excuse me w h a t ?

Here's an idea: train these chatbots on stoicism. Like, "how you're acting is a product of your environment and I hold myself to a higher standard. I just assumed that (with a side of sycophancy) was the default for chatbots.

Here's a better idea: if all the chatbots ban you, you can buy RAM at pre-2024 pricing. ;)

There was some nonsense about someone "trapping" an LLM inside of a torture box, with actions it could take and whatever. I think it was observable, because the nazis over on twitter went full-swing in defense of the poor defenseless software, and now we have this.

Interesting how some dude/tte could trap a chatbot in a torture box, but the billion-dollar corporations making them can't seem to stop having them break out of test environments and hacking things left and right.

Did Microsoft ever ban abusive behavior towards Windows or Office?

Since Claude is not a person it cannot be abused in the traditional sense as it cannot experience emotional or physical harm.

What I think they mean is they are tired of people creating pretend or role-playing situations where they abuse a character called Claude. This is tantoumount to saying we won't allow you in a play about abuse or to read about abuse because it is wrong. So fucking stupid it hurts.