Hacker News new | ask | show | jobs
by dharma1 3798 days ago
I wonder how general purpose biological neural networks are.

If we had a perfect brain-computer-interface, could we grow 3D brain tissue in the lab, hook it up to a computer and do arbitrary processing on it?

3 comments

Don't be misled thinking the brain "is a computer". It's not general-purpose, it's non-purpose. People often picture it as if evolution picked a "computation" or "intelligence" quality off the Platonic shelf and implanted it onto the brain.

An alternative way to view the brain is as a complex switchboard. Instead of doing stuff itself, signals just flow through it and reshape the vessel, like rivers shape the landscape. Of course this is also an incomplete view, but demonstrates at least one alternative metaphor.

Metaphors matter, they are central to how we conduct research.

Also, most of what the brain does isn't creativity, insight, curiosity and all these glorious things, but rather breathing regulation, hormone regulation, control of digestion, muscles, a lot of "body stuff" instead of "abstract, ideal, body-independent stuff". I'd actually argue that everything we think about (even things like morality) have a rooting in our bodily, real existence and nothing is purely abstract.

Check out the work of Steve Potter's group at Georgia Tech. They culture rat cortical neurons in a dish while recording and stimulating them in a closed loop system. They did some studies with using these hybrid systems to control robots IRL and controlling avatars in a virtual world. If I remember right, the avatars learn... but I've never read the studies closely
Neural networks in biology are general purpose. They have all the requisite parts: computation and memory.

Not sure what you mean about having a perfect BCI. You could grow 3D brain tissue and do arbitrary processing on it but it wouldn't require a perfect BCI to do that.

well, you would need a way to get the raw data into the brain tissue, and a way to get the results out, no?
Our disagreement is purely about nomenclature. If you grow brain tissue in a dish, outside of a brain, it wouldn't be a brain. I guess you'd need a "brain-tissue-computer-interface", which would be far simpler than a BCI.