3 ms·

I think the point is kind of like using arithmetic to find the area under a curve. occasionally you'll get lucky and can work out the area of an arch, but for a

by jfoutz 10y ago

I think the point is kind of like using arithmetic to find the area under a curve. occasionally you'll get lucky and can work out the area of an arch, but for an arbitrary curve, what you really need is calculus. So, also not a neuroscientist. So i think the general point is, their current approach is to turn off neurons and see how things change. with careful evaluation, you should be able to pin down the role of that specific neuron. With a chip, it seems like you'd want to investigate the behavior when output when output is the same as the input - just a pass through. maybe you can deduce the effect of that specific gate. My (poor) understanding is the standard tools weren't enough to identify stuff like ripple adders (or whatever's on the chip) so maybe they're using arithmetic when they need calculus.

AstralStorm 10y ago
Sorry, but integral as done by a neural network is not generally reversible by taking out terms, unless it's a strictly linear no threshold neural net. This one is not. Same problem with the binary functions like most kinds of digital electronics - they're typically combined in highly nonlinear ways. However, knocking out a whole IC and seeing what it did is more reasonable and about the level we are at when understanding functions of subsystems in the brain.
- jfoutz 10y ago
  Exactly, that's a much more elegant way of putting it. Simply removing components isn't a powerful enough technique to understand the meaning of the component.
  - irascible 10y ago
    I don't really understand what "meaning" means in this context, and indeed for the brain itself. I could see some sort of classifI cation scheme.. like "this group of neurons maps many to few outputs".. these "amplify signals from the inputs", as something valuable.. and indeed it seems that training an ANN to answere questions about functional units would be a good Avenue to explore. But asking for what it means is pointless, and seems more like an ontological question, for which the only answers are things like "it means that vision works" or "it means that slight imbalances in dopamine don't kill the brain". I think trying to understand how it works isn't as important as duplicating these functions in hardware in order to produce an oracle.

by jfoutz 10y ago

AstralStorm 10y ago

Sorry, but integral as done by a neural network is not generally reversible by taking out terms, unless it's a strictly linear no threshold neural net. This one is not. Same problem with the binary functions like most kinds of digital electronics - they're typically combined in highly nonlinear ways. However, knocking out a whole IC and seeing what it did is more reasonable and about the level we are at when understanding functions of subsystems in the brain.

jfoutz 10y ago
Exactly, that's a much more elegant way of putting it. Simply removing components isn't a powerful enough technique to understand the meaning of the component.
- irascible 10y ago
  I don't really understand what "meaning" means in this context, and indeed for the brain itself. I could see some sort of classifI cation scheme.. like "this group of neurons maps many to few outputs".. these "amplify signals from the inputs", as something valuable.. and indeed it seems that training an ANN to answere questions about functional units would be a good Avenue to explore. But asking for what it means is pointless, and seems more like an ontological question, for which the only answers are things like "it means that vision works" or "it means that slight imbalances in dopamine don't kill the brain". I think trying to understand how it works isn't as important as duplicating these functions in hardware in order to produce an oracle.