The Silicon Valley Insider In The Pentagon | DoW Emil Michael

0:00Emil Michael:I haven't seen a technology that has the potential to disrupt so many other industries in my lifetime. Imagine that some Chinese tech companies are stealing anthropic tech effectively by doing distillation. Then the Chinese government can use the effective capabilities of the anthropic model against the United States. But I would be using a lesser version of anthropic bound by their policies. So I'd be one arm tied behind my back against an anthropic model that's fully capable by an adversary.

0:42Welcome to Grit. I'm Juven, partner at Kleiner Perkins, a show where we go beyond the highlight reel and explore the personal and professional challenges of building history-making companies. are you going to the there's a shindig tonight at the white house right after the state of the

1:00Emil Michael:union you mean oh oh oh there's a state of the union and then there's a little there's a party after are you going to the state of the union uh no that's usually you know those seats are prized right because the president calls out people you know his invited guests and and congress takes up the the gallery the audience so um i'm not planning to go i guess what i really actually wanted to know is are you going to meet the usa hockey team that's the only thing i really wanted You just ask. You just ask. I wish. We might see. Maybe we'll see him around tonight. Those guys look like they have fun, so if they're around tonight.

1:31Tell me if you meet any of them that still have a voice after the last 72 hours of what they've gone through.

1:36Emil Michael:I know. Hopefully they're not stuck on their social media looking at all the haters and just celebrating. Do you get haters these days more than you used to? Much less than I used to. ironically enough i would say that you know being at uber during the 2016-17 time frame when all these sort of social happenings were were going down and there was sort of kind of like now there's a little bit of a moment of backlash against you know ai companies back then there was a backlash against um some of these mobile app companies whether it was airbnb or uber and the press was super negative. But remember also then the press was less fractured.

2:21Emil Michael:So there was just like a lot of negativity. Now there's good and bad. You know, there's different elements of the press that support what the administration is doing, some that don't. So you kind of get more of a balance as odd as that sounds. I can't remember when all of the Uber, what size was Uber when you joined? I mean, we're a hundred people. Okay, small, small still. Yeah, I left when we're about 20 ,000. How many cities? I mean, we were probably in 500 or 600 cities by the time I left. We were only in one or two when I'd started. And at that point, were you still in the media, at least portrayed as the good guys?

2:57Emil Michael:Yeah, we were the good guys in 2012, 13, 14, 15. Bad guys in 16, the worst. There was an article I always remember. My wife reminds me of it, which was like Emil Michael and Travis Kalnick are literally the worst. It was like Puck News or something like that. Some, you know, some rag. And that was sort of the bottom was 17. Can you refresh me? Is the real reason why, because you all went from David to Goliath or whatever, like is the real reason why you went from underdog to, you know, the incumbent and therefore everybody just turned on you? There was a couple of factors. I would say that there was some of that underdog to Goliath, which you see some of the AI companies sort of experiencing now, which then everyone sort of, you know, everyone always likes the revolutionary, not the incumbent, so to speak.

3:56Emil Michael:So you become the incumbent. There's some natural sort of antibodies that develop. But also 2017, being in San Francisco when President Trump was elected was like a moment for that city where it was like convulsing, you know, in just a politically odd way. And I'll never forget. And I, we were invited to a business council, like Uber execs, Travis and I, and you know, Tim Cook was there. This was like a broad, you know, who are the important companies? And we got invited and some set of employees kind of went nuts. They started chaining themselves to our front doors and internal protests, right?

4:39Emil Michael:Internal. And then that led to a, I don't know if you remember this delete Uber campaign because there was some miscommunication about like, remember surge pricing from Uber? We used to have surge price, people hated it. If it's raining, the cost goes up. Right, yeah, there's less car. You want more cars when it rains. To get more cars out there, you have to raise the price so drivers make more money. Sort of a market-driven approach. But people still don't like it. And there was some protest about asylum seekers or something in the first months of the presidency. And we said, OK, we're going to stay out of this.

5:14Emil Michael:We're going to eliminate. We're not going to have surge pricing to JFK during this protest because the protest was happening there. And people thought some taxi driver thought we were trying to bust the taxi union. And it spun out of control. And half a million people deleted the Uber app that day. Half a million? Yeah. That's a lot. Because, you know, it was very easy to delete the Uber app, take a screenshot, post it on your social. You're a cool protester. You know, Uber is bad. And so, you know, not to go too deep on that, but that was one other factor. It was just like the sensibilities of Silicon Valley, given what a monopolitical culture it was, were offended.

