• ubergeek@lemmy.today
    link
    fedilink
    English
    arrow-up
    11
    arrow-down
    3
    ·
    3 days ago

    Neural networks in AI barely relate to any biological neural network.

    Neural networks in AI are essentially a “scored” pachinko machine, with each peg having different numbers, which cause a “score” to go up for the “right” answer.

    Basically, just a really fast, and expensive sieve filter.

    • Zacryon@feddit.org
      link
      fedilink
      English
      arrow-up
      3
      ·
      2 days ago

      I said “inspired by” and not “exact digital replicas”.

      In classical MLP networks a neuron is modeled as an activation function depending on its inputs. Connections between those are “learned”, basically weights which determine the influence of one neuron’s output on the next neuron’s input. This is indeed Inspired by biological neural networks.

      Interestingly, in some computer vision deep learning architectures, we have found structures after the training procedure which are even similar to how human vision works.

      There are a bunch of different artificial neural network types, most – if not all – inspired by biology. I wouldn’t be so bold to reduce them in that absurd manner you did.

      • ubergeek@lemmy.today
        link
        fedilink
        English
        arrow-up
        1
        arrow-down
        2
        ·
        2 days ago

        I said “inspired by” and not “exact digital replicas”

        its not, though. Its best described as inspired by a big pachinko machine, with weighted pegs.

        It is almost in no way inspired by. Thats just propaganda being put out to make AI more palatable, and personable.

        There are a bunch of different artificial neural network types, most – if not all – inspired by biology. I wouldn’t be so bold to reduce them in that absurd manner you did.

        I would be, because it’s factual.

        If it was “inspired by” it would be able to tell the difference between running over a person, and avoiding a car, by example. It wouldn’t start hallucinating when asked simple questions, because a biological brain acts in congruence with it’s inputs.

        Which happens because of a web of interconnections and spanning of multiple sphere’s, with two major ones acting as checks on the other. Which is nothing like any current AI model.

        In current models, each token has a limited number of interconnects, and always to a neighbor node. That is nothing like a biological neuronal network.

        • Zacryon@feddit.org
          link
          fedilink
          English
          arrow-up
          1
          ·
          2 hours ago

          its not, though. Its best described as inspired by a big pachinko machine, with weighted pegs.It is almost in no way inspired by. Thats just propaganda being put out to make AI more palatable, and personable.

          Get your facts straight.
          The multi layer perceptron was first proposed in 1943 and was indeed inspired by biological networks: https://doi.org/10.1007/BF02478259

          You can be sure this wasn’t to make it “more palatable”, wtf.

          Regarding the rest of your reply:

          You seem to be expecting a fully functioning digital brain as replica of the human brain. That’s not what current ANNs in modern AI methods do.

          Although they are in their core inspired by nature (which is why I originally said that advancements in brain research can aid the development of more advanced AI models), they work structurally different. And ANNs for example are just simplified mathematical models of biological neural nets. I’ve described basic properties before. Further characteristics, like neurogenesis, transmission speeds influenced by myelinated or unmyelinated axons, different types and subnets of neurons, like inhibitors, etc., are not included.

          There is quite a large difference between simplfied models which are “inspired by” nature and exact digital replicas. It seems you are not accepting this.