on a phone without a keyboard, but with a large display, a computer in your pocket. And by the way, none of the traditional players, and if you remember back then it was Motorola and Nokia as the dominant phone producers, got this. In fact, they fufu’ed this device which was going to cost too much, not have a keyboard, you couldn’t talk on it easily. That’s how innovation happens. Somebody chooses to drive it. That’s what happened…
I was at Kleiner Perkins when we backed Jeff Bezos in 1996. He drove e-commerce. I was there when we backed Google in 1998, I think Larry and Sergey drove that change. So you rely on entrepreneurs with real passion to be smart enough to make the technology breakthroughs and take really smart shorts on goal.
Now, if you know hockey, not all goals go in. In fact, most shots don’t go in. But the occasional shot that goes in makes up for the ones that didn’t go in. And that’s how we look at investment. We take high risk bets on entrepreneurs trying to build a great vision and a great company. And occasionally they’ll build a small company. Very often they’ll build a small company and consequential to that, no, but not to us. But most entrepreneurs, ambitious entrepreneurs at least, strive for something larger than a small company. It’s like a category defining company. It’s a reputation defining company. It’s a reason to get really rich. All those things, motivations fall in. And the drive and energy you see and the persistence and going through the highs and lows of entrepreneurship is what makes these law changes happen or not. That’s why I like to say while experts extrapolate the past, entrepreneurs invent the future they want or fail trying to do that.
Preet Bharara:
I’ll be right back with Vinod Khosla after this.
I am going to get to your predictions now, Vinod. Number one, and this is a good segue into a larger conversation about AI, prediction one, and you made these predictions a few months ago, “We will be capable of having subject to no political barriers.” That’s a big caveat, but, “We’ll be capable of having subject to no political barriers, AI based near free tutors for every child and doctors and lawyers for every citizen 24/7 that will amplify our professionals 10 times, expanding accessibility and affordability.” I note that you didn’t put a timeframe on that.
Vinod Khosla:
I didn’t put a timeframe because most of my predictions were 25 years. This one I think will come true in the next 10.
Preet Bharara:
Should I look for other work?
Vinod Khosla:
Possibly. Well, podcasting may still be important.
Preet Bharara:
Okay. Well, I don’t know. I mean, we have a podcast series, a special miniseries on AI. And you could train a large language model on me and my content. I don’t know that I’m necessary in the future.
Vinod Khosla:
Yeah. Or pick the best if you and the best of the next best podcaster.
Preet Bharara:
Yeah, well just in second place.
Vinod Khosla:
Yeah. But look, in this country, if you are in the lower half of society, you can’t file for bankruptcy without hiring a lawyer. So just when you have no money, you can’t, I’m told. Shouldn’t somebody have an AI lawyer that can file for bankruptcy when they have no money instead of having them come up with the money to hire a lawyer? In this country, the average patient citizen sees their primary care physician once a year. In Australia, they see them four or five times a year. Shouldn’t every US citizen have that? And for most villagers in India, a primary care doctor may be a two-day ride into the city and back when AI is great at incorporating expertise.
And so my son is building a primary care doctor. In India it has very high diagnostic accuracy rate for primary care doctors. We have another company, Limbic Health, that’s building a mental health therapist. Their diagnostic accuracy is 93%, which is higher than the diagnostic accuracy of a therapist typically doing intake. This is studies in the UK at the National Health Service clinics, mental health clinics. These are large numbers. This was a study with 64,000 patients. So isn’t that a great service to have?
I’m arguing if you know India, there’s a payment service that has eliminated the need for Visa and MasterCard fees, which is that tax of 2 or 3% on the economy when they’re used. It’s essentially free called a UPI or Universal Payment Interface and Personal ID for every citizen voluntarily. Shouldn’t they also offer free primary care and free mental health services and free tutors? My wife has a nonprofit called ck12.org that has AI tutors available to every child. They can do better than your classroom teacher. Why? Because this tutor can map the gaps in knowledge, what’s called knowledge tracing in a student’s gaps of learning individually. In fact, if you give them a problem, the AI can predict what’s the probability they’ll solve it correctly because they know where the gaps in learning are and what’s in the… And then they can tutor to these gaps in learning.
