platinumrad

> At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first - they believe no one else will act responsibly, so they must do it themselves, despite the risk.

OpenAI are mercenaries, Anthropic is a cult. I know which I prefer.

Are there consequences for lying in sworn testimony?

Funny. Dario seems like the biggest snake in the industry to me and has leaned the hardest into doom marketing out of all of the influential leaders. With Altman (or Google), it's a transaction, and that's something I can live with.

ICE manages to harass and detain plenty of citizens.

Counterpoint: no it's not.

I don't think the rando providers like Novita, Phala, AkashML, etc., which frequently provide discounts to compete for traffic, are VC-funded. They have to actually make money to stay afloat.

API are not subsidized. We know this is true from the existence of independent inference providers (a lot of which are crypto companies that would otherwise just be mining if inference weren't actually profitable).

I'm not a fan of Langley but this is just whinging about having the slightest of restrictions put on your toys.

These things aren't binary. There is a large middle ground, which is where Pangram sits today. It's good enough to be useful, but I absolutely would not be willing to fail a student on the basis of a Pangram positive.

If you are implying that this Claude watermark may be good enough to enter the "evidence to fail a student" category then we have a major disagreement.

You're upcharging 50-70 dollars to flash a firmware image?

If you take it as given that the watermark detector won't have "plausible false positives" then, sure, your argument becomes trivial. I'm not taking that as given.

You know that dlopen does not compare.

Reads like LinkedIn slop.

I think instruction following carries outsized weight in these evaluations.

LLMs aren't people.

We're speedrunning all of the worst parts of China, except privatized and we're not even going to get some nice trains or cheap housing out of it.

The question you're asking has nothing to do with who has the burden of proof when it comes to claims about Pangram, but I'll answer it anyway.

Today, the best way is probably Pangram. Tomorrow, it might not be, especially if they try to push their recall up.

You might have to make peace with the fact that there may not always be a tool that does what you want.

You can't make a blanket statement like this without knowing how the watermark is implemented.

Pangram is subjectively very useful and I personally subscribe, but the burden of proof is on them. The product is very much "trust me bro" and I fear that if they ever try to improve recall both their precision and reputation will tank.

Unfortunately for static type weenies like me (and you, presumably), types don't seem to matter at all, or Python and Javascript wouldn't be on top. There's no reason to believe that Ada's type system is so unique that it alone can help AI self-correct, and Rust, Haskell, ML, Typescript, etc. can't.

Careful. If you admit to using models that weren't trained by OpenAI or Anthropic then you might hauled in front of Congress: https://www.scmp.com/news/china/diplomacy/article/3362616/us...

I have a lot of problems with Aschenbrenner and the "Rationalist" community as a whole, but this is not a factually correct characterization of Effective Altruism.

> and neither will it help with getting it installed on corporate networks

This might not be something the author cares about. Compliance with corporate policies is certainly not something I've prioritized when giving away open source software for free.

This is what https://dctransparency.com/ai-firms-pour-record-money-into-w... buys you.

The genius who brought us AI 2027.

Yes. For example, third-party inference providers serve DeepSeek V4 Flash just as cheaply as DeepSeek themselves, if not even more so. This is very strong evidence that the low price of the model is not subsidized.

We can guess based on the decisions of other inference providers who serve these models.

I get that this is a nice benchmark in that the result can be judged visually, but it seems extraordinarily far from what people actually use these models for.

The message is a lot better than what Amodei and Altman are pushing at least.

Does it refuse to work on "cyber"?