Dario Amodei posted a note on Anthropic’s position on open-weight models today.
The note is pathetic, and is filled with delusions and American exceptionalism. It is clearly begging for the US government to ban open-weights, while simultaneously trying to pretend they don’t want that.
It has an audience of one, and that is JD Vance. They’re begging him to gather all of the pedos, rapists and grifters in a room and demand that Kimi K3 be banned. For like, the kids or something.
It repeats the laughable American exceptionalism farce that Chinese models can only succeed by distilling off American models, patronizing Vance by citing a widely mocked speech he gave in Paris1. Given that Opus 5 is randomly inserting Chinese in English conversations, this is a particularly ironic claim.
Again, Anthropic desperately needs to push this lie: They need everyone to believe, particularly their many shareholders (yes a pre-IPO company has shareholders), that this Chinese challenge is really a mirage – merely a second-rate knock off, he wants you to believe – and surely Anthropic has a moat…they just need to add more crocodiles!
Only China has pretty much dominated AI research over the past year or two. Chinese papers are flowing fast and hard, and have legitimately been the primary innovators that have been moving the space forward. A big reason might be that Chinese firms don’t have a money printing press or infinite runway, so they’ve learned to do more with less, and almost all of their inventions have been to that end. And when you can do more with less, you can then do way more with the same.
It’s one of those situations where getting the smartest people working on the problem beats dumping infinite hardware and endless data centres on the need. And various innovations provided by some of these Chinese groups have been precisely what has moved the US industry forward as well, given that China has been pretty open. I suspect they’re going to stop being so open.
That China is now on par is not remotely surprising, and it was clear this was coming.
It also laughably argues that the real danger is non-US governments getting powerful, unrestricted AI. Again, the target audience is JD Vance, but to the rest of the world the US is the danger. I have infinitely less worry about China or other “authoritarian” governments having advanced AI than I do a war-crazed, murderous, terroristic, thieving, pirate nation like what the US has become, led by a dipshit constantly tweeting (sorry, “Truthing”) random threats to annihilate civilizations, with all checks and balances being utterly abandoned. Maybe when you stop the constant war-mongering and murdering school children and blowing up random boats and gloating about all the resources you stole and “joking” about invading an unending series of countries you’ll be in a position where this doesn’t read like some sort of Pravda 1970s farce. But right now, you are indisputably the bad guys.
If Amodei believed an iota of the tripe he wrote in this joke of a letter, he’d close shop and secure-wipe all the hard drives. But he’s a two-faced, self-serving megalomaniac, and can feel the capitalization eroding around him, so this desperate, unbelievably pathetic note is the result.
But again, it has an audience of one, and Amodei was willing to humiliate himself with this farcical nonsense to do so. I would not be remotely surprised if the US starts banning models like K3, though they’re going to be a bit surprised when every other Western nation tells them to go fuck themselves.
Footnotes
-
Which is hilarious because Vance was just repeating what Anthropic told him, as they desperately try to pretend they have a moat. It’s like two Wikipedia articles citing each other, each demanding credibility by pointing at the other. ↩