Anthropic's Fable Backlash, Nationalizing AI, Inflation Heats Up & California's Broken Elections

All right, everybody. 好了，各位。 Welcome back to the number one podcast in the world. 欢迎回到全球第一播客。 It's your favorite podcast. 这是你最爱的播客。 It's your podcasters's favorite podcast. 这是播客人最爱的播客。 It's the podcast that everybody likes to rip off for 50 episodes and then they quit. 这个播客，人人都想抄，抄了50集就放弃了。 But the All-In podcast is not quitting. 但 All-In Podcast 不会放弃。 We're doubling down with the original quartet. 我们要加倍押注，原班人马回归。 Everybody's here for a big big week, lots of debates, and we got to start with Anthropic. 大家都来了，这周大事不断，好戏多的是，我们得先从 Anthropic 说起。 Anthropic released a mythos level model with some interesting guardrails and uh seems as though Daario is back to blogging. Anthropic 发布了一个神话级别的模型，带了一些有意思的护栏，而且 Dario 似乎又开始写博客了。 Okay, the model's called Fable 5. 好，这个模型叫 Fable 5。 No idea why they're calling it that or released on Tuesday and it tops every benchmark nearly every benchmark. 不知道为什么起这个名字，周二发布的，几乎刷新了所有基准测试。 The tokens, however, cost twice as much as Opus 4.8, but it should use less tokens overall because it's theoretically that much better. 不过 token 成本是 Opus 4.8 的两倍，但理论上因为模型更强，整体用的 token 会更少。 Unclear if this is going to alleviate some of the hand ringing around the cost of tokens, but you remember in April Anthropic famously did not release Mythos because publicly uh it had some strong hacking capabilities. 这能不能缓解大家对 token 成本的焦虑还不好说，不过你们记得吧，四月份 Anthropic 没有发布 Mythos，原因是公开的说法是它有很强的黑客攻击能力。 So they gave it to some folks and we had um the CEO of Nesh of Palo Alto Networks who said, "Hey, Mythos was the real deal." 所以他们把它给了一些人试用，其中有 Palo Alto Networks 旗下 Nesh 的 CEO，他说：「Mythos 确实名副其实。」 and that they used it to seal up some vulnerabilities inside their shop. 他们用它修补了内部的一些安全漏洞。 Obviously, this new model is sensitive uh to topics like boweapons, hacking, all that stuff are blocked in it. 显然，这个新模型对生化武器、黑客攻击这类话题都有屏蔽。 But they got a big developer backlash and then I'll get your feedback after I explain this. 但他们遭到了开发者的强烈反弹，说完这些我再听你们的看法。 While using Fable, Anthropic stores all the prompt data you enter for at least 30 days. 在使用 Fable 的过程中，Anthropic 会把你输入的所有 prompt 数据至少保留 30 天。 And so that's a bit of a privacy issue. 这多少是个隐私问题。 And if Fable 5 detects you're doing frontier AI research, in other words, using their model to make a better model, competitive model, they were downgrading you, but they weren't telling you. 如果 Fable 5 检测到你在做前沿 AI 研究，也就是用他们的模型来开发更好的、有竞争力的模型，他们会悄悄降级你，但不会告诉你。 And this was buried deep in a 319 page document. 这条规定藏在一份 319 页的文件深处。 So many kuruffles, freakouts going on on X over this, but they've since walked it back a bit, quote unquote, in Wired. X 上为此炸锅了，但他们随后在《连线》杂志上「稍微」收回了一些立场。 We're changing Fable 5 safeguards for Frontier LLM development to make them more visible. 我们正在修改 Fable 5 针对前沿 LLM 开发的安全限制，让它们更加透明。 I'm going to stop there. 我先说到这。 Chimath, what's your take on anthropic and these extraordinary models? Chamath，你怎么看 Anthropic 以及这些超强模型？ They keep dropping and how they're handling the release of them. 它们不断推出，他们处理发布的方式你怎么评价？ Are they being thoughtful? 他们是在认真考量吗？ Are they being dramatic and drama queens? 还是在搞戏剧性的危机公关？ A little bit of both. 两者都有一点。 Where do you stand on it? 你怎么看？ You know, after getting Nesh's feedback and seeing the latest model come out, have you played with it? 你得到了 Nesh 的反馈、也看到了最新模型发布，你有没有亲自试用过？ What's going on at 8090 and the software factory in terms of benchmarking it as well? 8090 和软件工厂那边在基准测试方面进展如何？ It's a really incredible model and so kudos to these guys for continuing to push the boundaries of the closed frontier lab models. 这是一个非常了不起的模型，向这些人致敬，他们持续推动着闭源前沿实验室模型的边界。 They're firing on all cylinders. 他们全力以赴。 I think that it creates a pretty obvious risk now and I think that obvious risk is twofold. 我觉得这现在带来了一个相当明显的风险，而且这个风险是双重的。 One is Anthropic has essentially shown their hand which is that they will increasingly take in prompts, evaluate the prompts and decide what to do with them before they generate output to you. 一是 Anthropic 基本上已经摊牌了：他们会越来越多地接收 prompt、评估 prompt，然后决定在生成输出之前对其做什么处理。 And I think if you were a person, you should generally now think there's a risk of censorship. 