Hacker News Reader

Comment by anakaine | Hacker News Reader

I disagree with cancelling the OpenAI subscription. I've been getting some help from o1 for both python and php recently, and o1 was doing massively better for the python stuff (it ran, deepseeks didn't and wont with prompt refinement).

View on HN

Also for some philosophical stuff DeepSeek just won't do it. I'm working on an essay about spirituality and sometimes it just responds that it doesn't know how to work on those types of problems and we should do something fun like math or games, claud tends to reply with something more like "I have to be honest with you, reincarnation is not real" and ChatGPT doesn't seem to care about that kinda thing at all.

Just don’t ask it about anything related to Tiananmen square or president Pooh..

I’d guess they didn’t quite a bit of fine tuning to censor some more sensitive topics which probably impacts the output quality for other non technical subjects.

Would fine-tuning by using a LoRA paper over the censorship to a large degree?

Why even bother decensoring it (except academic curiosity ig)? There are a million other ways you can learn about those subjects.

The people making the model probably don't really give a shit about politics and just did the minimum to avoid being embarassed, but if people start jailbreaking it they will be forced to care.

Because the best way to learn is through interrogation.

I don't give a damn about ideology I just want everything ever thought or written searchable and interactive

IIRC thezvi's summary post on R1 mentioned that R1 is amazing for general reasoning and is very clearly a successful proof of concept/capability but a lot of effort seems to have been put into making o1 Good At Code as a practical matter, whereas R1 seems to have been more a research project which proved out the approaches and then was released without sanding the rough edges off because that wasn't the point.

Were you running a local model?

Reply on news.ycombinator.com