# dario-amodei-abc-news-2026-133753620
- Transcript ID: `ebd7aa59fa103ac8`
- Source tier: `official_webvtt`
- Original language: `en_or_source_language_unlabeled`
- Translation status: `agent_translated_reviewed` (20/20 chunks)
- Original source: https://abcnews.com/video/133753620/
- Raw evidence: `data/transcripts/web/dario-amodei-abc-news-2026-133753620.json`
> 中文是派生翻译，英文是留存的源文本。中文未复核时，以英文和原始音视频为准。
## 00:00:00
**中文**
人工智能巨头 Anthropic 以其聊天机器人 Claude 闻名。他们将坐下来面对——或者说，他们正处在 AI 军备竞赛的最前沿。该公司的首席执行官兼联合创始人达里奥·阿莫代伊（Dario Amodei）发出新的警告，称 AI 的进步速度及其对经济和国家安全构成的潜在风险令人担忧。
**English / Original**
Artificial intelligence giant Anthropic is known for its chatbot, Claude. They're going to sit down in front of the-- or they're sitting down with the front of the AIs arm race. The company's CEO and co-founder, Dario Amodei, is out with a new warning about the speed of AI's advance and potential risks it poses to the economy and national security.
## 00:00:19
**中文**
在接受本台林赛·戴维斯（Linsey Davis）独家采访时，阿莫代伊解释了 AI 公司为何应考虑放慢脚步。非常感谢您接受采访。当然，您今天公布了多项提案，描述了一种 AI 未来及其发展可能导致的多种失业情景：该国失业率可能从 5% 到 10%，甚至达到我们前所未见的水平。
**English / Original**
Sitting down exclusively with our Linsey Davis, Amodei explains why AI companies should consider slowing down. Thank you so much for the conversation. Of course, you've announced a number of proposals today describing an AI future and the development that could potentially lead to a number of scenarios for unemployment in this country, anywhere from 5% to 10% to levels that we have never seen before.
## 00:00:46
**中文**
请为我们解释这些情景，以及您认为最佳的解决方案是什么。首先要明确的是，我认为没有人知道将会发生什么。我们只看到，AI 是一项非常强大的技术。我认为它比以往的技术具备更多通用认知能力。它发展得很快。因此，尽管面对以往技术时，人们已作出适应、就业市场也已恢复，我们至少应考虑一种可能性：在这个案例中，这可能更具挑战性；如果没有干预，恢复未必会像过去那样充分发生。
**English / Original**
Explain those scenarios for us and what you think are the best solutions. So to be clear, I think no one knows what's going to happen. We just see that AI is a very powerful technology. It has more, I think, general cognitive capabilities than previous technologies. It's moving fast. And so while with previous technologies, things have adapted and the job market has recovered, we should at least consider the possibility that that may be more challenging in this case, that it may not happen as much without intervention.
## 00:01:16
**中文**
所以，我们大致列出了在各种情景下会采取的做法。有一种大致正常、照常营业的情景，基本不需要做什么。还有一种情景是，失业率升至 5%。我认为届时适当的应对措施包括：工资再保险等理念、鼓励公司不裁员的留任激励措施，以及改进失业保险。
**English / Original**
And so we kind of laid out what we would do in various scenarios. There's kind of a normal business as usual scenario where nothing much is required. There's a scenario where, yeah, unemployment goes to 5%. And then I think the appropriate measures to take in response there are things like, there are these concepts around wage reinsurance, retention incentives to encourage companies not to lay off people, improvements to unemployment insurance.
## 00:01:47
**中文**
这些措施旨在帮助人们度过一个颠簸但相对常见的转型期。如果情况变得更极端、恢复所需时间更长，那么当然，我们应设法帮助人们找到新工作，也应设法利用巨大的繁荣，让每个人都更富裕。但我认为与此同时，我们需要讨论经济支持。
**English / Original**
These are things to help people through a bumpy but relatively usual transition. If things go more extreme and they take longer to recover, then of course, we should look for ways for people to find new jobs, look for ways to use the enormous prosperity to help everyone be more wealthy. But I think in the meantime, we need to talk about economic support.
## 00:02:09
**中文**
我们可以设想提高资本利得税之类的措施，也可以设想对 AI 公司征税，或提高一般的企业税；还可以探索通过某种普惠资本账户，让人们分享繁荣的方式。这就是一系列可能的措施。并不是说我预测这些事情一定会发生。
**English / Original**
And we could imagine things like raising the capital gains tax. Imagine things like taxes on AI companies or the corporate tax in general, looking at ways to let kind of universal capital accounts to allow people to share in the prosperity. So that's the range of things. And it's not that I'm predicting these things will necessarily happen.
