• veng@lemmy.world
        link
        fedilink
        English
        arrow-up
        7
        arrow-down
        11
        ·
        9 months ago

        It’s literally a marketing term for a bunch of structured algorithms at this stage - not some sentient witchcraft

          • veng@lemmy.world
            link
            fedilink
            English
            arrow-up
            4
            ·
            9 months ago

            I guess the point is that its complexity is overrated, but still definitely not ‘simple’.

          • Miaou@jlai.lu
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            1
            ·
            9 months ago

            … It is simple, the idea exists since 40y ago, it’s just being done at scale

      • Miaou@jlai.lu
        link
        fedilink
        English
        arrow-up
        0
        ·
        9 months ago

        I bet I know much more on the topic than you, but please enlighten me on which part of this is complex?

        The core concepts of DNNs are taught in high-school, and putting them together can done by a Bachelor student. Shit, people often advise writing a NN libraries as a good learning exercise when picking up a new programming language.

        I think mathematically illiterate people assume that incredible results necessarily imply complexity, but that’s simply not the case here. Or the idea that unknown things are necessarily complex, maybe.

        The main reason DNNs are popping up is because we finally have the hardware for it. And the second reason is that tech companies have the resources (both financial and in terms of available data) to throw at it.