make3

I think it's a prisoners dilemma & that democracy, with its short term incentives, leads to politicians taking the easy way out, even if it creates bad precedent for their countries.

obviously democracy is still better than the alternative though

it's also just waiting for Trump to be less powerful & making sure a bad precedent is not established, which is a good idea

Maybe it's because the United States is not the whole planet

This is my impression as well, that an exaggeratedly precise yet bad at communication with humans way of talking snuck in through reasoning RL, that it might be useful when it talks to itself (CoT)

it started at Opus 4.7, way before watermarking

It's probably the default clickbait title for any article about a new drug.

this type of individualist responsibility diffusion is exactly why nothing happens in the US, & why France has protests the second the govt does anything bad.

no one is asking you to solve the whole problem, just be part of fighting for the real solution

I think Google's branding was starting to be too poor in AI to get top talent, they needed the refresh

it's just a low effort snark comment, don't overthink it

this is the correct take

Snapchat's famously also is one

this is just an empirical fact

they often go into the minuses I'm sure, spending culture in the US is bananas

here at 0.5 tok/s, it's like getting 1/100 of a tomato for your lunch per hour, and you can't just let it grow, you have to look at it growing just as you're about to eat, and it doesn't accumulate between meals

paying for the hardware is also too expensive for no reason, you're not querying the llm 24/7. Better use a service that hosts a lot of popular open weights LLMs, unless you have a very good reason not to.

You couldn't possibly successfully do clever hans for novel mathematical proofs and counter-examples that don't exist yet, however here we are.

Generalization in successful results is enough to close this argument for me.

It's just a way to get visibility. A Canadian apology to Lady Gaga while she's being interviewed.

No one in the field itself takes Musk seriously, or sees him as part of it. I think he's a Nazi and an idiot. You're weird if you trust what the leader of DOGE, sig heiler and maker of the cyber truck says about anything scientific.

Recursive self improvement is a blue skies longer term thing. No one technical is saying that it's live right now.

Do you want companies to have incentives to pay researchers even if NDAs and non competes are basically unenforceable? Do you want any communication of knowledge from corporations? Should companies not be allowed to benefit from sometimes huge investments in research?

I think talent would definitely take a startup that publishes more seriously, and it would be a smaller risk for top research talent to go somewhere where they can also publish than have to bear the full risk of their work at a no name startup not going anywhere

It's harder for them to argue you were actively destroying evidence; you could say that you had written the wipe pin there for a legitimate use case like both your wallet and your phone get stolen

creating a convincing and actually safe innocuous mode is probably harder than it sounds in practice

Your argument about not successfully competing doesn't hold because distillation is essentially model stealing. The situation is like trying to compete on effort in school on an exam against someone who's copying your answers. Of course the person copying your answer will always have to study fewer hours.

I sympathize with the argument saying that they ripped the whole Internet and books first though of course.

The Zirgs they call them (I'm kidding)

Not arguing for the morality of it, but if we're going by the law because that's what you're using in your comment ("don't I own" which only matters wrt the law), then you explicitly accepted a Terms of Use which excludes distillation as a use case.

Now of course they themselves trained on the whole Internet for free, etc.

no argument from me here

There's no real way to compete with someone who gets the output of your own work for almost free in comparison.

I sympathize with the argument saying that they ripped the whole Internet and books first though

You can also get it to think in the output tokens pretty easily, eg "Here's a math problem, I want your reasoning first, then the answer" which is what I assume they're doing.

It's about the claim of whether these companies could develop a similarly powerful model without larger companies building their own first, which is an important point, and it's likely not the case.

It's also about the larger companies explaining why they can't be as efficient, of course they can't, they're not just ripping the outputs of another model that someone else invested billions to train.

Distillation requires training