These are wonderful services to have and they’re about as expensive or maybe a little more expensive computationally than a Google Search. So this should be available to all 7 billion people on the planet. By the way, hugely deflationary because we spend hundreds of billions of dollars on medical expertise, which will be near free.
Preet Bharara:
So these are all the good upsides of Ai. And I know you’re an optimist and a cheeleader.
Vinod Khosla:
And let me give you one other fact. And this study on mental health, just published with a different study than the one I mentioned on diagnostic accuracy. It was a study on inclusion and equity. Diversity and inclusion went up 25% when they used the AI therapist relative to human therapists.
Preet Bharara:
Well, that’s certainly good, but now I want to talk about the pessimistic side. There are people who are very smart and who have been involved in the creation and development of AI who are very worried. And there’s a raging debate everywhere you look, including on our podcast, about what the nature and scope of regulation should be.
I want to read you a few quotes from one of your own pieces back in September and then have you explain why you think this. “If Western values are to win globally, we must approach AI cautiously but not at the expense of losing the AI race.” You’ve also written, “Aggressive AI regulation or calls for slowdown if enacted would add further material risk that we lose some momentum in this race.” And then finally, “To address these risks inadvertent and advertent, we must substantially increase our research and investment in safety technologies but not aggressively regulate AI.” Explain your thinking about the balance between the freedom to improve and develop AI versus regulation to protect us.
Vinod Khosla:
Humanity faces many risks. We know an asteroid can hit our planet and extinguish humanity. There’s a certain probability of that. There could be a pandemic that could be even more severe than COVID. That’s a risk. Humanity faces a large basket of risks. And then there’s many acts like Putin who may start a nuclear war. So there are many risks and humanity has to manage a basket of risks. I’m not saying the probability is zero that a sentient AI in the next five or 10 years takes over and destroys all of humanity, which is the scenario that some people postulate. Is that a possibility over the longer term as a risk? Absolutely. And very smart people Yoshua Bengio in Canada and others are very worried about it and genuinely worried. The problem is it’s being amplified beyond reasonableness because it so matches the headlines the press wants to write about science fiction.
Preet Bharara:
Right, but can I just pause for a second? How can you say it’s being overemphasized when the thing we’re talking about that you’ve just conceded is a possibility is extinction. Isn’t it hard to exaggerate that risk?
Vinod Khosla:
There is an extinction risk with an asteroid hitting the planet.
Preet Bharara:
Right. But the asteroid is not capable of regulation by the Congress.
Vinod Khosla:
Yeah, but I’m getting to my point. What is the larger risk with AI? The largest, and orders of magnitude larger risk with AI, is bad nation states, our enemies, win the race. First, interfere, which they will in 2024 elections in a significant way. And if our AI is behind theirs, we’ll lose. More importantly, if you take a 10 or 20 year perspective, whoever wins the AI race will win the economic race and hence the right to influence many parts of the world, whether it’s Southeast Asia, Africa, Latin America, India. If you get the winning AI, we’ll get winning economic power and then the ability to influence. That’s the race either western values will win or China will win. I don’t think North Korea is in that race, but they can do lots of bad things. They can’t do good things. China can do both good things globally. I think what’s at risk is Western values winning as a political system, much bigger risk than the sentient Ai. So I’d rather not lose to China and work hard on the sentient’s existential risk that people take off. Sometimes in AI it’s called hard takeoff.
Preet Bharara:
I want to go back to your predictions because prediction two is also a doozy and see if you can explain it in a way that shows how momentous a prediction it is. “There could be a billion programmers all programming in natural language. Many parts of the craft of coding would be commoditized and the entire field would be opened up to those without a computer science degree, democratizing computer coding as well on its way in the next 10 to 15 years.” Could you explain to people how radical a development it would be if people could program code in natural language, what that means?
