This transcript is generated with the help of AI and is lightly edited for clarity.
ARIA:
Reid, thank you so much for being here in New York to film this Reid Riffs. Great to be in person—always, when we’re chatting. So let’s dive right in. The Financial Times recently reported that Big Tech is using guarantees to keep roughly $300 billion of AI exposure off its balance sheet. And the basic arrangement is that separate companies take out loans to build data centers or buy chips and have separate agreements with tech companies to cover losses if things go wrong. And so this helps finance the buildout, but it means that people from the outside can’t really see the full risk that’s happening.
And when you look at Alphabet, they alone [more than] doubled their exposure under these credit backstops in six months, from about $17 billion to $44 billion, while recording a liability of just $815 million. And so this might be a smart way to finance data centers, which you and I probably agree that we need to finance the AI revolution. Or this could be companies that are taking on more risk than they should and obscuring it from the government, the American people, etc. What do you think about this way that we’re financing?
REID:
Well, so I think there’s two different issues here. So one issue is taking the risk, which I think they should do. I think it’s worth it. I think intelligence at the scale and cost of electricity, applied to a whole wide variety of things, is a good thing. And yes, there might be—like, it is not guaranteeing that every dollar will be efficiently spent or effectively spent, because, you know, when does the revenue exactly come in? And at these numbers, you get worried if it’s, call it, four years versus two years and that kind of thing, and there’s some legitimate kind of financial things. Now, there’s the other thing, which is the lack of transparency, which is almost always a bad thing, because it’s kind of like, if you’re hiding something, why are you hiding it?
And so people should be like—in a sense, like, my recommendation would have been if you said, “Look, we need to do this kind of legal structure because we want to still be reporting our books,” because there’s weird things about how expenses and other kinds of things come on the books. And people say, “Oh, you should be doing stock charges as expenses.” And there’s all this stuff. You say, “Look, we want to report our books legitimately, legally, this way. And so we’re going to do that, and so we’ll do these deals, but we’re going to tell people that we’re doing these deals so that they’re aware of them.” I mean, it’s just kind of always the case: when you’re hiding something and you’re obscuring something, like, that’s mostly a very bad thing. Why are you doing that?
ARIA:
And so, I mean, we talked about this last week. AI has sort of broken containment from a news perspective. Now everyone’s talking about it—everyday people. We’re talking about different things, and it’s even reached the government and the president. I will just start with that. And so, very important question, Reid. And I will not be doing any impersonations. “Many people think that the words ‘artificial intelligence’ are inaccurate and very ineloquent relative to AI, or artificial intelligence. A far more elegant and accurate description of this new phenomena would be Superior Intelligence, SI, or Extreme Intelligence, EI, or Supreme Intelligence, SI. This is a poll, and I would appreciate everyone voting.” What do you think is the best name for this ever-growing revolution?
REID:
Well, what I put in Impromptu and Superagency is amplification intelligence, right? Because I’m very focused on how it helps human beings in terms of how it operates. Now, in terms of that Truth Social tweet, it neglected natural stupidity or some other things as possible answers. But I think that the real question—and by the way, if I were really picking a term that I actually probably like more than artificial intelligence, it’d be synthetic intelligence.
ARIA:
Oh, interesting.
REID:
Right. Because part of this synthetic intelligence is—it’s kind of like, it’s not artificial, it’s real intelligence, but it’s being constructed in various ways. And artificial has weird echoes in human language, like artificial sweetener, or, you know—and it is artifice, right? So that thing—it’s kind of got a negative connotation that I think is broadly unhelpful.
ARIA:
And I think it also keeps some of the connotations that we were talking about: that we’re not trying to mimic humans, we are not trying to replace humans. We are trying to create an intelligence that will augment, amplify, help us do our jobs. But again, you know, a lot of people, especially in Silicon Valley, it’s like, “Let’s just build this. We don’t want to deal with government, we don’t want to deal with politics. Let’s just, you know, bring about the future through technology.” But this is a time when we’re going to have to have politicians and Silicon Valley sort of come together for these new rules. And so, on a more serious note, the AI debate in Washington—that wasn’t serious. Anytime I quote Truth Social, you know, I’m not exactly thrilled.
