My mention the other day of Japan’s Fifth Generation computer project prompted a reader to send along this link, which I thoroughly enjoyed. It concerns the Human Brain Project, currently being funded by the EU, and if you’re offended by procreating inanimate objects, you probably shouldn’t read it. But I feel the author’s frustration. The HBP is a ten-year effort, in its second year now, with a goal of doing brain and neuron simulation. Unfortunately, as the author correctly states, we have no reproducing way of knowing how to do that yet:
We can’t simulate the brain of C. Elegans, a very well studied roundworm (first animal to have its genome sequenced) in which every animal has exactly the same 302-neuron brain (out of 959 total cells) and we know the wiring diagram and we have tons of data on how the animal behaves, including how it behaves if you kill this neuron or that neuron. Pretty much whatever data you want, we can generate it. And yet we don’t know how this brain works. Simply put, data does not equal understanding. You might see a talk in which someone argues for some theory for a subnetwork of 6 or 8 neurons in this animal. Our state of understanding is that bad.
And things don’t get any better at the other levels that this project can be approached from:
. . .microscopically, we have no clue. It looks pretty random. We collect statistics (with great difficulty), and do tons of measurements (also with great difficulty), but not on humans. Even for well studied animals such as cats, rats, and mice, it’s anyone’s guess what the fine structure of the connectivity matrix is. As an overly simplistic comparison, imagine taking statistics on the connectivity of transistors in a Pentium chip and then trying to make your own chip based on those statistics. There’s just no way it’s gonna work.
That’s actually a very good analogy – and remember, as I always like to point out, that computer processing chips were designed by humans, and are thus far, far easier for humans to understand. Your chances of success with that statistical approach to a Pentium chip are not good at all, but they’re a lot better than the chances of it working on brain tissue. This brings to mind the famous “Can A Biologist Fix a Radio” paper (which I’ve also referenced here, here, here and here), which made almost the exact same point, imagining teams of biologists taking radios apart, one after the other, and cataloging the size, shape, and color of all the parts. It’s a telling point. (Where I part company with that paper, though, is its suggestion that there are clearly better ways to do such analysis that the biologists are ignoring – I think that the situation is just as bad as described, but I don’t see a clear remedy for it yet myself).
The (anonymous) author of the brain project rant finishes up thusly:
So, the next time you see a pretty 3D picture of many neurons being simulated, think “cargo cult brain”. That simulation isn’t gonna think any more than the cargo cult planes are gonna fly. The reason is the same in both cases: We have no clue about what principles allow the real machine to operate. We can only create pretty things that are superficially similar in the ways that we currently understand, which an enlightened being (who has some vague idea how the thing actually works) would just laugh at.
Exactly so. The same goes for too many other beautifully rendered simulation, but it’s especially true for the central nervous system. The brain is the darkest of the black boxes. That’s not a reason for despair, though. We’re going to learn an incredible amount as we start opening it up, but we’ve just barely started. Beware anyone who tells you otherwise.