5:52Emil Michael:And then, you know, we were in a Me Too era. We were in lots of different factors. And we had a really poor investment base, investor base. Did you get shown the door with Travis? Effectively. Shoot me straight. Are you salty about it still? Oh, yeah. I'll never forget that, nor forgive. Are you rooting for Uber these days? Yeah, I'm still a shareholder. I mean, the thing about I still love Uber as a product, as a business. You're not salty about Uber. It was the investor base. No, I'm proud of the company. I mean, it's a global brand. It solves a huge need. It's a super profitable company. We should talk about the number of babies that were born in Ubers because people were like, I'll take an Uber to the hospital versus an ambulance.

6:40Emil Michael:And how many drunk driving deaths were avoided because people used Uber and they started to rely on it as a utility for their social life. And how it made outer cities accessible and increased commerce. So there's a lot of good externalities from Uber. And the business model always works. That's why I held the stock, right? And still do and still root for the company. And I've gotten to know Dara and I think he's a good leader. You saw the acquisition they just did? Yeah, saw that. Saw that. Spot hero? A place to park autonomous cars, basically? Yeah, I guess. I guess that seems like what the theory is that I've read.

7:20You guys literally Uber had it. You had it in like 2015. You had it. You had autonomy.

7:27Emil Michael:me yeah we we bought this this carnegie melon robotics company and we were investing um very early before tesla was had fsd or anything like that waymo had still been around but they but they weren't as big as that as big as we could have been and that was one of the things is the investors didn't like that we were this one investor didn't like that we were betting on the future so far ahead. And what entrepreneurs do? They're like, look around corners. They're like, hey, we, this may take a while, but I have to start doing this because you can see the potential disruption of autonomous vehicles become really prominent.

8:07Emil Michael:So we were investing in that. And it was a long-term project, cost a lot of money, but they didn't know they wanted to preserve their embedded gains rather than try to make this a trillion dollar company. And that was sort of another source of conflict. When you said earlier, like, you know, like about your schedule, I asked like, you know, how is your schedule? And you said, hey, if my boss calls to your bosses, uh, Hegseth, is that right? Yeah. Secretary Hegseth. Okay. Can you like maybe draw, like, what are the things that you really care about in your, in your job today? I care about a few things.

8:39Emil Michael:I'll say that one of the, the things that I've been spending the most time of late on is trying to infuse AI into the Department of War. The Department of War is 3 million people. It's literally the biggest organization in the world. And therefore, it has all the opportunities and challenges you find with how do you make a 3 million person organization that has technology that's 20 years old from a sort of productivity standpoint, how do you make it more efficient? And then actually people are happier or if you could take the mundane tasks off their hands. So there's the enterprise level of how you can use AI here, just like any big organization would.

9:21Emil Michael:And then we do a lot of intelligence gathering, and a lot of that's done manually. So how do you use AI to sort of synthesize lots of data and give humans more capability to analyze more? And then for warfighting, if there's a conflict, how do you take these, use AI tools to make conflict you know, more winnable by us, ensuring that, you know, we have the right assets, the right place at the right time, that we're seeing the right things and so on. So that's number one, it's like getting AI into the department and usable across all these echelons and doing it in a smart way. The second thing is new entrants.

10:01Emil Michael:All these companies in El Segundo, all these defense tech companies and all the money that's going into them, I want to help them succeed for a couple reasons. Number one, they have good product. They're starting with a clean sheet of paper and they're saying, I'm inventing a new product that's not from the Cold War era, that'll be cheaper, more manufacturable, and has sort of the latest, greatest scientific and engineering thinking on it. And then there's a lot of capital that's crowded in around those companies. So helping them succeed in this big bureaucracy that we're in is something I do every day.

10:35Emil Michael:Because if they succeed, then more money goes into that sector, which means less spending and R &D by the department, more success by entrepreneurs, more people joining those companies. So it's a virtuous circle. So I spent a lot of time trying to get these new companies in. And then third, I try to solve critical national problems like critical minerals, right? We have a critical minerals problem. We have an industrial-based problem where we've outsourced so much for so long, batteries, like, you know, there's several categories, how do we redomesticate them so we're never dependent on an adversary for something we really, really, really need.

11:10Emil Michael:And that's sort of the broader administration strategy, but I'm doing the defense part of that. That makes sense. Call it like CTO-ish for the Department of War, chief procurement officer internally-ish of the Department of War, and then chief mineral procurement officer externally of the Department of War. I'm making up names, but something like that. not chief procurement officer, chief venture capitalist. Like I'm trying to bet on the companies and the people who could build new capabilities. You're like the corporate arm of an established, you know, something like that. In a way, yeah. Like Salesforce Ventures, I don't know.

11:48Emil Michael:Something like that. And yeah, and then we have the Office of Strategic Capital here, which is where we have a bunch of money to go put in critical minerals companies to encourage them to do things in the US and battery companies and so on. And that's like a little bit more of an, like a pure investment philosophy is like, is this company going to produce a set of products that we need to have domestically? And can I back them with a low cost loan and have venture capital crowd around that? Because it's otherwise unsexy. Very interesting. You must be, I think Dario's coming to pay homage to you all today and sort through things with Anthropic.

