PS. This is Text from Bing AI.

  • TheWildOutside@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    “Meta had partnered with Dell” Oh so if you hit a key to hard while inputting a prompt the entire model will break?

  • dethorin@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    Bing AI as a source is shit. Many times the quoted primary sources don’t support whatever Bing AI claims.

  • stddealer@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    Putting “safety” mechanism in foundational models is dumb imo. They are not just text generators, they are statistical models about human languages, and it shouldn’t have made up arbitrary biases about what language should look like.

    • api@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      It’s not hard to fine tune base models for any bias you want. “Zero bias” isn’t possible. There’s always some bias in the training data.

  • FPham@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    All they need to do is make it 180B and most people will have no way to abuse it. (Or use it)

    • gfth45fghmnfs@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      Llama 3 180B model on every quantization & compression steroid available might bang. Also it will show how far or close oss llms are from gpt4

  • llamaShill@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    That’s the prevailing idea based on all the info we have so far:

    • Llama 1 training was from around July 2022 to January 2023, Llama 2 from January 2023 to July 2023, so Llama 3 could plausibly be from July 2023 to January 2024.
    • In August, a credible rumor from an OpenAI researcher claimed that Meta talked about having the compute to train Llama 3 and Llama 4, with Llama 3 being as good as GPT-4.
    • In an interview with Lex Fridman published Sept. 28, Mark Zuckerberg has said they’re always training another model and already working on the next generation when talking about Llama.
    • At Meta Connect on Sept. 27 - 28, they said more news about Llama will be put out next year.

    WSJ published an exclusive on Sept. 10 that said Meta’s next LLM won’t start training until early 2024, meaning a release wouldn’t happen until much later, but they may have been mistaken since this seems to contradict Mark’s recent words. Meta could have also accelerated their plans to stay relevant in the LLM race, especially since leaks about their LLM development have shown they’ve put more emphasis on productizing Llama and incorporating it within their apps.

  • Sabin_Stargem@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    My speculation is that the “safety” of major LLM providers like Facebook won’t be substantial. They are probably designed so that “popping the lock” isn’t difficult. A fig leaf and shrug to placate outsiders, while allowing their actual audience to carry on.

    • Available_Screen_922@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      I’m not sure that it’s plausible to add meaningful safety to open source LLMs.

      Incidentally, that does worry me some in the long term.

  • FPham@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    Ooooh, I really want to see the marvelous “mechanism” to prevent open source model from misuse.

    • cvdbdo@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      *Provided you use the non fine tuned model straight from Meta’s download page so that they’re safe.

    • a_beautiful_rhind@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      They can ingrain those refusals pretty deep to where it will be irritating to use and hard to fine tune out. Vicuna has a bit of this.

      • FPham@alien.topB
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 year ago

        The point is Meta till now always also released base models.

        That would require them to release only fine-tuned model like llama2-chat was. Then they can bork it which would be irreversible. But if they give out base model without some stuff - it can be easily added.

    • mrjackspade@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      I tried coaxing an answer out of it, and as far as I got was

      1. One random redditors comment saying “next year” was announced at Meta Connect
      2. “It makes sense” due to the spacing between 1 and 2
  • obvithrowaway34434@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    I’m more interested in the next Mistral release, none of that corporate “safety” bs. It would also be good to have a truly open-source model (that releases both weight and the training data).

  • Ilforte@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    Maybe. If it won’t be killed because safety. I also think that it’s implausible it’ll be much better than what we have, or even catch up to 3.5 fully.

    • dogesator@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      Already Mistral 7B fine tunes reaching parity with gpt-3.5 in most benchmarks.

      I’d be very surprised if Llama-3 70B fine tunes don’t significantly outperform GPT-3.5 in nearly every metric.

    • Amgadoz@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      Well, two areas where Llama can improve are:

      1. Multilingual capabilities
      2. Mixture of Experts architecture