我觉得如果你是个人用户，现在大体上应该意识到存在审查风险。 If you're a company, I think it's almost a non-starter. 如果你是企业，我觉得这几乎是不可接受的。 And the reason is because you could accidentally trip one of these things without even knowing it. 原因是，你可能在毫不知情的情况下触发其中某条规则。 A downstream scientist using the cloud APIs could trip it. 你公司里使用 Claude API 的下游科学家可能触发它。 A business executive inside your corporation could trip it. 你企业内部的业务主管可能触发它。 A person doing scientific molecular research could trip it. 从事科学分子研究的人可能触发它。 And all of a sudden, you'll get cut off from a very important source of business differentiation for yourself. 然后你突然就被切断了一个对你非常重要的业务差异化来源的访问权限。 So I think if you take both of those two things together, we're at this very unique moment in time where I think companies need to start underwriting this next phase of AI, which is how do I have control? 所以我觉得如果把这两件事合在一起看，我们正处于一个非常特殊的时刻：我认为企业现在需要开始为下一阶段的 AI 做规划，也就是，我如何拥有掌控权？ Who am I allowing to learn off of all of this information? 我要允许谁从这些信息中学习？ Do I want to have single point of failure risk with respect to AI? 我想不想在 AI 上存在单点故障的风险？ And I think the answer is that you need broad diversity and a governance approach that's better managed. 我觉得答案是：你需要广泛的多样性，以及一个管理得更好的治理方式。 One thing I'll say about anthropic 有一件事我想说关于 Anthropic， is they tell the truth. 就是他们说实话。 Yes. 对。 It's just that the truth sucks when you actually take it and you eat it and you're like it's in your tummy and you're like no this is not good. 只是当你真正把这个实话消化掉，放进肚子里之后，你会觉得：不对，这不好。 I don't like this. 我不喜欢这个。 And so there's the censorship risk on the one hand and then there's just the governance business risk for enterprises on the other. 所以一方面是审查风险，另一方面是对企业来说的治理与业务风险。 Both are not good. 两个都很糟糕。 Freeberg running your own company now. Friedberg，你现在自己经营公司。 Uh, and I'll let you back clean up here. 我先让你补充一下。 Saxs, obviously you'll have a lot to say, but Freeberg, running your company, do you worry about getting rugpulled by one of these companies, investing your time into one of these platforms, having them then constrain your use of it and or take what you're doing at Ohio, put it into their model and make it available to other people. Sacks，你肯定有很多要说的，但 Friedberg，你自己运营公司，你担心被这些公司突然断供吗？你把时间投入到某个平台，然后他们限制你的使用权，或者把你在 Ohalo 做的东西放进他们的模型，再开放给其他人。 What 你…… What concerns do you have as a business owner at the forefront of your field using these tools? 作为站在领域前沿使用这些工具的企业主，你有哪些担忧？ It's a great question. 这是个好问题。 the the terms of service is pretty clear that they can't take what you're doing and put it into their model that they won't 服务条款说得很清楚，他们不能拿你的数据来训练他们的模型，他们不会这么做。 do you believe them when you read that I 你看到这条条款会相信他们吗，我…… I'm not sure to be honest there are things I'm concerned about but we've decided to throw caution to the wind because we do a lot of proprietary work in genomics what that means is we're looking at different genes or gene variants trying to estimate the impact that that gene or gene variant might have on a living organism like we work in plants obviously in agriculture 说实话我不确定，有些事情确实让我担心，但我们还是决定放手去做，因为我们在基因组学方面有大量专有工作，具体来说就是我们研究不同基因或基因变体，试图估算该基因或基因变体对活体生物的影响，比如我们主要做植物，也就是农业领域。 And we will do things like RNA guide design for gene editing based tools. 我们会做 RNA 引导设计，用于基因编辑工具。 We will do predictions on what gene may or may not have some phenotype. 我们会预测某个基因是否可能影响某个表型。 So there's a lot of genomic modeling work that we have found these models to be incredibly valuable at supporting us with over the last couple of years. 过去几年里，我们发现这些模型在基因组建模方面对我们的支持非常有价值。 So going back about 6 months, we were able to very simply and cleanly do things like design a genetic construct that you would then use to make a specific protein and the the tools was very easy to use to do that sort of research and extremely valuable. 大约六个月前，我们还能非常简单、干净地完成诸如设计基因构建体这类工作，然后用它来制造特定蛋白质，工具用起来非常顺手，极具价值。 This is the kind of work that many scientists would spend a lot of time doing and these models were very good at doing very quickly. 这类工作很多科学家会花大量时间做，而这些模型能很快完成。 Over the last couple of weeks, they've begun to restrict the ability to use the models to do that work. 但过去几周，他们开始限制使用这些模型来做这类工作的能力。 And the premise is that there's some sort of bioweapon uh type risk that folks have theorized could happen using these tools. 