앤스로픽 CEO 다리오 아모데이가 AI '프런티어 속도 조절' 계획을 발표하고 오픈AI 샘 알트만 등 업계 지도자들이 지지 의사를 보이면서 AI 안전 및 속도 완화 논쟁이 뜨겁다. 그러나 엔비디아 젠슨 황 CEO는 트럼프 대통령의 주장처럼 AI 반발은 '조작'이며 규제가 불필요하다고 반박했다. TechCrunch 팟캐스트에서는 지도자들의 속도 완화 의지가 진지한지, 구체적 세부사항 부족과 자율규제·시장만으로 안전을 담보할 수 있는지를 놓고 토론이 이어졌다.
번역된 본문
애플 팟캐스트에서 듣기 스포티파이에서 듣기 AI 안전과 잠재적 속도 완화를 둘러싼 큰 논쟁이 벌어지고 있습니다. 앤스로픽 CEO 다리오 아모데이가 최근 '프런티어의 속도를 조절하라(pace the frontier)'는 계획을 발표한 반면, 엔비디아 CEO 젠슨 황은 트럼프 대통령의 주장을 공개적으로 받아들이며 AI에 대한 반발은 허위이며 규제는 불필요하다고 말했습니다. 최신 TechCrunch 'Equity' 팟캐스트에서 커스틴 코로섹, 숀 오케인, 그리고 저(앤서니 하)는 아모데이와 오픈AI CEO 샘 알트만 같은 임원들이 정말로 속도를 늦추고 싶어 하는지 논의했습니다. 많은 업계 지도자들이 아모데이의 계획을 지지하는 것처럼 보여 저는 놀랐지만, 숀은 이 계획이 세부사항이 부족해 보인다고 지적했습니다. 커스틴이 기존 규제와 자유시장의 작동만으로 충분한 안전장치가 될 수 있는지 묻자, 숀은 "지금 연방 정부는 광범위하게, 더구나 이 분야에서 구체적으로 규제를 집행하려는 의지가 전혀 없어 보인다"고 답했습니다. 또한 그는 이렇게 덧붙였습니다. "소비자 선택이 이 시장을 움직이는 것 같지도 않다. 이 회사 중 하나가 정말 나쁜 짓을 저질렀을 때 해지 건수 같은 데서 영향을 받을 것 같지 않다는 의미에서 말이다." 편집된 대화 내용을 계속 읽어보세요. 앤서니 하: 숀, 어떻게 생각하나요? 우리가 정말 '프런티어의 속도를 조절'하게 될까요? 숀 오케인: 그 말은 왠지 영화 '포인트 브레이크'에서 나올 법한 것 같다. 그냥 아주 캘리포니아식 표현이다. "형씨, 우리 프런티어 속도를 조절할 거야" 같은. 글쎄, 지난주 이 쇼에서 말했듯이 어떤 형태의 속도 완화도 올 것 같지 않았다. 이유는 대부분 이 회사들이 그런 방식이 통하는 구조가 아니라고 생각하기 때문이다. 그런데 앤스로픽의 다리오와 오픈AI의 샘, 어느 정도는 일론 머스크까지 이런 이야기를 듣고 나니 일주일 전 제 말이 바로 뒤집힌 셈이 됐다. 물론 이들이 그 말을 믿는지 여부는 나중에 다룰 수 있다. 다만 제가 완전히 틀렸다고 생각하지는 않다. 다리오가 블로그 포스트에서 제시한 느슨한 계획과 다른 지도자들이 이야기한 내용에도, 위험과 리스크가 있다고 이들이 생각하는 이유에 대한 주장뿐 아니라 속도를 늦춘다는 것이 무엇을 의미하는지에 대한 구체적 내용이 여전히 많이 빠져 있는 것 같다. 앤서니, 지난주말 이 이슈가 터졌을 때 자세히 읽었으니, 그들이 하려는 말이 무엇인지 이해하는 부분에서 나보다 관대한지 궁금하다. 앤서니: 저도 의문이 있습니다. 저에게 정말 눈에 띄었던 점은 업계의 많은 사람들이 이 계획에 꽤 빠르게 결집한 것처럼 보였다는 사실입니다. 아모데이가 블로그 포스트를 발표하기 전에도 며칠 전부터 다른 사람들이 이 계획의 변형된 버전을 게시하는 것을 보았습니다. 제가 이해하기로, AI 안전 커뮤니티에서 나온 이러한 제안들은 오랫동안 조금씩 고안되어 온 것들이며, 아모데이가 혼자 생각해낸 것이 아닙니다. 그런데 많은 사람들이 이를 수용할 준비가 되어 있어 보였고, 어떤 면에서는 합의가 존재한다는 점이 고무적입니다. 하지만 동시에 이것이 진정한, 의미 있는 속도 완화 조치를 의미할지에는 다소 회의적입니다. 커스틴 코로섹: 우선, 합의는 좋은 일일 수 있습니다. 하지만 카르텔 형성의 시작일 수도 있습니다. 그러니 우리가 여기서 누구에 대해 이야기하고 있는지 분명히 합시다. 우리는 프런티어 AI 연구소들에 대해 이야기하고 있으며, 이는 구체적이고 중요한 사실입니다. 방금 세부사항이 명확하지 않다고 하셨지만, 여기서 일반적인 아이디어는 독립적인 제3자 평가기관이 앤스로픽이나 오픈AI 같은 회사 내부에서 일하며 AI 안전 관행과 사고를 모니터링하는 것이라는 점은 알고 있습니다. 하지만 그 주변에는 많은 모호함이 있습니다. 또한 민주 국가의 주요 AI 기업들이 안전 문제를 협조하는 또 다른 제안도 있습니다.
Listen on Apple Podcasts Listen on Spotify We’re seeing a big debate over AI safety and a potential slowdown, as Anthropic CEO Dario Amodei recently published a plan to “pace the frontier,” while Nvidia CEO Jensen Huang has publicly echoed President Donald Trump’s claims that the AI backlash is a hoax and regulation is unnecessary. On the latest episode of TechCrunch’s Equity podcast , Kirsten Korosec, Sean O’Kane, and I discussed whether executives like Amodei and OpenAI CEO Sam Altman are serious about wanting to slow down. While I was surprised at how many industry leaders seemed to throw their support behind Amodei’s plan, Sean noted that the plan seemed short on details. And when Kristen asked whether a combination of existing regulation and the workings of the free market could provide enough safeguards on their own, Sean replied, “We have a federal government at the moment that certainly does not seem eager to enforce regulations broadly speaking, let alone specifically speaking here.” Plus, he argued, “There does not seem like there is a ton of consumer choice driving this market, in the sense of, okay, if one of these companies does something really bad, then they will see an impact in how