They can take bribes as effectively as any other department. Actual review of robots is presumably out of scope.

Can you point me to the statute that makes it ok for humans to learn from copyrighted work, but forbids AI? I’ll wait.

I am unaware of any statute or constitutional provision that confers rights to AI.

We’re all distillers now.

Probably leads to long term site death as fewer new people have the patience for that, and older users age out/die.

Who’s going to keep coming back to post one 40-char comment every two days?

It’s a real problem. Being able to pull in reputation from other sources might help: link established GitHub or (shudder) LinkedIn accounts? Or build the long-fantasized anonymous web of trust reputation thing?

But I would expect they would at least use post-training to discourage identifying as the wrong company.

A lot of API customers don't want their chatbot or agent to identify as Claude or GPT, they want it to identify as <customisable marketingname>.

That's probably the true reason.

That only makes sense if you don't sell access to the model for 3rd party product development. One of those products is chat personas/agents. The company integrating the model will want it's own identity.

I don’t find Sonnet usable at all. Especially for the price. I’d much rather get half the usage of Opus.

So does AGI just mean infallible?

Currently it either means “better than the median human at every task including things like counting to 1000 or folding clothes” or it means “for each task, better than the top-n human at that task”.

It’s clearly not actually that “generalized” yet because it’s unable to do a number of very simple things that almost any 6-year old could do, such as count to 100 without using any tools.

It’s still a very specific type of intelligence, with some real breadth to it, but not general intelligence.

I think if you ask the median human to count to 100... more often than not you'll hear "One, Two, skip a few, Ninety-Nine, One-Hundred".

Agreed, that was a really fallacious point. It’s like saying buying car insurance didn’t pay off because you didn’t get in a wreck.

Not that I entirely buy missiles == insurance, just that the unused == wasted framing is too simplistic.

At least with a wreck there is a real risk factor you might experience yourself. What is the risk factor for ICBM based MAD? Zero. It has never happened before, to anyone. Might as well buy an insurance for alien invasion based damages to your home.

True, the metaphor breaks down because the act of buying insurance doesn’t change the likelihood of being in a wreck.

Yep. Classic “giant series A, company gets amazing office space” vibes.

How does that work when hundreds of people are working on a project, with a few new people joining every week?

It boggles me that anyone would argue against standardization, or believe that “just get hundreds of people to agree today and tomorrow and every future day” is a better solution.

In my own experience it works pretty well. In fact better than the autoformatters.

It was essentially not an issue. Sometimes I’d grumble about someone, but it would never get in my way the way black or ruff would regularly.