Transcription
The largest risk in AI, far exceeding any other risk, is AI in the hands of pick your favorite dictator.
It feels like at the moment there's just drama on X around who's getting most funding and who's stealing talent from who. The frontier labs, there's only a handful that matter. You know, it's like if you're a nuclear power plant developer, the one thing you can't afford is a nuclear explosion.
What kind of mechanisms do you think we need to see in terms of business models or incentive structures or perhaps economic redistribution to ensure that the value created by AI is shared? Well, that's a harder question. [Music]
I'm in Palo Alto to visit venture capital legend Venode Kosler. His firm has backed a handful of great Kiwi companies: Rocket Lab, Lanzitech, and most recently Leaf Foods, a company extracting protein from plants. Venode has invested across space, climate solutions, nuclear energy, and biotech. But today, I want to ask him about AI and defense and where he believes the biggest value will be created by both.
Well, thank you so much for doing this, Venode.
Happy to do it. Now, I don't know if you know this, but last time I interviewed you, December 2023, I got engaged a few hours later. So, who knows what's going to happen to me when I walk out of the Coastal Ventures building today. Anything could happen.
Okay, we won't speculate.
We won't speculate.
What's better than engagement?
Well, being married. Being married is a fantastic thing. I'm the biggest advocate for it now. So, thanks to, you know, you're part of that story, which is fantastic.
Great. Now, look, I want to talk to you very deeply about AI. And you made the first early contrarian at the time investment into OpenAI before it was obvious. And you've also written about AI going in two distinct potentially directions, utopia or dystopia.
For the last sort of five or 10 years, when the world has discussed the risks surrounding AI, it's very much been around the deployment of a use case potentially going wrong, like an AI applied biology something going wrong due to a lab leak or self-driving cars kind of crashing into people. But it seems like most of the use cases so far have been seemingly pretty great. And it feels like the larger risk is, at the moment at least, in the leaders of Frontier Labs and the rivalry between them. How do you view the human management of AI potentially being a broader catalyst type risk?
To me, the largest risk in AI, far exceeding any other risk, is AI in the hands of President Xi or President Putin; bad leaders, or people we would consider bad leaders, or different philosophical uh approaches is where the danger of AI is greatest. So there's many dangers AI can be used, but they're generally far smaller than President Xi or President Putin or pick your favorite dictator, uh able to do things that one wouldn't that wouldn't be considered acceptable in western society.
Do you think that there's alignment on that and a focus on that among the leaders of the frontier labs?
It feels like at the moment there's just drama on X around who's getting most funding and who's stealing talent from who, as opposed to the focus on that ensuring that this is done right, in the right hands.
I think the frontier labs, there's only a handful that matter, a very small number, generally agree with this danger and are cognizant of it, paying attention to it, and will do more and more over time in this area. In fact, it should be one of the largest areas of research, especially in academia, but also at the labs. And I think they're all very aware of this and realize that if things even start to go wrong, it'll kill their businesses. And so I I do think they're appropriately cautious. You know, it's like if you're a nuclear power plant developer, the one thing you can't afford is a nuclear explosion. And you know, your whole company gets destroyed if one of those events happens. And I think this is very much like that. And so people are cautious.
You've written and you've mentioned it there, quote, most AI ventures will end in financial loss. In aggregate, more money will be made than lost, but by a small concentration of world-changing companies.
What kind of mechanisms do you think we need to see in terms of business models or incentive structures or perhaps economic redistribution to ensure that the value created by AI is shared?
Well, that's a harder question. Uh let's separate what capital markets drive and what social and political orders can drive. They're very different things. So this idea of uh capitalism means because the impact of AI is so transformative. It doesn't matter whether you're designing new seeds for crops or brasses in New Zealand or uh or you're designing best jet aircraft. No matter where you look, AI will have an impact. And I think the world has realized that. And so every startup uh there's many more startups coming up now going after problems or opportunities. Uh some existing issues, some completely new things that never existed, like new kinds of entertainment would be an example. Could you do a different form of storytelling, for example? uh just one example of a consumer product. So because there's so many opportunities, there's many more startups going after it, and because of that, most will fail. But the ones that succeed will make much more money than so I like to say most startups and most investments will fail, but more money will be made than lost. So the important thing if you're an investor in these areas is to be in the best ones and trying to be a part in the most attractive opportunities.