## 00:02:33
**中文**
我预测的是我们应当做好准备。您认为眼下——无论是社会还是政府——为应对这些紧迫挑战，最应当做的事情是什么？如果问我们今天、此刻需要做什么，那就是收集更好的数据，了解正在发生什么。围绕 AI 目前对劳动力市场的影响以及未来会发生什么，已经有很多争论。
**English / Original**
I'm predicting that we should be ready. What do you think is the most urgent thing that we ought to be doing right now-- and that could be society. That could be government-- to address these urgent challenges. If you ask the thing we need to do right now, today, we need to collect better data on what's happening. There's been a lot of debate about what's happening in the labor market with AI now and what will happen in the future.
## 00:02:57
**中文**
我主要担忧未来，但通过观察当下，你能看到未来的早期迹象。我认为，如果我们能就正在发生什么、AI 对就业的影响朝哪个方向发展达成共识，那么我们就会处于更有利的行动位置。杰弗里·辛顿（Geoffrey Hinton）通常被认为是人工智能的教父之一，他最近谈到 Anthropic 时说——我相信您听过这段引述——Anthropic 正在“从主要关心发展 AI 安全，转向赢得 AI 竞赛”。
**English / Original**
I'm mostly worried about the future, but you see the early signs of the future by looking now. And I think if we can get agreement from everyone about what is happening, about what direction the job effects of AI go, then I think we're in a better position to act. Geoffrey Hinton, who is largely considered to be one of the godfathers of artificial intelligence, recently said of Anthropic-- and I'm sure you've heard this quote-- that Anthropic is "veering away from primarily caring about developing AI safety to winning the AI race."
## 00:03:29
**中文**
您如何回应这种说法？事实上，我没有听过这段引述，但我不认为它属实。我完全不知道他会以什么为依据——我不知道他的观点究竟基于什么。我的意思是，三个月前，我们刚刚顶住了战争部的压力，因为我们担心以可能破坏民主的方式使用模型。
**English / Original**
How do you respond to that? So I actually have not heard this quote, but I don't think this is true. And I don't know what he would base his-- I don't know what he would base his views on whatsoever. I mean, three months ago, we just stood up to the Department of War because we were concerned about using models in ways that might undermine democracy.
## 00:03:52
**中文**
因此，该部门对我们采取了法律行动，而我们顶住了。我们最近发布了一款在生物和网络安全防护方面非常强的模型。我们与不同合作伙伴及政府合作，修补网络防御系统，帮助模型做得更好。我们昨天发布的模型 Fable，是我们迄今发布过的最具对齐性的模型之一。
**English / Original**
Legal action was taken against us by the department because of that. And we stood up to that. We recently just released a model with very strong bio and cyber safeguards. We worked with various partners and with the government to patch cyber defenders and help the models do better. The model we released yesterday, Fable, is one of the most aligned models that we've ever released.
## 00:04:15
**中文**
我们一直持续开展 AI 安全和 AI 对齐方面的工作。也许他只是看到了 Anthropic 融到大量资金，或 Anthropic 向美国证券交易委员会（SEC）提交备案的新闻标题。但请看，我认为我们不应将金融事件或市场行为等同于 Anthropic 不再优先重视安全。我们一贯认为，安全与商业活动彼此互补，而非相互对立。
**English / Original**
We continue to produce AI safety and AI alignment work constantly. Now, perhaps he had just read the headlines about Anthropic raising a lot of money or Anthropic being on file with the SEC. But look, I don't think we should equate financial events or action in the market with Anthropic not prioritizing safety. We've always seen safety and commercial activity as complementary to each other, not against each other.
## 00:04:42
**中文**
您提到了 Anthropic 与战争部或五角大楼、特朗普政府之间的那场法律争端。还有报道称，Anthropic 目前也在与国家安全局（NSA）合作。Anthropic 与特朗普政府的关系目前处于什么状态？是的，我不能谈论仍在进行的——我不能谈论仍在进行的法律案件。
**English / Original**
You mentioned that legal battle between Anthropic and the Department of War or Pentagon, the Trump administration. There are also reports that Anthropic is also working with the NSA at this time. What is the status of Anthropic's relationship with the Trump administration? Yes, so I can't speak to ongoing. I can't speak to ongoing legal cases.