Vinod Khosla:
So think of the simple fact. You have to learn computer programming. You have to learn an application. Whether it’s Oracle or SAP or Uber or whatever application you’re using, you have to learn to use it. Human beings have to learn the language of computers, which buttons to press, which menu to access, what it does. And AI will enable computers to understand English and I’ll say, “Hey, in this Excel spreadsheet, add a row here and sum up these columns.” That’s English. And if I’m not clear, the AI will ask me for a clarification. “What do you want to do beyond column 17? Because there’s no data after column 17.”
So you’ll be able to converse with an AI in English language and it’ll then write the code. It’ll then do the things you’ve done. And then everybody can write code. You want a report in on something, you don’t call IT department and say, “Produce me a report.” You say, “This is the report I want,” and the AI writes it for you. That’s what I mean by a billion people being able to write code. And I think that’s happening very, very quickly and has surprised everybody.
Preet Bharara:
Yeah. Now when that happens, and I don’t know about coding and I certainly don’t have a computer science degree, when we get to the point where it’s just as efficient for people to code in natural language with the assistance of AI, does that mean that we get to a point where nobody knows how to actually code anymore and we become completely dependent on AI?
Vinod Khosla:
No, in all these areas, and when I was talking about doctors and to teachers and lawyers in coding, expertise will be in the system, but developing new expertise required that expertise. So there will be a number of programmers who are 10X programmers better than everybody else leading the way in what these systems can do. Or new disease diagnosis, for example, there’ll be 10X better. Expertise will be in AI, but the top 10 or 20% of any profession will be guiding further development of that AI or overseeing how it’s applied. Oversight will be very, very important into the human context. And so we will see oversight by humans in almost all these areas, at least for the next 10 or 15 years.
Preet Bharara:
Right. But that’s not very long. I mean, I’ve already-
Vinod Khosla:
No, it’s not.
Preet Bharara:
… made fun of the idea of 15 year prediction windows. But just for a moment, indulge me. That’s still possibly within the span of my lifetime. Now talk about 50, 60, 70, 90, 100 years from now, do you dare to make any potential prediction about what the world will be like with the assistance of AI in a century?
Vinod Khosla:
Yeah. So longer term forecasts are much easier to make than what will happen next 10 years. I can make three-year forecasts and I can make 30-year forecasts, but 10 is harder. What will happen, I think all labor will be free. And one of my forecasts is a billion bipedal robots doing more work 24/7 than humanity can do today, among the 7 billion people on the planet. Labor will be free. Expertise, whether it’s physicians or oncologists or teachers, will be free. We will have a time of great abundance.
In 2016, I believe I wrote a blog in Fortune, I believe, that said, “AI will cause great abundance, great GDP growth, great productivity growth, almost everything economists measure and increasing income inequality.” And we can come back to that. I think we will have a time of great abundance. We will have a distribution problem which we will need to address, and we will have the ability to address. Because you imagine 3% growth for 50 years or 100 years and compute the per capita GDP, it’s astronomical. I think that’s what will happen. Most things will be near free. A physician will be near free. Most medical care will be hugely deflated. Maybe we’ll still have to have MRI machines, but their throughput will be five times higher, which we are already doing by the way, by replacing say cardiac MRI technicians at Mass General. They’re already had twice the throughput they were before adopting AI self-driving for MRI machines and becomes more accessible for patients.
Preet Bharara:
But I’m confused. What does an economy look like if all of this labor is free?
Vinod Khosla:
I think the need to work will go away. People will work because they want to work on the things they need to work. And society will have enough abundance to assure a minimum standard of living. And people always say, “What will people do for jobs?” Most of the jobs in this economy are not great jobs. And I think those kinds of jobs will go away. People will have a higher standard of living than that person can afford today.
Preet Bharara:
But there has to be some income, I don’t understand how this works, unless we have a universal basic income.
Vinod Khosla:
So in 2016 is the first time I actually, in this assay, talked about the need for a universal basic income.
Preet Bharara:
Ah, that is the solution.
