If there was a complete model of the human brain, imagine all we could learn. We could perform experimental surgical procedures, we could study the effects of new drugs, or do anything else that needs human testing but is too dangerous for a real human.
That’s exactly what the Blue Brain Project is trying to do. According to them,
The Blue Brain Project is the first comprehensive attempt to reverse-engineer the mammalian brain, in order to understand brain function and dysfunction through detailed simulations.
In fact, the project claims that they may have a model of the human brain in the next 10 years.
That sounds really cool, but if you read further into the article, it gets a little harder to believe that this will be accomplished in 10 years.
For example, the project currently has simulated a rat’s 10,000 neuron Neocortical Column (NCC) in 2006. This is really impressive and shows that the project is progressing. It’s also very computationally expensive. To do this, the project used a Blue Gene supercomputer with 8,192 processors in it. Wow. That’s a lot of power!
However, the number of neurons in a human brain is probably around 100 billion. As in, 10 million times more than what they have simulated so far. I would try do the math to show how long it would take to reach this level with Moore’s law, but I think its pretty obvious that 10 million times in 10 years isn’t likely. Can you imagine how many computing resources this will take? Even with Moore’s law (which may not hold for another 10 years), this will become an astronomically expensive system to create and maintain (imagine the electric bill!). Almost to the point that I don’t believe it’s possible within 10 years.
Big Blue Brain has to have thought about this haven’t they? They have and they have a plan to address it. Their model is massively parallel. This means they can add more (relatively) low-end supercomputing resources to achieve almost a linear speedup. In this way, they will not have to buy the latest and greatest supercomputer to be effective, but instead can use slower, but cheaper, computers to do the job.
Also, the NCC that the project has simulated is very similar between rats and humans, we just have more of them than rats do. That’s pretty exciting since now it’s just a matter of simulating more NCCs to model a human brain. Will it be that simple? Doubtful, but it will be simpler to use the current NCC as a basic model, rather than having 100 billion unique systems all functioning at once.
An interesting question one could ask when the project succeeds, “Will it have consciousness?” The answering is somewhat chilling. If this system really does become conscious, it will be really important that this brain (robot?) have some sort of ethics. However, this may be difficult to do properly.
The Blue Brain Project really is a neat idea, and while it has had a lot of success simulating a rat’s NCC, it will be much more difficult to simulate the entire human brain for a few reasons.
- 100 billion neurons versus 10,000.
- Huge computing resources needed.
The project does have a few things working in its favor though:
- Humans mostly just have more NCCs than a rat, rather than completely unique NCCs.
- The model is parallel, so adding computers will give an almost linear speedup.
The fact of the matter is though that the computer system needed to do these computations will be way too expensive, despite the parallelism. There needs to be some sort of revolution in computing before the necessary resources become realistic. If Moore’s law follows, this project will see more gains in its computing power, but it is placing a tall order to get that power within 10 years.
So, is this project achievable? Yes, they’ve already shown that this sort of model can succeed. Will it happen in 10 years? I highly doubt it.