fizzle

Most of the power consumption comes from training and optimising models. You only interact with the finished product, so power per query is very low compared to that required to develop the LLM.

Everyone knows.

I doubt there’s anyone here who is not already aware of these details.

Common sense says you can’t get convicted as a sex offender for grabbing someone by the arm if you had reason to believe they were in imminent danger.

These kinds of stories have been around since the dawn of time for all sorts of things. Whenever you investigate the details it turns out that either the story is completely made up, or the offense was much more serious than it sounded.

Like the woman who sued macdonalds for getting third degree burns because their coffee was too hot.