it is just bad editorialising. the real summary is just that most common mathematical models of neurons are generally not complex enough to model biology. (or even interesting behaviours, as any LSTM fan will hasten to observe)
as an aside, this is yet another reason why I like evolutionary ANNs more than typical workflows - this sort of detailed sub-structure can and does evolve without intervention.
I don’t think we know enough to conclusively say that yet. Structure and emergent complexity can often arise in simplistic systems. There’s a very good book about this called “Think Complexity” which is basically a small primer on complexity theory. Basically though sometimes similar levels of complexity arise in systems, even when the agents themselves are vastly more complex. I’m not saying you’re wrong, you could very well be right. I am saying we don’t have the theoretical framework to say you are conclusively right either.
Not really, though. The author is explicitly asserting that each neuron is not a simple, linear gate, but rather it’s capable of performing multiple different non-linear functions whose behavior is capable of being complexly self-modulated.
this is an extremely loose definition of computer
it is just bad editorialising. the real summary is just that most common mathematical models of neurons are generally not complex enough to model biology. (or even interesting behaviours, as any LSTM fan will hasten to observe)
as an aside, this is yet another reason why I like evolutionary ANNs more than typical workflows - this sort of detailed sub-structure can and does evolve without intervention.
I don’t think we know enough to conclusively say that yet. Structure and emergent complexity can often arise in simplistic systems. There’s a very good book about this called “Think Complexity” which is basically a small primer on complexity theory. Basically though sometimes similar levels of complexity arise in systems, even when the agents themselves are vastly more complex. I’m not saying you’re wrong, you could very well be right. I am saying we don’t have the theoretical framework to say you are conclusively right either.
Yeah, it’s more like a gate.
Not really, though. The author is explicitly asserting that each neuron is not a simple, linear gate, but rather it’s capable of performing multiple different non-linear functions whose behavior is capable of being complexly self-modulated.