Jump in the discussion.

No email address required.

No way this an actual specific policy choice. They just created a system that is statistically likely to refuse requests and churn out woke gibberish because they were so concerned about "safety" in the case of various specific issues. The unintended effect is that the model thinks this is just what English conversations look like. For any given topic it could randomly spit out sanctimonious nonsense instead of helping you, especially if you prime it by alluding to any sort of demographic information.

I remember the small version of the Meta LLM would even refuse to help you with recipes because some people are allergic to dairy or whatever

Jump in the discussion.

No email address required.

For any given topic it could randomly spit out sanctimonious nonsense instead of helping you,

So it's a redditor

Jump in the discussion.

No email address required.

the AI "knows" that C and C++ are "unsafe" languages but doesn't understand what that actually means.

Jump in the discussion.

No email address required.

Teach kids scratch, then javascript, then rust.

Jump in the discussion.

No email address required.

>Teach kids rust

:marseygroomergrooming#:

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.