Jonah Lehrer's article about the Blue Brain Project is now on-line, so go have a read.

The project is an attempt to model the neocortex at a very high level of detail (down to the chemical level) using supercomputers. They've already successfully modeled a neocortical column of 10,000 neurons with 30 million connections.

This is an awesome project, and I look forward to seeing how it progresses, but a couple of things about it concern me. One, I think the model will be a wonderful tool for understanding cortical dynamics at a number of different levels, but I'm a bit more skeptical about how well it's actually going to be able to process input and generate meaningful behavioral output. They are building in neural plasticity, but I believe the overall approach is to build an adult brain. In other words, they are not modeling the developmental trajectory of the neocortex. Some theory suggests that the developmental process acts as a constraint on how much learning can occur at any given age. I'm not sure if these ideas are correct, but the analogy here would be to building an adult brain that then has to learn how to walk and talk from scratch. I'm not sure how viable that approach is going to be.

Secondly, I wonder if for the purposes of replicating neocortical activity, if the model isn't at too fine-grained a level of detail. If one goal is to distill the function of the neocortex into workable algorithms, modeling at the chemical level could be like trying to build an airplane by building a bird from scratch, cell by cell. It's more computationally expensive, and may be unnecessary. I don't believe that level of detail is necessary to replicate the most powerful aspects of neocortical processing, but hey, I could definitely be wrong.

Anyway, I still think the project is very interesting and exciting, and has a lot of things to teach us, so anyone interested in such things should definitely keep their eye on it.