• spaduf
    link
    fedilink
    English
    arrow-up
    12
    ·
    1 year ago

    In the bio-tech space specifically, most of these recent advancements (e.g. protein folding) are a direct result of advancements in Deep Neural Networks. This is because things that are impossible to model explicitly without a working understanding of the system are easily understood by a neural network. The reasons for this are precisely the same reasons why we ourselves are so good at the task. For a real world example of humans being applied to a similar task, look up the Japanese educational industry around chick sexing.

      • spaduf
        link
        fedilink
        English
        arrow-up
        9
        ·
        1 year ago

        Lol from Wikipedia, “Chick sexing is the method of distinguishing the sex of chickens and other hatchlings.” For further context, this is done by a highly trained individual who is able to briefly look at the cloaca of a chick and immediately recognize their sex. They are even known to be able to perform this on two chicks at the same time (one in each hand). Joshua Foer briefly talks about it in his book Moonwalking with Einstein where he attributes the skill to the incredible ability of humans to perform rote memorization. I think it is better understood as an example of the incredible abilities of these sorts of neural networks (digital and biological) for pattern identification.