Hacker News new | ask | show | jobs
by falafels 490 days ago
I listened to an interview a while back with the researchers did this analysis. They developed privacy-preserving techniques, avoiding having to read user conversations directly.

This is Anthropic we're talking about, they're rightly recognized as the 'ethical' AI company.

2 comments

There would be even better privacy if they published the weights so that they could be run by third parties that have no log policies or even better, locally.
Words I typed into a service I paid for are now in a huggingface repo. Fuck that.
Are they?
They aren't.