Anthropic IPO 是否对投资者和社会都是坏赌注?
Is Anthropic’s IPO a bad bet, both for investors and society?
Gary Marcus 对即将到来的 Anthropic IPO 高估值提出质疑,认为其收入已不再像 Q2 那样高速增长,中国模型追赶上后用户开始更谨慎地控制 token 支出,收入趋于平缓。他还认为 Dario 在医疗等领域的激进承诺不切实际,并呼吁为人类福祉考虑不要购买 Anthropic 股票。
From the person who told you that Oracle was at “peak bubble” when it was trading at over $300 (today’s price is around $115), here’s why I (not a professional investor so just giving my own reasoning!) have my doubts about the high valuation for the upcoming Anthropic IPO.
Either way, as a person who has been warning about risks of AI night and day for the last several years, I would like you to consider not buying Anthropic stock for the sake of humanity.
I will get to the moral side of the argument at the end.
But first, the bull case, as I understand it, and then what I think it gets wrong and leaves out.
Part I: The bull case
Anthropic makes strong AI models, and arguably, by a small margin, the best.
Back in October Brad Gerstner made the bull case with this chart:

And there is also this y-axis monstrosity

Both assumed revenue would go up and up.
That brings me to Part II.
Part II: Misconceptions
The first misconception was that revenue would continue to grow exponentially. I have often called that the trillion-pound baby fallacy, after this memorable tweet.

The point was that few exponentials last forever. And, sure enough, Anthropic has already fallen off Q2’s revenue rocket:

The most recent data I have seen shows that the revenue rocket is no longer accelerating as fast as before:

The green line represents the widespread fantasy, and roughly the rocket that Anthropic would have needed to make it as a gargantuan sustainable business. The darker bars on the right show that it has already fallen off the pace—which is why Anthropic probably won’t make it. Q2 was special, at the height of tokenmaxxing and before Chinese models (perhaps thanks to the largesse of Jensen Huang) caught up. Now people have options, and they watch their token spend more carefully, which means revenue has leveled out.
Another misconception is to believe that Dario can deliver on his wilder promises, especially around medicine. As I noted in a recent article, those promises are absolutely absurd. We are more likely to get hit by a meteor.
But what really matters the most is what’s been left out of the bull case.
来源:Gary Marcus:The Road to AI We Can Trust · garymarcus.substack.com