many cancellations there are or whatever.” Keep reading for a preview of our conversation , edited for length and clarity. Anthony Ha: Sean, what do you think? Are we actually going to”‘pace the frontier”? Sean O'Kane: That sounds, for some reason, like something from “Point Break.” It's just a very California statement, like, “We're gonna pace the frontier, bruh.” I don't know. I said on this show last week , I didn't think we were headed for a slowdown of any kind, in large part because I don't think these companies are structured in a way where that works. So this was a pretty immediate flip of what I said a week ago, to hear this from Dario at Anthropic and Sam at OpenAI and even Elon Musk to a certain degree — and we can get into whether we believe that or not. I will say, I don't think I was completely wrong, because even with these loose plans that Dario laid out in a blog post and and some of the other leaders have talked about, I feel like we're still missing a lot of specifics here, not only on some of the claims about why these people think there are dangers and risks, but also specifics on what they mean by slowing down. Anthony, I know you read this stuff pretty closely last weekend as it was all popping off, so I'm curious if you're maybe more generous than I am, as far as understanding what it is they're trying to say here. Anthony: I mean, I also have questions. What really stood out to me is the fact that so many people in the industry seemed to coalesce around this fairly quickly. Even before Amodei published his blog post, I was seeing other people post some version of this [plan,] just in the days leading up to it. My understanding is, these suggestions that have come out of the AI safety community; these are things that have been percolating for a while. They're not just things that [Amodei] thought up on his own. But it just seemed like a lot of people were kind of ready to embrace this and in some ways, [it’s] encouraging to see that there's a consensus. But it also makes me a little skeptical that this would represent genuine, major steps towards a slowdown. Kirsten Korosec: First of all, consensus can be a good thing. Also, it can be the beginnings of forming a cartel. And so, let's be clear [about] who we're talking about here. We're talking about frontier AI labs. and that's specific and important. You did say a minute ago that the details aren't really clear, but we do know that the general idea here is that independent third party evaluators would come in that would work inside [companies] like Anthropic or OpenAI and monitor AI safety practices and incidents. But there is a lot of vagueness around that. Then there's another proposal to have major AI companies in democratic countries coordinate on safety standards and limits. And then this international coordination piece. So those are the general ideas around this. I'll also add that there's been very considerable pushback from very specific people. And surprisingly, some people I didn't think would push back on this. The obvious one, I guess, would be Nvidia’s CEO Jensen Huang, who had some pretty choice words. I don't know what either of you think of […] some of his comments and the call he took from President Trump on-stage [at the All-In Summit] Sean: I think Jensen understands that, to a certain extent, he is seen as the adult in the room