/lit/ - Literature

This is an example that maybe is not ideal for this board, but just imagine an AI, a neural network like pic related. Each of the circles is called a "neuron", and it basically receives a number and decides if it's gonna pass along a number or not. These things are remarkably good at learning to do a few things, let's say classify images of dogs.

What this means is that, after trained, a neural network receives the image of the dog as inputs, the values of the colors in each pixel being the input, and those numbers go into the first column of circles there. Then each circle decides if it passes along the information or not, etc etc, until on the last neuron you get a single number that tells you if the image is a dog or not.
Let's say this neural network never makes mistakes. I could give you all the values of all the variables involved here. All the lines in pic related, all the weights on the connections between circles, everything. You would have a bunch of numbers in front of you. And then I'd say "show me where the dog classification knowledge is" and you'd be stumped. Science hasn't figured this out yet. We are just only starting to learn how to probe this bunch of numbers to retrieve useful knowledge for humans, but the fact is, in just a bunch of fucking circles connected you can somehow store relevant information for processing dog images into correct classifications.

If I used the same arguments as a lot of people in this thread we'd end up accepting that there must non physical properties aiding the decision process of the AI, since we humans can't look at the bunch of numbers of the individual components and understand how the decision was made. This is of course ridiculously stupid. It's similarly stupid for the human brain case.

>>13616207
that's just an oversimplification so primitive kaczynski tards can understand it