I wanted to ask you about exactly that, about how you pick the best ones, and OpenAI—sorry, KO Ventures outside of OpenAI has made a lot of investments into where AI is the ex-solution for a problem in in certain industry. When a founder comes to you today with AI embedded in their business or an AI potential solution for an industry, what do you is reminiscent of a say, comesque idea and approach versus something that is truly world-changing like an OpenAI?
You know, the biggest piece of that is usually the founder. So I would say assessing the founder has become even more important than it has been because the founder is the person handling the lever to leverage situations, and in AI the levers are bigger, and so you want a better founder to leverage it for maximum impact, I would say. So founders uh are the biggest characteristic. It's also true that because so many things are new, you want a founder uh who's a rapid learner. So experience is generally a disadvantage in this area. So the more experienced you are, the less likely you are to think out of the box, innovate, reinvent the business you're in. And so a founder that's a rapid learner and is willing to experiment and try and fail often on the way to iterating to success is the right paradigm. So not only do they have to be a great founder, persistent founder, but also a rapid learner. Those are some of the characteristics that matter the most.
So it seems like in your history of venture capital or venture assistance as you prefer to call it, your fundamentals haven't really try haven't really changed. You're still focused more on the founder rather than whatever AI application or problem they're solving.
No, we pay attention to both, but we pay much more attention to the founder than the specific opportunity. Because if it's a great founder and they pick the wrong opportunity in six months, they'll change what they're chasing.
When an AI venture does come into Kosa Ventures, how do you think about the defensibility between ventures leveraging existing AI foundation models like a cursive vibe coding agent versus taking a foundation model and applying it to a completely new industry, something like cell therapy development?
So there's opportunities in both uh and so we do both. In fact, we are doing software development. In fact, multiple aspects of software development. So, uh the company OpenAI just bought Vinsurf is really being an assistant to software programmers. We also have Devon, which is a coding agent. They'll give you a software engineer instead of assist a software engineer. And we have a company called Replet which is computer programming for those who are not programmers. So this is why I like to say I I think in five years or seven years there'll be a billion programmers on the planet. A substantial part of the global population will be able to write code. That's what Replet is doing. So we cover all these things and there's nuance markets for that. But at the other end uh developmental biology, how does a stem cell become an islet cell for to treat diabetes or a neuron to treat Parkinson's or a cartilage cell if you have bad cartilage and need replacement. Uh that's a different kind of thing and we're investing in that too. That's not been possible before ever. So it's a new area, a complete speculation that it'll even be possible. So that area and we've just invested in that is probably where OpenAI was six years ago or seven years ago in 2018 when we first invested in OpenAI.
If you consider the value creation from those two distinct types of approaches though, how do they compare? Because it and Bob Mcru recently has said this, the former head of research for OpenAI has said that if you have like an AI startup for lawyers, eventually the value of that is zero, right? Because then law is available to everyone. Similar to like being a software engineer would be available to everyone in the future and the cost of that is ultimately zero. The only potential value creation is if you are better than the rest in that certain domain. How do you think about that?
Um I do think there will be both of those. I think three things that should be free to everybody on the planet: An AI personal doctor that should be free and eventually will be free for every human being 24/7. A personal tutor for every kid on the planet that's near free now. My wife runs a nonprofit that does that. I'm very passionate about that. And as you mentioned, access to a personal lawyer would enable rights. And most people, if you're an apartment renter and you get kicked out, you you can't really file a lawsuit because you don't have the money to file a lawsuit. Um, so having a personal lawyer would be a great addition to to rights people have. So those are three examples. Um will there be areas in law or medicine to make money? Absolutely. But they may not be this initial starting point. So it will vary by area. By the same token, creating new value. You may not make money out of a model to say how does a stem cell become a cartilage cell, but you might be able to produce enough cartilage cells to transplant into people's knees when they go bad and make money that way. So there's many ways, and at different stages of the development of AI, different things will make money, but almost all of business will be radically transformed.
