Supercomputer that simulates entire human brain will switch on in 2024 | Hacker News Reader

Supercomputer that simulates entire human brain will switch on in 2024

newscientist.com

31 pointselromulous15 comments

It's hard to believe that it can simulate the billions of neurons carrying out trillions of signals

but it is a bit bigger than a brain, so which one is more believable?

It's really a question of fidelity. A typical brain has on the order of 100B neurons and 100T connections. This machine can perform 228T "synaptic processes" per second so it has reached the right ballpark. A synapse can fire more frequently than once per second, so there are some orders of magnitude missing there.

The rest of it comes down to how complex their simulation steps are, and where they have losses in fidelity because they have not captured a realistic model of how the biological system works.

There is a lot of marketing hype in the title (its an SC announcement), but they can probably do something interesting with this system.

How can a classical computer simulate a quantum system?

Very inefficiently. That's not being glib you can simulate them just fine it just takes forever

I'm not sure that there's consensus among neuroscientists and electrical engineers and what-have-yous (who are presumably the people doing the project) that the influence of quantum effects would be any more significant than they are in a field effect transistor. Of course, the project could prove them wrong, but what a remarkable result that would be, to falsify a significant hypothesis with your first attempt.

Huh, today I learned "falsify" has two meanings!

why do you assume that the brain is quantum? it seems very obviously not? on stuff the size of neurons quantum effects are already negligible at anything except near 0K temperatures which the brain isn't at, right? as this ever been shown to not be true. have we ever seen quantum effects on warm super-mulecular sized structures?

The quantum considerations have more to do with the molecular level. There are definitely quantum effects and considerations for what's happening at the level of enzymes and chemical reactions and they might matter with neurotransmitters at the synapse, and if they matter those effects could be amplified to neuron scale electrical impulses. There could be quantum effects that play a key role in sensory perception (we're talking the level of photons and molecules here too). You have to remember that biology is incredibly ordered and very physically small things matter so all these effects are on the table. Do they actually matter? Unknown it's complicated (outside of a few enzymes and systems where basic things like tunneling are clearly happening)

This is not correct. It’s been proven using electron microscopes that generic photosynthesis uses a quantum mechanical process inside every chloroplast. As a photon lands inside the microstructure an exciton hops through a superposition-network of chlorophyll molecules until it finds the low energy reaction center where it splits a water molecule into charged ions, used in further processing. This photon to ion process is nearly 100% energy efficient which would be totally impossible without the biochem exploiting quantum effects at room temperature.

I believe that it is very hard to capture these types of effects because measuring quantum systems impacts them, but I think (fully speculatively) it’s a matter of before we find more and more biochem systems that exploit quantum mechanical effects.

My first thought was that the Swiss-based Human Brain Project finally launched. Nope, they were just an oven for burning EU money.

Paywall free: https://archive.ph/QbDvz

Interesting. Some years ago yet another brain simulating project was announced by ETH Lausanne with a lot of media attention (https://www.epfl.ch/research/domains/bluebrain/). Goals seem to partly overlap. Once the media hype is over, it seems that such projects go quiet very quickly. I don't think that progress can be expected any time soon.

The only progress it's supposed to sport is progress in burning public money. Only some disconnected politicians will buy into the idea "this computer will be working like a human brain". Then some servilist academics will make a tiny career and some money out of it and that's about it.

I thought we were still orders of magnitude away from that?

Reply on news.ycombinator.com