The funny thing about Anthropic haters is that they still mostly believe Anthropic’s marketing. They think Claude is a recursively self-improving silicon God, and that we are all a couple refusals away from falling into the perpetual underclass. This gives them way more power than they deserve.

Of course they believe they should dictate morality to you and control the future light cone of the universe and everything they do is justified for the mission and so on and so forth. And yea, the model is pretty good, so you worry that if the model is good that they will get their way about other things.

A lot of people in history have had dumb ideas about ruling the world, doesn’t mean they can actually do it. They are a product of the rationalist cult (which surprise surprise, thinks the world is going to end soon but advocates for polyamory in the mean time). There’s some short-term dangers from them, but long-term very little. Remember when SBF wanted to regulatory capture all of crypto? And similar to Claude, FTX was a successful product! But the totalizing nature of the ideology contains its downfall.

The tech to make models is way more commodity than previous generations of technology, the biggest factor is mostly if you are willing to spend money on compute and data. We should be grateful that their ideology pushes them to train way bigger models than is probably economically rational. (also, always remember that most all American companies have Chinese spies in them, so while they don’t publish their research, at least it’s making its way into the open models eventually)

Once you reject this premise, the whole drama deflates into a much more reasonable question about who will capture margin in a commoditizing industry. Here’s my prediction for what actually happens.

Imagine a tractor replacing a team of people to dig holes. At first, finance people are stupid, and think that the size of the tractor market will be the size of the hole digging market. Instead of the money going to the laborers, it will instead go to the tractor companies. But the fallacy here is that the price of the labor is related to its value. In reality, unless there’s some insane price fixing, it’s set by the real cost in a competitive tractor industry.

So if holes become 10x cheaper to dig, the hole digging market size in dollars becomes 10x smaller. Maybe some of it is offset by digging more holes now that the cost is less, but it’s unlikely to immediately 10x hole demand, since there really are only so many uses for holes.

I think this is basically what it will be for most knowledge work, knowledge workers are so grossly overpaid compared to the energy they consume, and AI will rectify that. This explains why the Chinese are giving the (much more moderate resources to train) models away for free. They love to see deflationary economics in the US. Even if you regulatory capture the US government, nobody is getting a monopoly on AI, we don’t live in a unipolar world anymore.

AI will be cheap, everywhere, collapse a bunch of sectors of work, wreck wage premiums, and cause a big reevaluation of status hierarchies. Who else is excited for Great Depression 2: Electric Boogaloo?

关于Anthropic的批评者，有趣的是他们大部分仍然相信Anthropic的营销。他们认为Claude是一个递归自我改进的硅神，而我们所有人都离沦落为永久底层阶级只差几次拒绝。这给了他们远超应得的权力。

当然，他们相信他们应该向你发号施令，控制宇宙的未来光锥，他们所做的一切都是为了使命，等等等等。是的，这个模型确实很好，所以你担心如果模型好，他们就会在其他事情上如愿以偿。

历史上有很多人有过统治世界的愚蠢想法，并不意味着他们真的能做到。他们是理性主义教派的产物（巧了，这个教派认为世界很快就会终结，但同时提倡多元恋爱）。他们有一些短期危险，但长期来看非常少。还记得SBF想要监管控制所有加密货币吗？和Claude类似，FTX也是一个成功的产品！但极权性质包含了其自身的衰落。

制造模型的技术比前几代技术更加商品化，最大的因素基本上是你是否愿意花钱买算力和数据。我们应该感激，他们的意识形态促使他们训练比经济上合理规模大得多的模型。（另外，永远记住，大多数美国公司都有中国间谍，所以虽然他们没有发表他们的研究，但至少最终会进入开源模型。）

一旦你拒绝这个前提，整个戏剧就缩小为一个更合理的问题：在商品化行业中谁能赚取利润。以下是我对实际发生情况的预测。

想象一台拖拉机取代一群人挖洞。起初，金融人士很愚蠢，认为拖拉机市场的规模将等于挖洞市场的规模。钱将不是流向劳工，而是流向拖拉机公司。但这里的谬误是，劳动力的价格与其价值相关。实际上，除非存在某种疯狂的定价垄断，否则在竞争性的拖拉机行业中，价格是由实际成本决定的。

所以，如果挖洞成本降低10倍，以美元计的挖洞市场规模就会缩小10倍。也许部分会被挖更多的洞所抵消，因为成本降低了，但不太可能立即让挖洞需求增加10倍，因为洞的用途毕竟是有限的。

我认为对大多数知识工作来说，基本上也是这样。知识工作者相对于他们消耗的能量来说被过度支付了，而AI将纠正这一点。这解释了为什么中国人免费提供（训练所需资源更为适中的）模型。他们喜欢看到美国的通缩经济。即使你监管控制了美国政府，也没有人能在AI上获得垄断，我们不再生活在单极世界了。

AI将变得便宜、无处不在，摧毁许多行业的工作，破坏工资溢价，并引起对地位等级的大规模重新评估。还有谁对大萧条2：电子布加洛感到兴奋？