On the latter specifically as it relates to biotech, it feels like we haven't had that breakthrough moment, like that ChatGPT release moment when hundreds of thousands of people were downloading it and using it instantly. How far away from a breakthrough in that domain do you think we are?
Well, the thing with medicine is it's controlled by regulatory bodies which moves slowly. You can design a drug overnight, but it still takes years to test it in the physical lab, and then you might be able to accelerate that. And I think the next big trend in AI will be AI scientists. So my bet is in by 2035, 10 years from now, we will have many more AI scientists developing new science than human scientists developing new science. So that's an exciting decade ahead of new kinds of science and a greatly accelerated pace of scientific development. Um but when you get run into the regulatory bodies, it still takes time, and for good reason. You want safety if you're treating human beings.
Now I want to talk about some of your more recent investments, and I want to raise ClickHouse first. M, you led that Series C, 350 million US dollar round. That company competes kind of with Databricks, who my husband works for, and also Snowflake, two French founders who I've interviewed previously, although noting that the database with ClickHouse is kind of specifically for AI and the use case for AI agents. How much space do you think there is in these traditional industries like enterprise software and services for AI to rearchitect them?
You know, those three are very different trajectories. So Snowflake and Databricks seem to compete, but I think Databricks has really separated the market it's going after from Snowflake. So I would say four or five years ago Databricks was trying to compete with Snowflake. Today I would say they're in a different market, but Snowflake is trying to copy Databricks is what's happening. In fact, Databricks just bought one of our companies, Neon, I heard. So we are a very large shareholder in Databricks. So we want them to be very successful.
Yet you've also led the round for ClickHouse.
We also led the round for ClickHouse, and the reason is simple. Snowflake is a pretty old-style architecture that hasn't innovated as much as they should, and ClickHouse is more directly competing with the old Snowflake business with a much better and more open-source product. So, these are nuances, but I do think Databricks is doing something different than Snowflake, though. Snowflake is pretending it's the same business.
Okay, another one that I'm incredibly interested in is Mark Industries. You just led that round. Incredible business, incredible founder in in Ethan. There's a huge amount of venture capital flowing to defense technology. And I note without a war that is driving currently, thankfully, huge production volumes. How do defense tech companies, specifically hardware manufacturers, make money in the long run if there hopefully is no war?
So I don't think defense tech companies make money whether there's a war or not. Clearly they can make more money if there is a war and weapons are being consumed. But a couple of things have changed. Geopolitics has changed dramatically. If you look at the Trump administration, they have made it clear to the rest of the world that no company or country can rely on America. They're absolutely insured. And so every country will do something to be independent of the protection America would have afforded. Uh Japan's doubled or tripled its defense budget. Now they're close to North Korea and not far from China. And so every country will have independence, and the money will be made in preparedness and large defense spending. Um that's one thing, but it's also true the nature of war has changed pretty dramatically. Used to be heavy weapons, you know, big bombers, big aircraft carriers, those are sitting targets today. In a new kind of war is now possible. You know, a $1,000 drone can destroy a $50 million bomber, as we've seen in Ukraine. So, the nature of war has changed. The nature of warfare has become much more dynamic, and AI is starting to play a much, much larger role. Um so somebody like Mark Industries can take something that might be a multi-million dollar weapon and turn it into tens of thousands of dollar weapon that's smarter than the big, big weapons. And there's a beautiful book about five or six years ago I read called The Kill Chain, and it argued that the the old paradigm of war and defense spending is completely broken and will not stand up to a real war. And that's what's happened in Ukraine. That's what all the scenarios for Taiwan now say. And so that just creates new opportunity. And because it's mostly software-driven, yes, there's hardware components, and we are doing Mark Five aircraft in a company called Hermier. Very exciting. But uh I do think that creates a large opportunity, and it's very unlikely traditional defense contractors, people in the US like General Dynamics or General Atomics or Lockheed will be able to keep up. They're not used to the space of development. They have no skills in these new areas, and they're not rapid in their response. So I do think it's a whole new opportunity area for venture capital.
