U.S. AI Policy Is in a ‘Dangerous Place’ Says VC Martin Casado

[00:00:03] Exchanges, the Goldman Sachs podcast featuring exchanges on rates, inflation, and U.S. recession risk.

[00:00:12] Exchanges on the market impact of AI.

[00:00:15] For the sharpest analysis on forces driving the markets and the economy,

[00:00:20] count on exchanges between the leading minds at Goldman Sachs.

[00:00:23] New episodes every week. Listen now.

[00:00:33] Welcome to Tech News Briefing. It's Monday, October 28th.

[00:00:37] I'm Zoe Thomas for The Wall Street Journal.

[00:00:40] A handful of scientists and engineers are working to give electric motors a major upgrade

[00:00:47] using a technology pioneered by Benjamin Franklin.

[00:00:51] We'll tell you about the efforts to develop modern electrostatic motors.

[00:00:56] And then, Martin Casado, general partner at venture capital firm Andreessen Horowitz,

[00:01:02] says concrete risks from the artificial intelligence boom haven't materialized.

[00:01:07] But many U.S. officials are calling for a slowdown in AI development.

[00:01:12] At WSJ Tech Live, he shared his thoughts on U.S. tech policy and the outlook for venture investing.

[00:01:19] We'll have highlights from his conversation with our global tech editor, Jason Dean.

[00:01:23] But first, Benjamin Franklin is credited with inventing many tools we still use today.

[00:01:34] Bifocals, the lightning rod, even swim fins, according to the Franklin Institute.

[00:01:40] Now, a handful of scientists and engineers are creating modern versions of Franklin's electrostatic motor

[00:01:46] with the help of materials and techniques unimaginable in the 1700s.

[00:01:51] Here to tell us about these efforts is our tech columnist, Christopher Mims.

[00:01:56] So, Christopher, let's start with the basics.

[00:01:58] What is an electrostatic motor?

[00:02:00] So, a regular motor uses something called a Lorenz force,

[00:02:04] and it transforms the continuous flow of electricity through a tight coil of typically copper wire

[00:02:11] into the rotation of the axle of a motor.

[00:02:15] And if you run it in reverse, that's a generator.

[00:02:18] An electrostatic motor uses a totally different type of force.

[00:02:23] It's the attraction between a positive and a negative electric pole,

[00:02:29] or the repulsion between two poles of the same polarity.

[00:02:33] So, it's hard to visualize because normal motors just don't work like this.

[00:02:38] But just imagine, like, you know, two magnets pushing each other apart.

[00:02:41] And you get the idea?

[00:02:42] Has the electrostatic motor been used in any form yet?

[00:02:46] First working electrostatic motor was built by Ben Franklin,

[00:02:50] and he used it on the banks of a river to rotisserie a turkey, appropriately enough.

[00:02:57] It was part of a summertime electricity exhibition.

[00:03:00] He did all kinds of funny things, like he shocked people and stuff.

[00:03:05] And it's really just been this scientific curiosity,

[00:03:08] because as simple as it is to build,

[00:03:11] it works on a fundamentally different electromagnetic or physical principle than a conventional motor.

[00:03:19] And the reason that we don't see it in our everyday life is that typically it just isn't that great.

[00:03:25] It doesn't generate a lot of force.

[00:03:28] It doesn't spin that quickly.

[00:03:29] It's not super efficient.

[00:03:32] However, the same principle shows up.

[00:03:34] It's actually in your phone, in your pocket right now.

[00:03:37] There's electrostatic actuators in there that are making up some of the sensors inside of your phone.

[00:03:43] So, that's the one place that it's been used since the days of Benjamin Franklin.

[00:03:47] If it's got some drawbacks, what are the benefits of an electrostatic motor?

[00:03:52] So, if you fast forward to the present day, there are a couple of PhD electrical engineers who said,

[00:03:58] what if we could combine everything that we've learned about electricity and magnetism in the modern day

[00:04:05] and make a new kind of electrostatic motor?

[00:04:07] And that's what they did.

[00:04:08] It's this kind of marvel.

[00:04:10] It has these really fast-switching power electronics,

[00:04:13] which are the same things that make modern electric vehicles possible.

[00:04:17] It has this weird organic dielectric fluid inside of it instead of air,

[00:04:24] which keeps the various parts of it separated and insulated from one another.

[00:04:29] And it's got this very complicated printed circuit board inside.

[00:04:33] And the net effect of all of this is that you end up with a motor,

[00:04:37] which is just way more efficient for a lot of the applications that we use motors for,

[00:04:42] which are just all over the place, especially in industry.

[00:04:45] What's the timeline for something like this to become ubiquitous?

[00:04:49] When there's a technology this radical, it takes decades.

[00:04:51] What I would compare this to is the first lithium-ion batteries,

[00:04:56] which the public could buy.

[00:04:57] That was like the early 1990s.

