Hacker News Reader

Comment by bbor | Hacker News Reader

This is a funny offhand comment, but I think indicates a common mistake in this kind of computer-as-brain thinking that has lead to most modern thinkers (or at least most modern HNers) dismissing it: being lazy with the scales. Just because the brain is composed of connected cells doesn’t mean that those individual cells are the right place to start any analysis. Real neuroscience, AFAICT, has been pointing away from “each neuron is a step in a process” and pointing towards “each neuron is a small part of many different structures” for a while now

View on HN

Yes, for instance, there's a lot more going on in the axons and dendrites than was even initially thought.

My favorite example was neurons in your eye that see something and send signals out laterally so when they (after a specific time-delay) meet another neuron that is currently seeing the thing, they combine to trigger a signal that can be directly interpreted as movement in a specific direction at a specific arc-speed.

Just one of the ways visual information is heavily processed before traveling along your optic nerve.

When I worked in a neuroscience lab (although I'm certainly no neuroscientist and this was decades ago, so grain of salt), the "aha" moment for me came when the researchers told me that neurons are not like electrical circuits, in that meaningful signals don't flow through them like electrical signals through a wire. Processing is about patterns of electrochemical flow throughout large neural masses.

My PI said that a still faulty, but better, analogy is that processing in a neural mass is more akin to the ringing of a bell. It's about waves flowing through a medium.

Isn’t that true for transistors as well?

Transistors have some nuance to them, but a transistor can in theory be scaled all the way down to being made out of only a handful of atoms if it was made perfectly with no impurities, and it would still function nearly identically to the slightly larger versions. Of course, packing them tighter still has some problems due to quantum tunneling at the scale of handfuls-of-atoms, but that's a separate issue.

Meanwhile a brain cell is still a lot more complicated and has a place in an entire network of cells and the connections between them that cannot be described by the comparatively simpler equations one can use for a transistor.

I almost failed Computer Architecture class back in the day so I might be off base here, but I understand most transistors to be organized into discrete functional units - this part forms a register, this part a control layer, this part an adder, and so on. These units are organized hierarchically of course, but the whole system is based around the neat compartmentalization of transistors based on function. I’m not a betting man, but if I were, I would put all my chips on that not being the case for the brain.

Thus, my complaint of trying the (IMO) lazy approach of “the smallest unit of the computer — the transistor - will be equivalent to the smallest unit of the brain - the neuron”.

Reply on news.ycombinator.com