Hacker News new | ask | show | jobs
by feet 1439 days ago
Neural networks in their current implementation don't even come close to the complexity and computational power of a single real neuron. While that is the case, on an extremely basic level they are designed after concepts learned about real neuronal networks

I do mean extremely basic I can't emphasize this enough, so much so that I normally say its not even close

1 comments

I want to second the 'extremely basic' part.

But, yes, it is biomimicry. Computer scientists looked at actual neuronal networks in animals, and said 'let's try to do some extremely simplified version of that' and that is how we got "neuronal networks" (bad name, but what we gonna do).

I heard there was some work in adding inhibitory neurons at least. Progress!