So if these defense tech companies from that answer sort of become this defense as a service and are funded by this increase in defense spending, effectively financed by government spending to keep war at bay by developing and producing these types of autonomous weapons and hardware, how do you then stop those companies that are surviving off government funding to not become hijacked by government incentives, like those other companies you've mentioned historically have?
Well, it just, you know, national defense is one of those things. There's no you'll do the your buddy-buddy thing. You have to be able to fight a war. Every army in the world is constantly running war simulations and trying to see how what they need in terms of capability. So that's what will keep it much more dynamic, at least for a while.
The last time we spoke, Venode, in late 2023, we spoke very heavily on the risks and reality of what you called a technoeconomic war. Is there a way where autonomous weapons like the ones that MK Industries are creating, keeping peace, inadvertently accelerate the likelihood or existing likelihood of a global AI arms race? Could that be more destabilizing, and where do you place the risk of that right now?
So the first thing I would say is if you have a strong defense, it dramatically reduces the chances of a war. So if Iran had thousands of nuclear weapons, the chances Israel would attack it would be small. But because they don't, they can attack it, and they can try and prevent—I'm no fan of Iran, so don't get me wrong, but it's a good illustration. If Iran could respond, Israel would be much more cautious about attacking it. But because they can't uh with nuclear weapons, that's why nuclear weapons is such a big stake. You know, I hope we keep Iran from developing one, but you know, there's always a question of does the means at the end justify the means? Uh, I won't go into that here. So, that's sort of the war idea. But when I talk about technoeconomic war, I mean technology-driven economic power. So whoever gets the best uh technology, especially in AI, in energy production is another related area like fusion uh gets the biggest tool set with which to influence uh economies worldwide and hence political philosophy worldwide. I think that is the weapon people like President Xi would like to exercise, and that's the war we have to win, the economic war to win the influence and political ideology war globally.
This brings me back to my very early point around kind of the drama surrounding leading labs. That is exactly what's at stake.
That is what is at stake, and I hope it's western labs that win. Most likely it's an American lab versus the Chinese labs.
We've spoken about the world changing in almost every single measurable area. In a world where that is happening at pace, what remains the same? Venode, is there anything that humans attain or amplify for humans only?
Well, I I do think the nature of human beings and what we aspire for will not change. First, most of the jobs on this planet are not respectable jobs. I like to say if you're a farm worker working in the field picking lettuce for eight hours a day for 40 years, or if you're an assembly line worker at General Motors, mounting tires on a car for eight hours a day for 40 years, that's servitude because you need to survive. Those are not respectable jobs. I think humans will be able to be more human. They will be able to spend more time with their children, with their parents, the things that matter. Yeah. Because they won't be so rushed and driven by economic necessity, which drives especially the bottom half of the population on this planet in terms of economic status. So, I'm quite hopeful AI will let humans be more human. It'll free them up to be able to really be the things that their humanity drives them to. Now, status won't go away. That's also human. Um, trying to beat somebody at sports, your favorite rugby game won't go away. Uh, people will compete in sports, people will compete in the Olympics, people will uh so status, competition, those are very human. But so are caring and relationships and those other things or creation. You can make a living being an artist for once, which you couldn't in the old world. You either got to be an artist and a waiter or you did something more pragmatic. So I do think AI will free humans to humanity to be more human, and hopefully I can make money doing content maybe because you don't can't make much of it right now.
Thank you so much for your time.
Thank you. Appreciate you having me in. It's with Madison. Thanks for watching Markets with Madison. If you enjoyed this interview, please hit the like and subscribe buttons below. I do all of the research for these episodes myself, so your support means a lot. I always love to hear your feedback and ideas, so please drop any thoughts in the comments below. Now go put your money to