The most complex object known to humanity is the human brain – and no only is it complex, but it is the seat of one of the few natural phenomena that science has no purchase on at all, namely consciousness. To try to replicate something that is so poorly understood may therefore seem like hubris. But you have to start somewhere, and IBM and the Ecole Polytechnique Federale de Lausanne (EPFL), in Switzerland, propose to start by replicating “in silico”, as the jargon has it, one of the brain’s building blocks.
In a partnership announced recently the two organisations said they would be working together to build a simulation of a structure known as a neocortical column on a type of IBM supercomputer that is currently used to study the molecular functioning of genes. If that works, they plan to use future, more powerful computers to link such simulated columns together into something that mimics a brain.
In a real brain, a neocortical column is a cylindrical element about a third of a millimetre in diameter and three millimetres long, containing some 10,000 nerve cells. It is these columns, arranged side by side like the cells of a honeycomb, which make up the famous “grey matter” that has become a shorthand for human intelligence. The Blue Gene/L supercomputer that will be used for the simulation consists of enough independent processors for each to be programmed to emulate an individual nerve cell in a column.
The EPFL’s contribution to the Blue Brain Project, as it has inevitably been dubbed, will be to create a digital description of how the columns behave. Its Brain Mind Institute has what is generally regarded as the world’s most extensive set of data on the machinations of the neocortex – the columns natural habitat and the part of the brain responsible for learning, memory, language and complex thought. This database will provide the raw material for the simulation. Biologists and computer scientists will then collaborate to connect the artificial nerve cells up in a way that mimics nature. They will do so by assigning electrical properties to them, and telling them how to communicate with each other and how they should modify their connections with one another depending on their activity.
That will be no mean feat. Even a single nerve cell is complicated, not least because each one has about 10,000 connections with others. And nerve cells come in great variety – relying, for example, on different chemical transmitters to carry messages across those connections. Eventually, however, a digital representation of an entire column should emerge.
This part of the project is expected to take two to three years. From then on, things will go in two directions simultaneously. One will be to “grow” more columns (the human brain contains about l m of them) and get them to interact with one another. The second will be to work at a more elementary level – that is, to simulate the molecular structure of the brain, and to look at the influence of gene expression on brain function.
Assuming that the growth of computing power continues to follow Moore’s Law, Charles Peck, the leader of IBM’s side of the collaboration, reckons it should be feasible to emulate an entire human brain in silico this way in 10 to 15 years. Such an artificial brain would, of course, be a powerful research tool. It would allow neurological experiments that currently take days in a “wet lab” to be conducted in seconds. The researchers hope, for instance, that their simulated brain will reveal the secrets of how certain psychiatric and neurological disorders develop. But that is probably not the real reason for doing it. The most interesting questions, surely, are whether such an artificial brain will be intelligent, or conscious, or both.