Vinod Khosla:
That is the solution and the ability to afford it will be there. In today’s context, it’s hard to imagine, but do the following calculation. Let’s say GDP growth goes from 2% to 4% or 6%, which will happen with these increases in productivity and efficiency and free labor and others. We can get to physical goods too. I believe that those are part of the same economy and will be a little more of a constraint, but not as much as we think today, but that will allow that increment in GDP growth from 2 to 4% or 6% will allow for enough additional income to take care of everybody.
My bet is the best way to handle it is to take the incremental growth over 2% and put a percentage that into a reserve fund. We have that for oil in Alaska, for example. Alaska has a state oil fund from all the oil riches. Many of the oil companies from Norway to in the Middle East have these funds. Think of AI as very similar to our oil gusher we’ve discovered, which will create value that we should take some percentage off to make sure everybody gets to be protected at the basic livelihood level and take economic growth above 2% and use it to create reserves for the population.
Preet Bharara:
I want to ask you about innovation generally. Is it your view that the US has a vast advantage in technology and innovation over every other country? Or are we neck and neck with China or is it some other configuration?
Vinod Khosla:
I think the US has historically had an advantage because it’s been a risk-taking culture. I came to Silicon Valley in 1978 to go to Stanford. I came here with the dream of starting a company. Why did I come to Silicon Valley? Because I read an article about Andy Grove, a Hungarian immigrant coming to Silicon Valley and starting Intel. So he was a role model as an immigrant for a new immigrant like me. And so the Valley has become a magnet for people all over the world who want to take risks and want to achieve whatever their vision is, whether it’s making more money or changing the world with electric cars. So the US has a huge advantage. Around that has developed a culture of investors who will take the risks along with entrepreneurs. And they’re not really investors. I say, I’ve never called myself a venture capitalist. I’ve said I’m a venture assistant.
Preet Bharara:
Right. But it’s in all your bios that other people do of you.
Vinod Khosla:
I like to help entrepreneurs build their companies based on all the screw-ups I’ve done in my life that I talked about earlier, that ecosystem of investors encouraging risk-taking not safety, employees from an established company like Google being willing to join, give up safe jobs and safe salaries and join a startup and cut their income by 70% in return for stock options. That culture is dominant in the Valley and encourages the kind of risk taking. Those ecosystems are developing in other places. Shanghai is a place. There’s other places in India and Bangalore. But it’s very strong here and the Valley continues to attract talent from all over the world. And talent is the key fuel for the innovation,
Preet Bharara:
Right. I mean, how much of this is reliant on a good and generous and smart immigration policy?
Vinod Khosla:
Look, I think smart immigration policy helps a lot to get the best talent from everywhere. So many Russian PhDs looking to get out of Russia right now. Our companies are setting up offices closer to Russia in Turkey or Dubai or others because it’s harder to come into the UW. But a great immigration policy to attract this talent will let us win both the technoeconomic race that I talked about earlier with China, but also generally innovation race. And I think that is key. But there’s enough talent locally too. I mean, it’s not like only immigrants start companies.
Preet Bharara:
Right. They only become the CEOs of the major technology companies. Not all.
Vinod Khosla:
It is surprising how much that has happened, but that speaks to US culture. It’s so open, the right people get to the top, especially in parts of the economy that are dynamic and change rapidly where competence matters a lot. If a segment is not growing or growing at 1 or 2% a year, you can get away with doing next year what you did last year. And so old boy networks work in those industries. They don’t work as well when competence matters a lot, which is very true of the economy, technology economy.
Preet Bharara:
Yeah. What’s the role of the education system? So we have been told ever since I can remember that our elementary schools are not good. We’re behind in reading and math as compared to many, many countries around the world who don’t do as well as we do in the ways we’ve been discussing. Is the difference that our elementary schools and our high schools are not so great, but our universities are excellent? Or is it something else?
Vinod Khosla:
I think our universities are excellent, but our universities are also excellent at allowing for much more exploration. When I went to IIT, I had a much more prescribed curriculum than when I went to Carnegie Mellon for my master’s degree.
Preet Bharara:
Right. We should pause and explain to people what IIT is and how difficult it is to get into IIT in India.
