Hacker News Reader

HNHacker News

Top New Best Ask Show Jobs

Comment by snidane | Hacker News Reader

Parent Full thread

snidane·

200 neurons, yet still more AI complete than any concocted deep learning attempt with way more of so called "neurons".

View on HN

A 'neuron' might be a kind of the biological equivalent to the CPU or microcontroller.

So saying that the Tardigrade has 200 CPUs sounds pretty good, actually.

If you build a drone for example with 200 CPUs, you'd expect it to be pretty capable I'd imagine, including some form of AI or machine learning.

And it seems to me neurons by default are machine learning "CPU"s...

Really depends, doesn't it? 200x something like a attiny13 is still kind of useless in that regard

I've never tried to build a Tardigrade with 200 attiny's, so I'm not quite sure....ESP32s with 8MB of PSRAM on the other hand....would it be enough?

Well, not really. They are neither nodes only in a ML algorithm but neither are they complete processors. They are somewhere in between where they can do some limited “signal processing” (eg. throttle signal frequency, negate it, and have some form of primitive memory even), but they are likely not Turing complex in themselves.

Just dumping this somewhere, pretty informative video on how neuronal signalling works.

https://www.youtube.com/watch?v=ZKE8qK9UCrU

Great video! Love the visuals -- thanks:)

Would they need to be Turing complete? Seems to me they operate on a completely different principle.

Anyways my statement was analogical -- and as we know analogies are merely tools for extending thinking from one paradigm to another. Best not to take it TOO literally.

Reply on news.ycombinator.com