理由是有人从理论上认为，这些工具可能存在某种生化武器风险。 And as a result, we're losing the capacity to use these models for this very important scientific development work. 结果我们正在失去用这些模型做这项重要科学开发工作的能力。 As a result, we are likely going to end up needing to use open- source models and run them locally ourselves. 因此，我们很可能最终需要转向开源模型，在本地自己运行。 So the reason I walk through all of that is so people can really understand the context of what's going to happen here. 我说这些是为了让大家真正理解接下来会发生什么。 As folks like Anthropic say, "Hey, we're going to restrict access or censor the output of these models." 当像 Anthropic 这样的公司说「我们要限制访问或审查这些模型的输出」， It is going to force companies like ourselves who still want to take advantage of the capability of these LLMs to go and get open- source tools and run them. 就会迫使我们这样的公司，这些仍然想利用 LLM 能力的公司，去获取开源工具自己运行。 And what are the best open source models today? 那今天最好的开源模型是什么？ They're Chinese. 是中国的。 Yeah, they are. 是啊，没错。 And that is a major concern. 这是一个重大隐患。 the the American open source models are not as good as the Chinese open source models. 美国的开源模型质量不如中国的开源模型。 So the restrictions that Anthropic and others are putting upon themselves and upon the industry is forcing a lot of companies to go and get open source Chinese models and run them. 所以 Anthropic 等公司对自己、对行业施加的这些限制，正在迫使大量公司去使用中国开源模型并自己运行。 We're seeing this across the landscape with startups with large scale enterprises. 我们看到这种情况在初创公司、大型企业中普遍发生。 Everyone's making that move. 所有人都在走这条路。 Yeah. 嗯。 And this is something we've been talking about here with Apple and their silicon and how well you can run these models. 这也是我们一直在聊的事，关于 Apple 芯片以及这些模型在上面跑得有多好。 Also, I predict your next card you'll turn over is you'll start making your own models, Freeberg. 另外，我预测你下一步要翻的牌是：你会开始做自己的模型，Friedberg。 You'll take all this data you have. 你会用你手头的所有数据。 It's exactly right. 完全正确。 And we'll we'll start we'll start with the core model, combine it with our data, and then we'll have our own genome language model, our own prediction model that we'll then use internally. 我们会从核心模型开始，结合我们的数据，然后有我们自己的基因组语言模型、我们自己的预测模型，再用于内部。 And I think that's where folks are going. 我觉得这就是大家要走的方向。 But the reason I point this out is because a lot of folks are feeling the pressure from anthropic doing this. 我指出这一点，是因为很多人正在感受到 Anthropic 这么做带来的压力。 and they're feeling the pressure from politicians repeating the scary words that are being said by Daario and others. 他们也在感受到政客们不断重复 Dario 等人所说的那些恐怖措辞带来的压力。 And as a result, they are going to try and they are already trying to force either natural enforcement or politicized enforcement upon the model providers in a way that is ultimately going to benefit Chinese open source model providers. 结果，他们正在尝试，并且已经在尝试，以某种方式对模型提供商施加强制执法或政治化执法，而这最终将有利于中国开源模型提供商。 And that is a scary thing because we are going to damage our own kind of economic viability. 这很可怕，因为我们将损害自己的经济竞争力。 You can't just stop AI. 你没办法直接叫停 AI。 As much as everyone says AI is doomsday, by stopping AI or trying to stop AI through this political action and social, you know, kind of behavior, you are fundamentally going to give someone else the advantage because the AI isn't going to go away. 无论多少人说 AI 是末日，通过政治行动和社会层面的所谓行为来叫停 AI，你实际上是在从根本上拱手让人，因为 AI 不会就此消失。 So the reason I describe what we're doing is for everyone to really understand in Grock that you can't just turn off AI and turn off access to these things. 所以我描述我们在做的事，是为了让每个人都能真正深入理解：你没法关掉 AI，没法关掉对这些东西的访问。 What you will do is you will force the hand of someone else to now have an unfair advantage because you're unfairly restricting your models and your access to those models. 你做的只是逼别人拿到不公平的优势，因为你在不公平地限制自己的模型和访问权限。 Well, well, really well said, Freeberg Saxs. 说得太好了，Friedberg。Sacks， What's your take just in terms of the chessboard in terms of we got the business case here, we got the local case, but there's a lot of issues here with Anthropic in terms of them essentially throwing up every red flag to get regulated to induce and it'll be our second story, the Bernie Sanders seizing the equity of these companies. 你怎么看这盘棋局？我们有商业层面的问题，有本地化层面的问题，但 Anthropic 这边还有很多麻烦：他们基本上在竖起所有红旗要求被监管，而且会引出我们的第二个话题，就是 Bernie Sanders 要没收这些公司的股权。