There’s a fair amount of excitement on the Internet about efforts to make an artifical rat hippocampus. This idea strikes me as, well, pretty weird. I am a little doubtful as to whether it could work (I can think of a lot of reasons why it wouldn’t) but to be honest, it doesn’t matter whether it works or not; the point here is that they have developed a way to interface a chip into a brain and do some decent data processing.
Some problems with this project:
1) I suspect there are probably hundreds of millions of neurones in the hippocampus. At the very least a million. Thus they will need the same number of connections. I would be amazed if they managed to create a chip that could hook up all the connections. But if I’ve read the article properly, they aren’t even trying to preserve the exact connections between the neurones, and I don’t think that’s a good idea.
2) The lab that I work in at Cambridge does a lot of recordings from neurones in the rat. We only try to record from one neurone at a time, and we know what we’re doing. Despite that, we still have problems with the signal to noise ratio when listening to neurones. Often we’ll have to do some serious waveform analysis to get the signal out properly, and that takes a lot of processing power. Recording accurately from millions of neurones simultaneously, online – I’m not saying it’s impossible – but it is years, if not more than a decade, beyond our current capabilities.
3) “They had to devise a mathematical model of how the hippocampus performs under all possible conditions…” I’ll put it this way. The scientists themselves state that no-one knows how the hippocampus works. Even by stimulating it with electrical signals, ‘millions of times over’ is not enough to make a proper mathematical model. And never mind how the hippocampus works, we don’t know how neurones work.
4) Neurones do more than just send electrical signals. There’s a very complex interplay of chemical signals, ion channels and lots of other stuff, most of which we don’t know about. Even if it were possible to mimic the electrical signals of a hippocampus on a chip, you’d still be missing everything else.
Still, like I say, it doesn’t matter if it doesn’t work; it’s always useful to do research on chips that can interface with neurones. However, I don’t think it’s realistic to expect this to work any time soon.
mssv 