Elf Sternberg

Neural-net analogies are poor because there's no evidence the brain is just a neural net. It could well be a cellular automata with a robust but distance-poor signaling system; in which case, transhumans will have to come up with a more efficient but equally robust signaling system to make sure that all parts of a mind stay in sync.

J…ein. The brain is, by definition, a neural net. Any collection of biological neurons is a neural net. That's why everyone makes sure to refer to the "artificial" when talking about ANNs.

As for signalling, I happen to also remember that biological neurons don't propagate electric signals instantly. If they did, synapses wouldn't fire, but sorta sputter, if they did anything. If neurons didn't act like capacitors, our brains wouldn't function.

Our neurons aren't analogous to the wiring on a CPU chip. They're more like combination wire/capacitor. The neurotransmitters crossing the synapses cause a change in voltage drop across the neuron.

Lately, I've been wondering if human cognition isn't a self-emergent property of … well all of the brain's variable aspects: neurotransmitter quantities, the individual ion-propagation time constants of each neuron, the number of connections between each neuron, and how a given set of signals to each neuron are summed to produce an internal voltage change.

The notion that a flaky memory "isn't physiological but due to our brains getting full" is something of a misstatement. All that says is that the physiology is insufficient.

:)

That brings us back to the analogy I made earlier. I'll admit, it has been 10 years since I last saw this stuff. But I do recall that ANNs are closely modelled after some experimental biology studies of living neural nets, and how those behave. That includes the whole "capacity measurement," which is determined solely by the number of neurons and the topology of their interconnections. It could, in principle, be used to measure the lower-bound capacity of biological neural nets, and not just ANNs. The problem in practice is: what's the base-unit of information stored by a BNN? For an ANN, it's bits, obviously.

But for BNNs, much less brains, much less human brains? Who even knows? We don't even know how to measure the information size of, say, the image of grandma's face. So how can we say how much space in your brain that image of Grandma occupies? ;)

Our brains are finite, nonetheless. So, their capacity … whatever the units of measure … is also finite. Now, I know your response, Elf: "Then, we need to find ways to enhance that capacity." True enough, and I'd love nothing more than to have an eidetic memory for the duration of my lifetime.

I just don't see it happening. If our best models of the brain, ANNs, are too crude, as you say (and I agree with that), then how can we hope to understand, much less modify, much less enhance the real thing?