## 00:05:03
**中文**
你知道吗？请看，请看，我能说的是：对我们而言，这始终是关于在我们的红线范围内，我们很高兴为美国国家安全服务。我们与 NSA 所做的事情也在我们的红线范围内。那您与特朗普总统的关系如何？我和特朗普总统谈过几次。
**English / Original**
You know what? Look, look, what I can say is, for us, this has always been about we've been excited to serve US national security within the bounds of our red lines. And what we're doing with the NSA is within the bounds of our red lines. And what's your relationship like with President Trump? I've spoken to President Trump a couple of times.
## 00:05:23
**中文**
我始终强调，我们正尽最大努力确保美国在 AI 领域领先；我们正努力——我们正努力推动经济增长；我们是爱国的美国人。Anthropic 最近曾建议暂停，因为 AI 已接近能够自我发展的门槛。这种暂停将如何运作，您支持这种暂停吗？
**English / Original**
I always emphasize that we're trying to do the best we can to ensure that America leads in AI, that we're trying to-- that we're trying to grow the economy, that we're patriotic Americans. Anthropic has recently suggested taking a pause because AI is on the cusp of being able to develop itself. How would that pause work, and do you support that pause?
## 00:05:53
**中文**
不，不。我想澄清一下。我们在一篇博客文章中所说的是：如果世界各地所有开发 AI 的人能够达成一致，并能建立可验证的机制来展示各自的 AI 进展速度，那么暂停、或者更可能是放慢 AI 进步的速度，或许是有道理的。
**English / Original**
No, no. I would like to clarify. So what we said in one of our blog posts was that we believed it would be a good thing if all the people developing AI around the world were able to agree and were able to develop verifiable mechanisms for showing their pace of AI progress, then it might make sense to pause or more likely, slow the pace of AI progress.
## 00:06:22
**中文**
AI 的进步速度非常快。我们已经谈过就业影响。我认为，如果我们稍微放慢一点，社会可以更好地适应。我也认为，正如我们就 Mythos 模型和 Glasswing 所讨论的那样，我们看到网络安全方面的风险。很快，我认为我们还会看到生物技术被滥用的风险，以及模型具有自主性、以人类未预期的方式行事的风险。
**English / Original**
The pace of AI progress is very fast. We've talked about the employment effects. I think society can adapt better if we slow things down a little bit. I also think we see risks around cyber, as we've discussed with our Mythos model and Glasswing. We're soon, I think, going to see risks around misuse of biology and around the models being autonomous and behaving in ways that humans don't expect them to behave.
## 00:06:51
**中文**
所以，如果我们有更多时间，我认为我们可以把所有这些风险处理得更好一些。现在，我认为现实中要让许多参与者共同放慢进步速度会非常困难。除了其他 AI 实验室外，还必须包括中国这样的参与者。因此，我对我们能否获得真正严格的护栏、并让它奏效持一定怀疑态度；但我的观点是，我们至少应愿意尝试。
**English / Original**
So if we had more time, I think we could address all of these risks a little better. Now, I think in reality, it's going to be really hard to get a lot of players to work together to slow down progress. It would have to be players like China in addition to the other AI labs. So I'm somewhat skeptical that we can get really serious guardrails and that we can make it work, but my point is that we should be willing to at least try.
## 00:07:18
**中文**
那么，明确一下：您会主张暂停，只是不确定实际该如何实施。必须是所有相关的 AI 参与者，包括地缘政治对手，都同意，而且它必须可验证。面对人工智能整体问题时，您如何确保中国等国家会成为善意、负责任的参与者？
**English / Original**
So just to be clear, you would advocate a pause. You're just not sure how it would actually be implemented. It would have to be the case that all the relevant AI players, including geopolitical adversaries, would have to agree to it, and it would have to be verifiable. How do you account for countries such as China to be good actors, responsible citizens when it comes to artificial intelligence overall?
## 00:07:49
**中文**
如果是我们在 AI 领域领先——而我认为我们应继续领先——那么就更容易让我们的对手同意至少采取某种最低程度的克制。我完全不信任中国。我不认为我们能达成很多共识。但例如，我可以设想提出这样一项提议：双方同意不使用 AI 来开发生物武器。
**English / Original**
If we are the ones who are leading in AI, which I think we should continue to do, then it is easier to get our adversaries to agree to at least some minimal amount of restraint. I don't trust China at all. I don't think we're going to get a lot. But for example, I could imagine proposing something around both sides agreeing not to use AI to develop biological weapons.
## 00:08:14
**中文**
我可以设想，这可能是现实可行的。
**English / Original**
That's something I could imagine could be realistic.