[00:05:00] And yet they were still a curiosity for years and years and years.

[00:05:03] And then this company called Tesla came along and had this radical idea of like,

[00:05:07] why don't we power a car with these batteries?

[00:05:08] And everybody said, that's crazy.

[00:05:09] They're going to catch fire.

[00:05:10] It'll never work.

[00:05:11] So you can see how many decades it's taken for us to just accept,

[00:05:15] oh, these super powerful batteries are everywhere.

[00:05:18] Same thing with these motors.

[00:05:19] It could take many decades.

[00:05:22] All right.

[00:05:23] That was our tech columnist, Christopher Mims.

[00:05:26] Coming up, we'll hear why a prominent VC says when it comes to AI,

[00:05:31] the U.S. government's stance on tech policy is in a dangerous place.

[00:05:36] That's after the break.

[00:05:43] With Ecolab Science Certified, we take cleaning off your plate

[00:05:46] so you can focus on what's most important to your restaurant, your guests,

[00:05:49] and having them switch from giving your restaurant a go to making it a go-to spot.

[00:05:53] Ecolab Science Certified.

[00:05:54] Count on a scientific clean.

[00:05:56] Learn more at sciencecertified.com.

[00:06:03] Martin Casado started his career at Lawrence Livermore National Laboratory,

[00:06:09] working on large-scale simulations for the U.S. Department of Defense.

[00:06:13] Now he's a general partner at venture capital firm Andreessen Horowitz,

[00:06:17] leading its over $1 billion infrastructure practice.

[00:06:21] Casado says while the AI boom has already taken off,

[00:06:25] it's not too late for investors to find opportunities.

[00:06:28] He spoke with our global tech editor, Jason Dean, at WSJ Tech Live.

[00:06:33] Here are highlights from their conversation.

[00:06:36] We're talking about the outlook for venture capital,

[00:06:39] and we can get later to whether there is anything in that category other than AI,

[00:06:44] but let's start with AI for now.

[00:06:47] Heaps of money has already been invested in this.

[00:06:50] Valuations are soaring.

[00:06:52] What type of AI startups are interesting now

[00:06:57] and not too expensive or too late?

[00:07:00] So I actually think it's probably good to draw a comparison to the Internet

[00:07:05] when thinking about this.

[00:07:06] So periodically in the history of the industry,

[00:07:10] we've seen the marginal cost of things go to zero, right?

[00:07:12] So with compute, the marginal cost of computation went to zero.

[00:07:16] So before we had people calculating logarithm tables by hand,

[00:07:20] and then we had a computer do it.

[00:07:22] That created the compute revolution.

[00:07:24] And then the Internet, the marginal cost of distribution went to zero, right?

[00:07:27] And so instead of sending things over the email or the mail,

[00:07:30] we do it over online.

[00:07:32] So when it comes to AI,

[00:07:34] it really feels like the marginal cost of language reasoning and creation

[00:07:38] are going to zero.

[00:07:39] And if that's the case, this is a super cycle.

[00:07:42] And if that's the case, we've got decades.

[00:07:43] So there's kind of no too late in that sense.

[00:07:46] We're still very, very early.

[00:07:48] Now what makes this different than the Internet,

[00:07:50] which is actually very important,

[00:07:51] is in the case of the Internet,

[00:07:53] you have this kind of early wave of investing in things like fiber.

[00:07:56] You have this big build-out.

[00:07:58] And that was done by companies that would take on debt.

[00:08:01] They would lay all this fiber.

[00:08:03] And then once the demand for that fiber dropped,

[00:08:07] we entered into a glut.

[00:08:08] And that created the dot-com crash.

[00:08:09] So right now, these models that you hear about from like Google

[00:08:14] and OpenAI and Anthropic,

[00:08:17] they're actually backed by these massive companies

[00:08:20] that have hundreds of billions of dollars on the balance sheet.

[00:08:24] And so you can't really look at that market in the same way

[00:08:27] and say, is it going to crash?

[00:08:29] It's not going to crash.

[00:08:30] Clearly, this is strategic value to them.

[00:08:32] Clearly, this is working.

[00:08:35] Clearly, this is the beginning of something great.

[00:08:37] And so the way that we think of the world is we segment it almost in two.

[00:08:40] We're like, there's these very large models that everybody's heard about.

[00:08:43] You've heard about Gemini and you've heard about OpenAI.

[00:08:45] And those are backed by very large companies and people are using them.

[00:08:49] But there's a bunch of other smaller models

[00:08:51] that do things like speech or music or images.

[00:08:54] And if you look at that cohort of companies as an investor,

[00:08:57] they're actually very successful.

[00:08:59] So if you take almost a foundational comparison against the Internet,

[00:09:03] you'd be like, the Internet, a lot of this was on debt.

[00:09:05] But it wasn't backed by these large companies.