Trump did say a few weeks ago that he’ll appoint a new AI czar and stand up an AI Force modeled on Space Force. And Trump dismissed safety concerns as a hoax, as he does with pretty much everything in everyday life. And he promised that the White House won’t hinder growth. I mean, you and I agree with that. We don’t want the White House to hinder growth either. But is he missing something crucial? Like, what do you think of Trump’s play here, calling AI [safety fears as] a hoax and saying, you know, he won’t regulate or anything on the federal level?
REID:
Look, I think there are some good, limited regulations that can be done on the federal level. I think the likelihood that you will get any regulations, and then followed by good regulations, is very low in this administration, because I think that, you know, when you look at, you know, kind of freedom coin and planes as, you know, kind of gifts and so forth, like, that whole range of things, more stock trading than the entire Congress combined, and all the rest of the stuff—I don’t think there’s going to be any play here other than things that are kind of self-enrichment. So I don’t think there’s going to be a concern for, well, how do we slow down terrorism, or how do we make businesses safer, or how do we make your average American safer?
All of which would be the kinds of considerations of what you do. You know, I think the previous AI czar was trying to kill every single regulation that, actually, internally, the various cabinet secretaries and so forth were trying to do. So my guess is the new AI czar will probably be that, although, you know, maybe they’ll appoint Buzz Lightyear as the AI czar. And, you know, it’ll be “to infinity and beyond,” you know, as kind of a way of doing this, and that would at least be a friendlier—sure, kinder—one than previous. And so, you know, I think that very little will happen there. Now, you say, well, what kinds of things should happen?
And for example, some of the simple things are, I think, since we have uncertainties, having a kill switch is a good idea. I think having kind of smart auditors looking at the stuff you’re doing. And I think part of what, you know, Dario and Sam have been calling for is, hey, there’s not only the old auditors, which you could still use in various things, but there’s new auditors who have a much more deep technical sophistication that can be kind of reporting on what you’re doing. I think that’s a good thing. I think, for example, an international treaty on cyber would be a very good thing, and apply that to AI. Now, you probably want—most people don’t realize how much we’re in a state of cyber warfare constantly.
Most people don’t realize that because most of it’s economic—well, it’s either anti-economic or anti-government. Neither of those two want reports as to what’s happening. And AI is only going to intensify that, and that’s going to be intensified through open-weight models. Matter of fact, reports I’m seeing are already seeing that intensifying through those. And so I think there’s a whole bunch of good things that could be done in theory. But the likelihood of any delivery in a griftocracy is very low.
Hey, y’all. Masters of Scale Summit returns October 20 to 22 in San Francisco. If you’ve joined me at Summit before, you know how different it feels from other business or tech conferences. Van Jones calls it “the highest concentration of heart with smart you’re ever going to find.” Every element—the lineup of speakers, the curated audience of founders—is designed to spark ideas that transform your business. You’ll see people you know from Possible—Reed Hastings, Amjad Masad, Kara Swisher—take the Summit stage live, alongside leaders like Nobel laureate Maria Ressa, Walmart’s John Furner, and Cloudflare’s Matthew Prince. Be in the room this year and walk away feeling it’s possible to scale your next great idea. Join us. Visit MastersofScale.com/Summit to learn more.
ARIA:
I think many people think that, you know, regulation at the federal level is better than having a patchwork of state regulation. But like you said, if we’re not going to get it at the federal level, I think we are going to see states come forward and say, “We have to regulate this technology.” You mentioned a kill switch. You mentioned outside auditors. And actually, Governor Newsom recently signed an executive order giving state officials until November 16 to recommend whether California should require a kill switch for frontier models, embed these independent auditors in the largest labs, and expand incident reporting. And so, a lot of the things that you’re talking about that you would seem to want—Governor Newsom also vetoed a bill with these, you know, nearly identical provisions, with some differences, in 2024.
