Jul 25

The Human Brain Project kerfuffle

Human-Brain-project-Alp-ICTWe’ve had a lot of controversy recently about the Human Brain Project, a European initiative that plans to spend a billion-plus Euros over the next few years to advance brain research.  To summarize ultra-briefly, a group of several hundred European neuroscientists have signed an open letter arguing that the Project is going in the wrong direction and needs to be redesigned; Henry Markram, the leader of the project, responded by saying in essence that the criticisms are misguided.  Because so many people have been waiting with bated breath to hear my opinion, I’d like to weigh in with a few brief comments.

First, it is striking that precisely the people who might be expected to support the Project most strongly are the ones who are most strongly criticizing it.  The goals of the Project can be summarized as providing computational tools for theoretical neuroscientists to use to create detailed computer simulations of the brain.  It might be expected that theoretical neuroscientists would love that idea.  But as a matter of fact the list of people who signed the letter is strongly weighted toward theoreticians, and includes many of the superstars of that community.  When the people who ought to have the most to gain are the strongest critics, it’s worth paying attention to.

Second, my own experiences lead me to agree with the criticism.  I don’t believe the HBP, as currently structured, is likely to give us anything of significant value.  I’ve done the sort of thing it is trying to do (on a vastly smaller scale), and concluded that the approach is misguided.

When I was a graduate student I spent several years trying to build biologically realistic computer models of the brain, or rather, of the dentate gyrus of the hippocampus.  I was able to create a model that replicated everything known about the electrophysiology of that brain area, but in the end I abandoned that approach as unproductive.  The basic problem is that biologically realistic models involve huge numbers of numerical parameters, and there is no conceivable way of setting values for them with the necessary precision.  Even my simple model of the dentate gyrus used over a hundred parameters; a sophisticated model of the cerebral cortex would need thousands.  And if even a few of them are wrongly set, there is a good chance that the model would fail to function correctly—would fail to carry out the right computations.

I continue to be a strong believer in the value of computer modeling, but I no longer believe in the bottom-up form.  By “bottom-up” I mean modeling that begins with detailed physiology and attempts to derive a theory from it.  “Top-down” modeling, in contrast, begins with a theory and attempts to implement it in a biologically realistic form.  Top-down modeling is valuable because it forces theoreticians to be think concretely, and because it gives insight about why the physiology takes a particular form.  But the goal of the HBP is to promote bottom-up modeling.

Henry Markram and his colleagues counter that brain simulation is not really their primary goal—that their direct aim is to create computer tools and databases of information about anatomy and physiology.  But there are grave doubts that databases of that sort are going to be useful.  It might be tempting to compare the idea with the databases the geneticists have formed and find very valuable, but there is a big difference.  DNA sequence databases are valuable because DNA is data.  DNA is precisely the data that a cell passes on to its descendants, and DNA databases transcribe that data with very high fidelity.  In contrast, the material that goes into neuroscience databases is largely arbitrary, and so is the structure imposed on it.

Consider, for example, the Neuroscience Information Framework, an NIH project to create “a dynamic inventory of Web-based neuroscience resources: data, materials, and tools accessible via any computer connected to the Internet”. The project has been running for years, and as far as I can tell hardly anybody finds it useful.  What reason is there to believe that the HBP will do better?

Let me try to put this in perspective.  I have written previously about the American BRAIN initiative, explaining why I have concerns about the direction it is likely to take.  But there is a major difference between BRAIN and HBP.  The BRAIN initiative, if properly focused, has the potential to give us revolutionary advances.  If not so focused, it will still direct resources to a range of valuable projects.  In contrast, the HBP, given its current focus, is likely to accomplish nothing of significant value.  Unless it is restructured, it is likely to be a mere waste of money.  If it can be revised into a project that directs resources to a range of valuable goals, even if they are not revolutionary, that will still be a substantial gain.  As far as I can tell, that is what most of the people who signed the open letter are trying to accomplish.

Older posts «