[00:09:08] And it was very hard to monetize where in the AI wave,

[00:09:12] you've got these great foundational technologies, which have great backers.

[00:09:16] But also, if you look at the startups around it, they're doing very, very well.

[00:09:19] So I would say very early, lots of success cases.

[00:09:23] And we're very optimistic.

[00:09:25] Within that, though, I mean, there's already huge valuations

[00:09:28] for some of the foundation model companies.

[00:09:30] And the barrier to entry there seems like it might be getting more difficult.

[00:09:35] Within that landscape, what is most interesting now?

[00:09:39] So there are things that are definitely working.

[00:09:43] So the big foundation models like the OpenAI, the Anthropoc,

[00:09:46] that's kind of not a game for private capital.

[00:09:48] But on the other side, there's these smaller models

[00:09:50] that are different modalities,

[00:09:52] which actually are these kind of very solvent companies

[00:09:55] that are doing very well.

[00:09:56] And then maybe many more will work in the future.

[00:09:59] Your leaders of Andreessen Horowitz, Andreessen Horowitz,

[00:10:04] were pretty vocal about supporting Trump

[00:10:06] because they thought it was good for the industry.

[00:10:09] And then Horowitz changed his view more recently.

[00:10:12] What do you think is at stake in this election

[00:10:15] for tech and for AI regulation?

[00:10:17] I think it's really hard to say.

[00:10:20] I'm going to answer a bit of a different question quickly,

[00:10:22] which is the doctrine of the US on tech has just changed.

[00:10:25] Like I remember in the early days of the internet,

[00:10:28] I mean, all my formal studies were basically

[00:10:29] in network security and cybersecurity.

[00:10:31] And in the rise of the internet,

[00:10:33] you had these very clear impacts to safety, right?

[00:10:36] Like we had worms that had taken down critical infrastructure.

[00:10:39] We actually had a shift in like US military doctrine.

[00:10:43] We went from like mutually assured destruction

[00:10:45] to like what was considered like asymmetry,

[00:10:47] where if we attacked a less like developed country,

[00:10:51] we'd be more vulnerable because we had the internet.

[00:10:53] Like it was very clear, it kind of changed the nature of risk,

[00:10:56] the martial nature of risk.

[00:10:57] And even then, what did we do?

[00:10:59] We invested, we created disciplines,

[00:11:01] like the national labs were on the forefront of it.

[00:11:03] I worked on it.

[00:11:04] I mean, like we viewed these things

[00:11:05] as we need to be the leader.

[00:11:07] It's very important to be the leader.

[00:11:08] And we need to like get researchers on it

[00:11:10] to make this stuff more secure.

[00:11:11] And we want academia and the national labs.

[00:11:13] Like that was our posture.

[00:11:14] That was like 20 years ago, 25 years ago, 30 years ago.

[00:11:18] Now the ASF is so strange because like,

[00:11:21] there's been no demonstration of risk, of marginal risk.

[00:11:24] Marginal risk meaning?

[00:11:25] Marginal risk means new risk that doesn't exist on the internet.

[00:11:28] Someone with Google, right?

[00:11:29] So like what is a model for risk that didn't exist before?

[00:11:32] Open research question.

[00:11:33] We don't have demonstration of risk.

[00:11:35] There's a lot of rhetoric, but it just isn't there

[00:11:37] like we've had it before.

[00:11:39] And yet there's like the white blood cells are out

[00:11:42] and they're like, oh, we need to stop it.

[00:11:44] And we need to pause it.

[00:11:45] And like, I am so okay with these conversations,

[00:11:48] but if they're baseless, you can't make any progress,

[00:11:50] especially when you're referring

[00:11:52] to this high level aphorisms.

[00:11:53] And so I think the posture of the country

[00:11:56] with respect to tech has kind of gone

[00:11:58] to a very dangerous place

[00:11:59] when it comes to like say global leadership.

[00:12:00] I think it has changed.

[00:12:01] I think we have a new doctrine.

[00:12:03] I could not tell you which of the two candidates

[00:12:05] would do the right thing with respect to this.

[00:12:08] But I will say where we are isn't a bad place.

[00:12:10] With that, we will have to end it.

[00:12:12] I'm afraid we're out of time.

[00:12:13] Thank you so much.

[00:12:14] This is fascinating.

[00:12:15] Okay, absolutely.

[00:12:17] That was Martin Cassano,

[00:12:18] general partner at venture capital firm

[00:12:20] Andreessen Horwitz,

[00:12:22] speaking with our global tech editor, Jason Dean.

[00:12:25] And that's it for Tech News Briefing.

[00:12:27] Today's show was produced by Julie Chang

[00:12:29] and Zoe Culkin with supervising producer

[00:12:31] Catherine Millsop.

[00:12:33] I'm Zoe Thomas for The Wall Street Journal.

[00:12:35] We'll be back this afternoon with TNB Tech Minute.

[00:12:39] Thanks for listening.