Let’s build a better brain
Guest post by E. Paul Zehr
Or should we first see if we can build any kind of brain at all? On the surface it seems like an almost trivial exercise. All you need to do is figure out how the brain functions, then run some computer simulations, use the outcomes of the simulations to create fully detailed models, test and retest the models with machine learning algorithms over many, many iterations, and then make a brain based on the successful outcomes.
So, pretty simple, then? There are some complications that make this idea, to borrow a bit of physics/engineering/mathematics jargon, a “non-trivial” problem. The main thing I want to talk about has to do with scope and size.
The cool thing about most of the body is that you can tell a lot about physiology (how it works) from the anatomy (how it looks). Function comes from form. In your cardiovascular system you’ve got a big muscular pump in the form of the heart that receives and pushes blood all around the body. Taking a good look at the heart along with all the piping coming in and out, allows a reasonable estimate of what it does and how blood flows in the body.
A real human brain contains about 100 billion neurons (the cells of the nervous system). Those 100 billion neurons might have on average 5000 connections from other neurons making synaptic connections with it. That means about 100 trillion connections. A pretty big number. Far bigger than the number of galaxies in the universe (estimated to be between 200 to 500 billion). Overall this is a huge number of connections to model.
This is part of what allows the nervous system to present with a much broader scope. Not because the anatomy is impenetrable or that much more complicated within different areas of the brain. It is certainly complex, but the general features of the connections from those 100 billion neurons form into tracts and bands of connections within the brain that can be reasonably identified (mostly).
The real non-trivial problem comes from the fact that the function—the behavior—of the brain cannot be directly predicted from anatomy. Enter those 100 trillion connections. The key thing is that the network activity in the brain emerges from the activity of whatever synaptic connections are active at any given time. It is a constantly shifting landscape of network activity.
A simple approximation is to imagine sitting in a boat that is rising and falling on the swells of the Atlantic ocean. Boats are all around you and you can see them rising and falling such that at any given moment you see different boats. Those boats all represent active connections between neurons that are expressed when you can see them and silenced when you cannot. To complete the metaphor, multiply by many trillions.
This is what makes building a brain such a daunting task. It’s not so much building something with brain-like connections, but rather a brain that functions like a real brain.
This is what makes the “Human Brain Project” such an interesting idea. This group is made up of institutions in Germany, the UK, France, Spain, Switzerland, Sweden, Israel, Austria and Belgium and is one of the finalists for a new EU program to create a “simulation of the human brain – an achievement that promises to revolutionize not only neuroscience, medicine and the social sciences – but also information technology and robotics”.
The focus of this project, clearly named to draw on the cachet of the first biological megaproject in the Human Genome Project, aims to bring together data and databases on brain study, figure out organizational principles and then build models with as much detail as possible.
It’s important to realize that the scope and extent of this project has never been attempted before. That doesn’t mean it will be successful—lots of thing never attempted before fail when somebody tries it out. What it does mean is there’s a reason to be cautiously optimistic that some major new advances in our understanding of how we can understand may be just around the corner.
This brings me to close with one of my favourite neuroscience quotes. The South African zoologist Lyall Watson (1939-2008) wrote: “If the brain were so simple we could easily understand it, we would be so simple we couldn’t.” But we are going to give it our best shot. I look forward with great anticipation to the undiscovered country that will be revealed by this endeavour.
E. Paul Zehr is the director of the Centre for Biomedical Research at the University of Victoria, where he is also a professor of neuroscience and kinesiology. He is the author of Inventing Iron Man: The Possibility of a Human Machine and Becoming Batman: The Possibility of a Superhero. This blog post was originally published on the website of graphic novelist Warren Ellis.