And so California is now trying to mandate the thing that Dario asked for. He asked for this kill switch, and yet Newsom vetoed this two years ago. So where do you stand? Are you in favor of this new bill, and why were you potentially opposed to it a few years ago?
REID:
So I was opposed to it a few years ago. There were actually some other provisions in that. Like, for example, you must validate that it has no negative impacts, and that validation was also squishy. Like, well, what negative impacts do you mean? Like, one person might run into something, five people, etc.? Like, and who litigates that, and how is that interpreted? And are you just going to create full-time employment for all lawyers in the state, and maybe in all the states, for a long time, doing this? Like, all of that kind of stuff, which is part of the reason that earlier bill was not a good bill. And I think that’s probably—I don’t think Gavin has probably changed his mind a lot. I think he pays a lot of attention.
He does learn. And he, like—he signed another bill even before this where he’s like, look, data is good, auditing is good, kill switch—obviously a good thing to add in. But in order to not be just massively restrictive on the whole industry, you have to be precise about what kinds of things you’re doing. You have to not make a—“Hey, we’ll just sort out all the ambiguities in the courts” is not the way that one should almost ever do policy, legislation, regulation, etc. And so I think that’s the difference between them.
And I think that part of the—you know, I know from discussions with Gavin that he actually, in fact, regularly calls, like, many of the tech companies and says, “Explain X, explain Y. What’s going on here? Why do you think this?” Etc. Because he’s trying to navigate, in his own way, the thing I was kind of calling for in a dual mandate, which is: How do we be successful at AI? And as governor of California, how do we help, like, California be responsible, but also help all of its businesses and people? And so that question is the thing he’s trying to navigate. And, you know, I think in the previous bill, Anthropic was supportive, but most of the other people weren’t.
I think in this one, more of the other people are like, “Okay, that’s now gotten sharper. That’s now implementable while we continue to develop and deploy services.” And so I think it has a much broader base of support.
ARIA:
I mean, I think in every industry, you know, you always worry about the regulators getting too close to the companies they’re trying to regulate. But I’d say that especially with AI, we need our politicians and our regulators to actually understand the technology. And so for them being in conversation with Dario and Sam and the heads of the other labs, it’s really positive, because we need to get this regulation right.
REID:
Yeah. And, well, and so one, you talk to multiple companies, because the companies are competing with each other and have different points of view and so forth. Then you say, well, but there are some things where the companies as a whole have a collective interest, or a blind spot they’d be missing. And that’s part of the reason why you talk to NGOs and you talk to universities and so on. And you should have all of those conversations. And I know that Gavin is actually having them, because I’ve also seen him show up at Stanford and talk to professors. And, you know, part of how I met Gavin the first time was he’s literally like, “I’m curious about what’s going on in tech investing.”
And he literally came by our offices at Greylock to say, “I want to talk to you about this because I want to learn about what you guys are doing.” And that’s obviously the kind of leader who’s like, “Hey, I want to learn about this and that.” I’m always much more sympathetic, even in cases where I disagree with the leader. It’s like, “I’m trying to learn. I’m trying to understand. I’m not just grandstanding.” And, you know, people will not be surprised: that’s a small number of politicians who have that intellectual curiosity.
ARIA:
Reid, thanks so much for being here today.
REID:
Well, and thank you as well. Always fun.
Possible is produced by Palette Media. It’s hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Shaun Young. Possible is produced by Thanasi Dilos, Katie Sanders, Spencer Strasmore, Yimu Xiu, Aman Suri, Danny Garrison, Trent Barboza, and Tafadzwa Nemarundwe.
ARIA:
Special thanks to Surya Yalamanchili, Saida Sapieva, Ian Alas, Greg Beato, Parth Patil, and Ben Relles.