Hacker News Reader

Comment by dcow | Hacker News Reader

Not sure if that analogy holds exactly. Neurons don't “compute”, they simply emit a signal that depends on how “saturated” they currently are, right?

View on HN

Neurons definitely compute. The dendrites that aggregate signals to initiate an action potential are not just simple summations but complex arrangements of signal promoters and inhibitors. The complexity of a pyramidal neuron is equivalent to a 5-8 layer artificial NN.

https://www.quantamagazine.org/how-computationally-complex-i...

https://www.sciencedirect.com/science/article/pii/S089662732...

And that result is just from a computer simulation of a neuron that focuses on the dendrites and doesn't appear to extensively model the nucleus.

"Unfortunately, it’s currently impossible for neuroscientists to record the full input-output function of a real neuron, so there’s likely more going on that the model of a biological neuron isn’t capturing. In other words, real neurons might be even more complex."

That simplifies neurons a bit too much. There are a variety of neuron types including mirror neurons and pyramid neurons. These can feature more or less branching even.

My point is they operate more like gates. They aren’t mini Turing machines.

neurons have gates, these gates interconnect and form logic networks.

these are histeretic, programable, and dynamic.

the neuronal body state, the electro-osmotic environment, the past history of state are primary effectors of structure,and function resultant in the logic.

You are correct. Neurons are not Turing machines. However, they are complex enough that they can be Turing machines so it can be fun to call them that.

I didn’t expect this topic to be so controversial to be honest. I’m not surprised though.

I fully admit I may be underinformed. I had thought, though, that neurons were not sophisticated in the way a Turing machine is and much more akin to memristors. However, that may be a memristor tinted view of the world.

Feel free to research it. The ‘saturated’ analogy is overly simplistic and elides the incredible complexity of what’s actually going on.

But neurons certainly aren’t mini Turing machines, either? Happy to inform myself more if you can point me in the right direction.

Cells perform a staggering number of parallel computations. Neurons are cells, and perform these same calculations and more in addition to emitting spikes. Spikes, rather than being the major form of computation, appear instead to be a means of maintaining synchronization.

I quite honestly did not think of it this way. But it makes some sense. I always thought cells perform, for lack of a better term, self-maintenance operations independently from their macro function (the receiving and emitting of spikes, in this case). I did not think of the internal operations of a cell being connected to the macro function in any sophisticated or meaningful way. I would love to inform myself better on this topic since it is not by area of expertise, and I have an admittedly cursory understanding.

I do think the field of psychopharmacology is probably sufficient to give an appreciation of the complexity involved at the same time as being useful to the practical understanding of how medications (and food) change behavior. It's also well studied and there are good books. I think Dr Stahl's books are the university standard. Like a lot of science though I think much of it is 'current best guess'.

For information on the modifying DNA expression in neurons that's under epigenetics and not as well studied.

Reply on news.ycombinator.com