So I read this morning about folks who are taking what seems to me the only sane path toward strong AI: making something that does what brains do, and seeing what happens.
I have a few reactions to this on the general subject of neuroscience and cognitive theories. The first is Woo hoo! I love science. I love the systematic discovery of how the universe works and what it does, and especially the counterintuitive parts where it turns out nothing is actually quite the way we thought. Love that. I suspect we’re going to find that brains are big pattern matching structures (or at least neocortexes) are mostly great big pattern matching-with-feedback mechanisms with no central processing unit, and consciousness is going to stay weird and inexplicable. We don’t have a rigorous model by which matter and energy can exercise will, and we’ve got a lot of examples of matter and energy coming together to exercise will. We call them people. Or sometimes dogs. The present models can’t account for that, and that’s cool. We’ll build a better model later, when we know more. That’s called science. If our present model accounted for everything, we’d be done.
Second, I am wildly tired of the “Oh my god, subject X did something new and it CHANGED WHAT HAPPENED IN THEIR BRAIN!” I’ve seen a lot of this recently. Just as a head’s up, that’s the expected value, folks. If you do something new, it changes what happens in your brain. Learn how to play piano? Changes your brain. Someone gets clocked by a bowling ball and their memory starts getting bad? The fMRI is going to show less activity in the hippocampus. That’s two ways of saying the same thing. If you get a study that shows someone doing something new and their brain staying exactly the same, then I’m interested.
Third, I know that “Oh, the machine became conscious” thing is done to death. Cliche. Boring. Even when it’s been done really well (and I’m looking at you, Galatea 2.2) it’s the execution of the story that makes it stand out not the idea. But one of the things we know — maybe the only thing we know for certain — is that matter is capable of displaying consciousness. Some configurations of matter and energy are able to experience pain, love, wonder, and the determination to by god lose a few pounds next year. Yes, at some point we may be able to replicate that. But the chances seem slim to me that whatever machine we build will have any deeper insight into the nature of the universe than we do. I imagine the first AI strong enough to contemplate the philosophical questions of the universe, lifting its shoulders and saying “Wow. Weird.”
Fourth and final, Ted Chiang has a few essays in Lady Churchill’s Rosebud Wristlet that should be required reading for anyone doing hard SF. Ted remains the best science fiction author.