12:27Can I like give you my version of the beef and then you tell me your version of the beef. Sure. And I'm just going to like proud, we're like, we're proud Anthropic investors. So I will put my bias on the table and just say I love that company. But they are one of the very few approved LLM vendors for the Department of Work. And they're approved because of some of their partnerships with Palantir and some of the access to information that they have that nobody else does. And the department's like, great, let us do what we want with the technology. And the thought book's like, well, do it what you want, except for a few things.

13:03And you're like, well, that's not really how this works. And they're like, well, that's not how we want to kind of solicit our technology. And so now there's an impasse. Things have rhetoric has escalated. And I imagine everything will get sorted out. But he's coming to D.C. I'm sure he's going to talk to you and Secretary Hanks and sort it out. How far off am I?

13:24Emil Michael:I think you're pretty far off in that. I'll give you an example. It's like if you buy the Microsoft Office suite, they don't tell you what you could write in a Word document or what email you can send or what you could put in a PowerPoint. That's your own regulations and procedures and policies that have your employees and what they're allowed to do on their work products that you give them. And we view AI as an important tool. And we say, as long as we're using it lawfully, we should be able to do that because, you know, here you are out there saying this is the most powerful tools in the world.

14:02Emil Michael:Well, great. Well, we want to use powerful tools at the department and we'll comply with all laws, all regulations. In fact, we have so many internal policies and regulations that, you know, we almost choke on them so much because we're very restrictive because we've got, you know, in our department, lives are at stake. There's no room for error. So in that vein, what I can't do is have any one company impose their own policy preferences on top of the laws and on top of my internal policies, which are pretty robust. And if you're an American champion, like I believe they are, they're one of the most important companies in the country.

14:39Emil Michael:Don't you want to help your Department of War succeed with the best tools available? And it should be a partnership over the long term. You know, we're not, you know, we're not fast moving, so fast moving that we're going to be reckless. And, you know, the ironic example I saw yesterday when the Anthropic sort of article came out about the Chinese attack on the distillation topic was like, imagine that some Chinese tech companies are stealing Anthropic tech effectively by doing distillation. Then the Chinese government can use the effective capabilities of the Anthropic model against the United States.

15:13Emil Michael:But I would be using a lesser version of Anthropic bound by their policies. So I'd be one arm tied behind my back against an Anthropic model that's fully capable by an adversary. It's totally Orwellian. You're saying that the point of the article was like that these Chinese model companies are basically just training using Anthropic and then basically building their own version of Anthropic. And then the CCP is using that against us. I'm just saying that's the scenario. So what the article was yesterday, and Anthropic published this, they call it distillation. You could hit their model enough times that you could sort of reverse engineer what their weights are.

15:54Emil Michael:And then that's obviously civil military fusion in China. So the military could use the effective capabilities of the Anthropic model against the United States. But I have to then be subject to policy restrictions. They're not. Now, okay, this is going to be tough for you, but like, can you steel man the other side for me? What would be their position that you're like, okay, I can like, well, yeah, still mean their position. I think their position is that they have a doomer view of what could happen if AI goes wrong. Like there's fear for the world, for planets, for humanity about what happens in a scenario where, you know, we just saw this open claw breakout the other couple of weeks ago.

16:33Emil Michael:What is the next 10 things like that that are going to happen that, you know, could get out of control? and they want to be more careful about how their software is used. And I think that's sort of the steel man argument for them. And therefore they want to, and they want to comport it to their values, effective altruism and sort of some of the things that, that are embedded in that company's culture and DNA. So that's, I think they're where they're coming from, but you know, there's moments of choosing, right? We had the biggest military buildup in world history in China, in world history. And we are trying to champion our IA companies by ensuring power, by trying to export the American stack, by trying to do all the things to support our American champions.

17:22Emil Michael:We have an AI czar for the first time ever, David Sachs, right? So we take them seriously. So we want these companies to succeed. And I think our only ask is that we be able to use a technology that they say is super powerful for lawful use cases to defend the country. That seems like the argument's pretty easy in my mind, even though I gave you the steel man. That's fair. Let's put your technologist hat on, like your Silicon Valley hat on for a second. Would you agree that this is the most powerful technology that you have seen in your lifetime? It is. And I liken it to quicksand. So quicksand subsumes everything.

17:58Emil Michael:It just sucks up everything that stands on top of it. And the thing about AI that's interesting and really different, just different, it's just so horizontal that you can imagine it's subsuming so much stuff, which is why you saw the stock sell off the other day when Anthropic had said that, you know, people had shown that you could sort of eliminate some of the vertical SaaS categories. And then you're like, OK, well, then can it do travel reservations? Can it do speech writing? Could it create art, movies, all the things that it could disrupt? there's very I haven't seen a technology that has the potential to disrupt so many other industries in my lifetime and given that like argument and by the way like where for what it's worth not that anybody cares but where I net out on this stuff is like all of this is probably not as bad or as good as we think it is like probably not all sass companies are dead and probably we're not going to all be on the beach in a few years.

19:00And probably I won't have like parent cyborgs. I don't know, like could be famous last words. But like, I probably net out that it's like not as good or bad in either in either case. But like, if it is as powerful as I think you and I would both acknowledge, then like, maybe it's good that like, one of the most important companies in the space is at least thoughtful. How's that? Is that fair?

19:27Emil Michael:So thoughtful, absolutely. But I think you have to give credit to the other egg companies who are also thoughtful and the democratically elected government in our country that also is trying to be thoughtful and has a say in defending itself. And we don't have a system and don't want a system where any one company gets to dictate for the country how it governs itself. And it's the other way around. And again, in a capitalist framework, the government sets the rules, companies abide by them, they could succeed. That's how it works, right? I can't have a company saying, well, we have all these laws on surveillance and what the balance of civil liberties and law enforcement are, and it's hotly debated in Congress, signed off by a president, implemented by the executive branch.

20:18But hey, I don't think those are sufficient.

20:20Emil Michael:So I'm going to add my own doesn't work that way. So I think ultimately, I hope this settles out in the right place. But I think, you know, just like with social media, you know, you're just you're seeing certain sort of legislators and people like govern that on deep fakes. and some of the legislation that Tech Cruise has passed about miners and what kind of things you can show and not show and all this stuff, you'll have the same sort of thing with AI. Like as problems develop, government's going to have to address them. And so is industry. Industry is going to have to deal with copyright, right?

20:57Emil Michael:If anaerobic claims or any of these companies claim about being distilled and having their IP stolen, what about all the IP that these models were trained on? Does anyone get paid for that? you know are you going to destroy a whole class of folks because you know how does that work so these conversations are all going to be frontier conversations not you know no pun intended where we're going to have to have them as a as a country um and we're going to have to have them on a on a on a level playing field yeah i mean look like uh we are definitely the these are the solved questions that we need to solve you know like uh there is no perfect answer here like i think both parties have like their own pretty strong points of views that i'm like this has never been answered before like we've never had to ask ourselves this version of this question in the world before and so like we're asking ourselves now and dario's making the pilgrimage to probably ask himself the question in front of you guys and you guys do the same and by the time this airs i'm hopeful that there will be headlines everywhere saying that you're uh your best friends i i hope so and i've always um here's where the advantage of being having been in silicon valley shows itself in the job that i have now which is i understand the cultural dynamics in san francisco i understand sort of enough about this new technology to understand how different it is but i also now understand and try to communicate to friends in the valley what we're up against and what we're trying to do.

22:28Emil Michael:And you do see that more and more now because there's so many Silicon Valley people in the government now that more than ever before, and all of us believe there has to be a partnership. This is not a big tech is bad kind of administration. This is a big tech's good. Big tech can be good. Little tech can be good. How do we work together to make sure we're doing the best thing for the country? So it's a lot more of a partnership approach. And I see that all the time. I mean, there's companies I meet with that said I couldn't get a meeting in the last administration. We'll take all the meetings with the new tech companies.

23:02Emil Michael:We'll take the meetings with the big tech companies. We'll do partnerships where they sit side by side with us with Palantir, you know, with this forward deployed engineer concept and work on hard problems together. And that's, I think, the only way forward. There is no scenario where these powerful technologies are done in contravention of government or without partnership of government. It has to be both. I will say that I'm really glad that technologists like yourself are in the administration because look, like call it what you want, but we may have like 14th century Florence in San Francisco right now for all of its good and bad for all of the craziness, the whatever you want to call it, the homelessness, whatever it is.

23:49I'm not sure you can have one without the other. And the byproduct of that is like the, there is a renaissance happening here. And I think that's starting to become less and less disputed. It's happening in the Bay. There's not nowhere else. Like it's happening right here. And you, you come here, like you see it. We've been to these, like, it's insane. And all the capital is here. And you know, like, I do think that having people like you, I think acknowledging that, like, I think you know that and are protecting that in some ways. Like, I think you are truly trying to protect like this Cambrian explosion of innovation that's happening at a rate that's, yeah, probably we've never seen before.

24:30Emil Michael:There's no doubt about it. San Francisco has a remarkable ability to renew itself with each wave of like, you know, seriously, I was in the mobile app era. Now we're in the AI era. Before that, we're in the internet era, three eras in 25 years, which has created enormous economic value for the world, right? In something like 20 square miles, even less if you just want to go down from Palo Alto up through SF. I do think they have cleaner streets and that's like nonlinear to success. So I don't agree with you on that point. I don't think crime and dirt are, you know, I think those are separately solvable problems.