Vinod Khosla:
The Indian Institute of Technology is a collection of institutes, started in the 1960s in India. I don’t know the exact statistics, but I think one out of 300,000 people who apply get in, much higher selectivity than Harvard or Princeton or Yale or Stanford.
Preet Bharara:
Or MIT.
Vinod Khosla:
Or MIT. In fact, I was talking to somebody, this was about 10 years ago, well accomplished person in India, in the technology business in India, and he said, “My son’s backup school is MIT if he can’t get into the IIT.”
Preet Bharara:
Right. So that’s an extraordinarily ambitious and selective university, and I interrupted you, but you were comparing that to an American university unfavorably. Why?
Vinod Khosla:
Because they’re much more rigorous and the ecosystem around it doesn’t allow for experimentation and trial. If you look at most parts of the world and most parts of US business, they reduce risk in businesses to the point where the probability of success goes up. But the consequences of success are minor or inconsequential. And the venture business is about starting companies with a lower probability of success, but much larger consequences of success. Create a new Google, create a new Meta, create a new Tesla, pick your favorite. And that’s cultural.
Preet Bharara:
Vinod Khosla, thanks for your time and your insight. It was a treat having you on the show.
Vinod Khosla:
Thank you very much.
Preet Bharara:
My conversation with Vinod Khosla continues for members of the CAFE Insider community. In the bonus for Insiders, we discuss the promise of new climate technology and what the American dream means to him.
Vinod Khosla:
Most immigrants can break through with enough skill, persistence, and talent. And America allows for that environment, allows for us to create what is essentially value for America. It’s a symbiotic system.
Preet Bharara:
To try out the membership for just $1 for a month, head to cafe.com/insider. Again, that’s cafe.com/insider.
BUTTON
To end the show this week, I want to shift gears for a moment. This week marks the official start of spring in the Northern Hemisphere as people all over the world celebrate the march equinox. That word, by the way, is derived from two Latin words, equi, which means equal, and knox, which means night. It’s when the earth’s two hemispheres get an equal amount of sunlight in a day. Here where I am in New York, the days are getting longer and feeling warmer. Now I like the briskness of fall, but spring always brings with it a certain optimism, a feeling of rebirth and renewal. But as it comes upon us this year, it’s hard to square the promise and cheer of spring, the anxiety and uncertainty of this political season and the state of the world right now.
As some of you may know, I have been known to enjoy a poem or two. A little while ago, I spoke on the show about the question of whether or poetry was dead. Many, many of you wrote in to tell me thankfully that it wasn’t. I will look to poets from time to time for inspiration and meaning during uncertain times. So with that in mind, I came across this poem lately that I want to share. It’s by the great Emily Dickinson. It’s called A Light Exists in Spring.
“A light exists in spring, not present on the year at any other period when March is scarcely here. A color stands abroad on solitary fields that science cannot overtake that human nature feels. It waits upon the lawn, it shows the furthest tree, upon the furthest slope you know. It almost speaks to you. Then as horizon step or Noons report away, without the formula of sound, it passes and we stay. A quality of loss affecting our content as trade had suddenly encroached upon a sacrament.”
If you have a favorite spring poem or something else giving you inspiration at this moment, please send it to me at letters@cafe.com.
Well, that’s it for this episode of Stay Tuned. Thanks again to my guest, Vinod Khosla. If you like what we do, rate and review the show on Apple Podcasts or wherever you listen. Every positive review helps new listeners find the show. Send me your questions about news, politics, and justice. Tweet them to me at @PreetBharara with the hashtag #AskPreet. You can also now reach me on Threads, or you can call and leave me a message at 669-247-7338. That’s 669-24-PREET. Or you can send an email to letters@cafe.com.
Stay Tuned is presented by CAFE and the Vox Media Podcast Network. The executive producer is Tamara Sepper. The technical director is David Tatasciore. The deputy editor is Celine Rohr. The editorial producer is Noa Azulai. The audio producer is Nat Weiner. And the CAFE team is Matthew Billy, Jake Kaplan and Claudia HernĂĄndez. Our music is by Andrew Dost. I’m your host, Preet Bharara. Stay tuned.
