Aaron Rainbolt on Nostr: I guess that's comforting. I still would probably not do it though. Reasons why: * ...

I guess that's comforting. I still would probably not do it though.

Reasons why:

* Many "open-source AI" things are not open-source at all, but are only source-available if even that. Llama is an example of a source-available-but-not-even-that AI.
* AI is extremely difficult to audit to make sure it's doing things that are ethical. Much AI is open-weight (which is a hilarious example of true write-only memory - you can tune the model but you can't see how it was trained), and even fully open-source AI can't be audited without pouring over a corpus of data that would make your head explode. I don't want my AI advising me to do things because it happens to have a racist or manipulative nature thanks to being trained on the wrong data.
* AI when used in the wrong context is just so useless. AI writing tools have made much of the Web's content objectively bad, boring, and useless. AI image generators are actually neat when used properly but oftentimes make horrific results that people try to pass off as useful. When used to solve pointed, specific problems, AI can shine, but in other instances it can't be trusted to do things right and even using it can have a negative effect on work.