Zeth0s

I am mainly thinking about matching navigation history with identifiable information... You are right, It's a tricky thing...

I also wonder, if lemmy becomes a thing, with numbers in the same order of magnitude of reddit, if and how gdpr will affect server admins... Having a privacy anonymization tool built in by design might avoid headaches on the long term

Same experience. It is also increasingly ignoring instructions, always ends up with similar answers whatever the request is formulated. It's a pity

I am thinking more of a Meta "threads" -like situation. Not necessarily malicious, just a different privacy expectations between user and provider

Thanks, makes sense

Unfortunately that is not the case. Closed sourced software for small communities are not safer. My company had an incredibly embarrassing data leak because they outsourced some work and trusted a software used also by the competitors. Unfortunately the issue was found by one of our customers and ended up on the newspapers.

Absolutely deserved, but still, closed sourced stuff is not more secure

What if database entries are encrypted, so that a person cannot match email and username with the requests in the urls?

Users' client create encryption key on client side. Would it make sense?

I used mainly for programming support, brainstorming on techs and libraries, and to refine technical documentation. But support for programming is becoming a bit of a pain honestly...

Thanks. It would be interesting to understand if some anonymization technique could technically be created, and eventually implemented in the future.

Because it looks a pretty dangerous situation, given what people discuss on social media nowadays...

Thanks for the clarification!

Thanks! I tried vicuna, but I didn't find it very good for programming. I will keep searching :)

What local models are you using that are better? Not trying to argue, honest interest

I find that recently the effort needed to get the "right" answer is much more than in the past for gpt-4. That's my impression. At the end I am finding myself more often going back to google, stack overflow, manuals, medium...

I believe they distilled the model to much for performances, or the rlhf is really degrading the model performances