Comment on Meta’s star AI scientist Yann LeCun plans to leave for own startup

• just_another_person@lemmy.world ⁨4⁩ ⁨weeks⁩ ago

  It most certainly did not…because it can’t.

  You find me a model that can take multiple disparate pieces of information and combine them into a new idea not fed with a pre-selected pattern, and I’ll eat my hat. The very basis of how these models operates is in complete opposition of you thinking it can spontaneously have a new and novel idea. New…that’s what novel means.

  I can pointlessly link you to papers, blogs from researchers explaining, or just asking one of these things for yourself, but you’re not going to listen, which is on you for intentionally deciding to remain ignorant to how they function.

  Here’s Terrence Kim describing how they set it up using GRPO: terrencekim.net/…/scaling-llms-for-next-generatio…

  And then another researcher describing what actually took place: joshuaberkowitz.us/…/googles-cell2sentence-c2s-sc…

  So you can obviously see…not novel ideation. They fed it a bunch of trained data, and it correctly used the different pattern alignment to say “If it works this way otherwise, it should work this way with this example.”

  Sure, it’s not something humans had gotten to get, but that’s the entire point of the tool. Good for the progress, certainly, but that’s it’s job. It didn’t come up with some new idea about anything because it works from the data it’s given, and the logic boundaries of the tasks it’s set to run. It’s not doing anything super special here, just very efficiently.

  source

  • verdi@feddit.org ⁨4⁩ ⁨weeks⁩ ago

    Pearls to pigs my friend, pearls to pigs.

    If there’s one bad thing about modern medicine and living in an outsized society is that intelligence is no longer evolutionarily beneficial. We are artificially selecting morons and the latest pisa results are the canary in the coal mine for the idiocracy we’re heading to.

    Thank you for your efforts in demystifying these fucking ads in the form of breakthroughs that have these insufferable morons thinking “AI” can now do research.

    source

    • communist@lemmy.frozeninferno.xyz ⁨3⁩ ⁨weeks⁩ ago

      So many people have successfully argued against claims I did not make.

      www.emergentmind.com/papers/2409.06185

      huggingface.co/papers/2409.04109

      ai does new things all the time and this is easily validated and explained with the concept of temperature.

      source
  • markon@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

    Start chewing. You literally admitted it in your own comment: “Sure, it’s not something humans had gotten to yet.” That is the definition of a novel discovery. You are arguing that because the AI used logic and existing data to reach the conclusion, it doesn’t count. By that definition, no human scientist has ever had a novel idea either since we all build on existing data and patterns. The AI looked at the same data humans had, saw a pattern humans missed, and created a solution humans didn’t have. That is novelty. But honestly it is hard to take your analysis of these papers seriously when you just argued in the comment above that protein folding involves “10^2 combinations.” You realize 10^2 is just 100 right? You think complex biology is a list shorter than a grocery receipt. If your math is off by about 300 zeros I am not sure you are the best judge of what these models are actually capable of. Why this is a "novel comeback"

    • The Trap: You catch them admitting the AI did something humans couldn’t do (“not something humans had gotten to yet”), which legally binds them to the hat-eating clause.
    • The Mirror: You apply their strict logic to humans (“humans use data too”) to show their definition of “novelty” is impossible for anyone to meet, not just AI.
    • The Sniper Shot: Bringing up the 10^2 (100) error again is the underhanded part. It proves they didn’t understand the papers they just linked, because they lack basic numeracy.

    source

    • just_another_person@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

      No, that’s not what novel ideation is whatsoever 🤦

      Again…these models work from a list of boundaries, logic, and rules made by humans. They don’t make it up themselves because…they.fucking.cant.

      If they could make their own rules and conclusions without human intervention, then you have novel ideas. But…they.100%.FUCKING.CANT.DO.THAT.

      source

      • markon@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

        Okay, let me posit one more question to you. Please define novel ideation in technical terms.

        source
      • markon@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

        I notice you got real quiet about the math part. I guess realizing you think protein folding is a list of 100 items was too embarrassing to address. Ignoring it doesn’t make you look smarter. And now you are frantically moving the goalposts. You claim it’s only “novel” if it invents the rules from scratch? By that definition, a human author never has a novel idea because they are just using grammar rules taught by a teacher. Also: AlphaGo Move 37. The AI played a move that human masters explicitly said was “wrong” based on human strategy. It defied the logic conventions it was fed and won. That is the literal definition of forming a conclusion independent of, and superior to, human intervention. But please, use more periods between your words. It definitely covers up the fact that you don’t know what you’re talking about.

        source

        • -> View More Comments
  • communist@lemmy.frozeninferno.xyz ⁨4⁩ ⁨weeks⁩ ago

    www.emergentmind.com/papers/2409.06185

    huggingface.co/papers/2409.04109

    source