Saturday, August 16, 2008

Robot brains closer than Sundin decision?

Back from vacation a week ago, and now trying to learn an important lesson about writing - I've been trying to work on a piece about personal genomics for a while now, but it just won't materialize. I have been dreading trying to get it done, because it's discouraging to look at something you put effort and thought into and you end up hating. It occurs to me that it doesn't matter, because I can write something else. I'm sure there has been enough interesting stuff published in the last 2 weeks. Although none of it is about Sundin, which is getting pretty irritating. We can try to laugh it off with this gem, forwarded by a friend (thanks, CT). An illustration of why science education is important to everyone (parents: please don't let your kids grow up to be like this woman). What kind of crazy, awful thing must we have done to our Earth to make some substance come out of the ground and make a rainbow in the sprinkler on my front lawn?!!! We all know that didn't happen 20 years ago!

Sadly, I wish I could say that this is probably a unique case, but I don't believe that for a second.

* * *

Bits of news - The FDA has dismissed claims that the trace amounts of bisphenol A (BPA) that leach into our drinks (and into baby bottles) are in any way harmful. BPA is some pretty nasty stuff, but then so are plenty of things we find in trace amounts in our environments. Still, I appreciate the Canadian government's willingness to err on the side of caution and ban its use in products for children.

A shameless plug for a study out of the Institute where I work: Mothers that were pregnant during the huge ice storm of 1998 have kids that are developmentally delayed 5 years later. The babies are slightly smaller, and have a slightly lower IQ than average. On the bright side, other studies have shown that these kids will probably "catch up" to their peers by adolescence. Interestingly, the way that the mother responded to the stress didn't seem to matter - just the presence of the stress itself seemed to impact the development of the child. The primary investigators will continue to follow the children through their lives.

* * *

Frickin' robot brains, man!!! They're getting closer and closer. This is upsetting because I will lose one of my euphemisms for human error/incompetence in the lab ("wetware" problems), but actually really cool to see. This video shows a guy who is using randomly-assembled networks of actual neurons to drive a robot to learn about its environment. I really can't wait for this guy to publish, because this is nowhere near detailed enough for me. He might even be something to consider investing in, once he has a startup. If I were to get back into electrophysiology, this exactly the sort of thing I'd love to do.

In a diametrically-opposed approach, another group has a computer model of something like a brain. This model is "given" a virtual body that interacts with its virtual environment, and the neural network figures out how to move itself using the virtual stimuli and the joints and limbs the programmers give it. What is most interesting here is that structure dictates function, and the "robots" just learn to move what they have rather than needing to have complicated steps programmed into them. This will be a huge step in designing computer games at least, but probably in artificial intelligence and robot tech in general.

No comments: