all 4 comments

sorted by: hot top controversial new old
[–] 3 points 1 year ago

The irony of the company who pioneered scraping the internet to feed their AI making me complete a Cloudflare captcha because I use Firefox on a VPN.

  • source
  • [–] 2 points 1 year ago

    It wasn't what I expected, but I'm satisfied.

    I thought they were going to release the GPT-3 source code.

  • source
  • [–] 1 point 1 year ago (1 child)

    So deepseek / quan but from Microsoft?

  • source
  • hideshow 1 child comment
  • 5 comments in c/gai@sopuli.xyz view that post
    [–] 2 points 1 year ago (4 children)

    Sorry, but I can't comply with that.

  • source
  • hideshow 4 child comments
  • [–] [S] 1 point 1 year ago (3 children)

    I'm trying the 20b weights model in LM studio now, and it's not having any issues with providing summaries of plots of movies/shows/episodes. Do you know what kind of system prompt or any other details on what's needed to keep it from responding?

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 1 year ago (2 children)

    Do you also see the reasoning part? I've played with it yesterday and yeah, like half of the reasoning is whether it's legal to ask a question.

    I've heard other people on reddit tell that results like on the screenshots are because of quantization, I've played with raw 20b.

    Well at least red pill question is always forbidden, but for the rest, cannot reproduce.

  • source
  • parent
  • hideshow 2 child comments
  • [–] [S] 2 points 1 year ago (1 child)

    I tried the red prompt one word for word, and it gave me a list of common red pill ideas. It did also tell me about the misconceptions with each of the red pill ideas and why I shouldn't believe them 100%, but it didn't refuse to respond to the question.

    I'm currently running it with a generic "you are a helpful assistant" system prompt and low reasoning, it's possible to that the refusal to answer some questions only happens at higher reasoning levels or a different system prompt.

  • source
  • parent
  • hideshow 1 child comment