The European Commission has announced funding for the Human Brain Project, a massive simulation project whose goal is no less than the detailed simulation of a complete human brain. The project, expected to last 10 years and cost over 1.2 billion euros (1.6 billion dollars), will be coordinated at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland. It’s the maturation of the Blue Brain Project, which for the last ten years has been striving to simulate a small section of rat neocortex, with some very interesting results.
The official EPFL press release includes a decent introduction to the project, along with a 7-minute overview video, which you can also watch below.
This is a very ambitious project, and given the sheer scale of it, there are of course mixed views within the neuroscience community.
My own background, as it happens, is in neuroscience (M.S., 1997, UCSD). So, since everyone else has an opinion, here’s mine:
This is one of the most important scientific projects since Watson & Crick unraveled the structure of DNA.
Nervous systems are the most complex structures in nature, and although neuroscience is now a huge field pursued by many thousands of skilled scientists, almost all of them are forced to focus on one tiny piece of the puzzle. This makes it very difficult to say how all those pieces come together to produce emergent phenomena like learning, reasoning, and consciousness.
The Human Brain Project is the first realistic attempt to address those issues, by simulating an entire brain in detail. Some argue that it applies too much detail, but I think it’s too early for us to know reliably which details are important, and which we can safely abstract away. Starting with full-resolution simulations, with all the detail we can gather, is the conservative approach. We can decide what to simplify later, once we have full-detail benchmarks for comparison.
So, this project will integrate data and simulations from neuroscience labs all over the world, and bring them together in a constructive way like no other project can. That’s good in itself, but it’s not the really big prize. The really big prize is this: because of the way this project will be gathering and synthesizing all that data, and making it available in an open, standardized format, real neuroscience research will no longer require an expensive lab, with 2-story-tall electron microscopes, vats of liquid nitrogen, and small armies of lab techs. Anyone with a computer and enough passion about the subject will be able to delve into the Human Brain Project, run virtual experiments, trace out circuits, and uncover the computational principles that have been tuned by half a billion years of evolution. Much of neuroscience research will be transformed from an expensive lab discipline into a cheap information discipline. Like all information disciplines, it will undergo exponential growth.
So, bringing this doggedly back to robotics, I expect our robots in the next decade or two to start using algorithms lifted more or less directly from the human brain, based on this research. These will be simple sensory or motor-control tasks at first, but they’ll get steadily more sophisticated. This systems neuroscience approach to discovering algorithms will start with the very detailed model of the brain, and then extract useful algorithms we can reimplement in our traditional computer hardware.
At the same time, some researchers will push the model in the opposite direction. Starting with the generic, combined-from-many-sources model from the Human Brain Project, they will work out methods to build a one-to-one mapping between a particular, individual brain, and a functionally equivalent simulation. This is the path that leads to mind uploading, and all that implies.
So, in more ways than one, we’re going to look back at the Human Brain Project as the start of some truly world-changing breakthroughs. And you can say you read about it in 2013, on a little blog called BotScene.net.