it doesn’t matter. the principle is that if x is the length of your context window, then at 0.4x the chance of hallucinations start increasing exponentially. we’re now at token windows of 1M, and all it does is shift that hallucination window further away, so the model ‘feels’ stronger because it takes longer before it hallucinates, but eventually it always does.

By the end of long workflows

Yes, this has been known for 10 years.

I don’t think it’s garbage. It’s highly effective use of technology to get a message across to the exact target audience they have.

Getting pretty tired of anything CGI being labeled AI slop. If you watched cartoons, an MCU movie or have done anything on a computer with graphics, you might as well call that AI slop too.

I prefer to reserve AI slop for people who offload their creativity to LLMs, which is not what these guys did.

It’s not so Iran can make more money, it’s for Russia. Also it helped distract us from the fact that indicted bibi is continuing with a 2nd genocide for his Greater Israel sideproject.