here. Which is interesting, because that's a dynamic that I think has changed. If you were to ask people a year, almost two years ago now, who was playing that role as these AI companies were really starting to blow up in big ways, I think a lot of people would have pointed to Microsoft and Satya Nadella, and I feel like Microsoft has — for a number of reasons that we don't need to get into right now — lost that grip on being the adult in the room. Certainly in part because Nadella said he was going to make Google dance , and then that never happened. and in fact, the opposite happened, in a lot of ways. Jensen, I think, understands not only that he needs to be pleasing the most amount of people and so he wants to say the most measured thing, he also is one of the most useful liaisons between this field and this administration, to the point that he's getting a call from the president while he's on-stage at this very well-attended and high-profile conference that allowed both of them to respond to all this mess. Kirsten: [To] your point about him being the adult in the room and the liaison, he is in a position where he could do that. But in my view, I would say that [his appearance,] which did feel a little staged, went a step too far and really highlighted the fact that Nvidia greatly benefits from AI moving forward in an unpaced manner. I hate to be cynical about it, but that was sort of my initial reaction to it. Anthony: Trump’s and Huang's interests are aligned, and also [Huang’s] telling [Trump] what he wants to hear. I did think some of the things he was specifically focusing on are worth unpacking a little bit. First he talked about this idea of, “We are not going to have a slowdown.” It’s interesting, just that phrase “slowdown,” which has come up a lot — but at the same time, both Altman and Amade [will] use the slowdown language, but the word they like to use more is “pace.” right? And these three proposals — the ones you just described, Kirsten — in theory, they could result in things slowing down, but they're not saying that explicitly. It's saying, “Hey, we're making some basic safety measures and having some people keep an eye on us.” But none of those things are intrinsically saying that things have to go slower than they are now, at least as I understand it. Kirsten: So I have a question for Sean. Couldn't you argue in a free market society, we have existing regulations in place, and also [in] the competitive environment in which there isn't protectionism, if there is something unsafe that is released by a company, that they would no longer be able to be a viable company? And there are regulators in place today that could go after these companies? Or do you reject that idea? Sean: In a vacuum, or in a market that is not being distorted by a whole bunch of different external pressures, yeah, I think there's an element of that. But we have a federal government at the moment that certainly does not seem eager to enforce regulations broadly speaking, let alone specifically speaking here. And we're also in this weird situation where — I don't really know the right way to put it, but it does not seem like there is a ton of consumer choice driving this market, in the sense of, okay, if one of these companies does something really bad, then they will see an impact in how many cancellations there are or whatever. Especially now,