Jul 17 2014

European Commission Human Brain Project Hubbub

In 2013 the European Commission awarded $1.3 billion to a project to simulate the human brain in a supercomputer. While everyone is excited about this prospect, and welcomes the infusion of cash, recently the project has come under public criticism.

More than 180 neuroscientists signed an open letter criticizing the way the project is being managed. The letter states:

“We believe the HBP is not a well-conceived or implemented project and that it is ill suited to be the centerpiece of European neuroscience.”

There appear to be two main points to the criticism – the first is that the money is largely going to computer scientists to create the software that will simulate the human brain. The neuroscientists complain that while the project is being sold to the public as a neuroscience project, in reality it is an IT project.

The problem here, they argue, is that we have not yet sufficiently mapped the connections within the brain to model them. The project, therefore, as currently constructed, is premature. This is certain to lead to failure and create bad press for European neuroscience.

They argue that the project should be funding the neuroscience that will create the data that the programmers will need for their simulation. Defenders argue that the neuroscience is already happening. The project was always meant to be an IT project to simulate the brain, not fund basic neuroscience.

The neuroscientists may have a point, in my opinion. Providing a large amount of funding for research can be tricky business. The potential problem is that the funding, if it is targeted, is like putting a giant thumb on the scale. Researchers need resources to conduct research, and therefore they will follow the funding. In other words, instead of researchers spending their time and effort on projects they find interesting and plausible, they will spend their time and effort on projects that are funded, even if they are not ideal.

By funding a specific goal prematurely, especially with such a large amount of cash, the project can create a massive waste of time and energy.

We often see this in medicine, where grassroots or patient groups raise money to fund research to find a cure. This could prematurely shift researchers into researching treatments, rather than asking more basic questions, and shift the balance away from what is optimal. It actually might slow progress toward a cure. The eventual failure and disappointment is also a PR disaster.

The second criticism comes from cognitive neuroscientists. The issue here is whether or not the brain project should approach the simulation problem more from the bottom-up or the top-down. In other words, should they focus on individual neurons and basic brain architecture (bottom-up), or should they focus on how the brain works and is connected in relation to the higher cognitive functions (top-down). This is a bit of a false dichotomy, but the question is a matter of emphasis.

The cognitive neuroscientists feel they are being left out in the cold and their contributions to the project are being neglected.

With 1.3 billion dollars on the table a cynical person might view this all as a fight over funding. Everyone thinks their contribution to the project deserves a larger slice of the pie. But even if you endorse this view, it is a fight worth having. Let everyone make the best case they can for their view and allocate resources accordingly.

This gets to the heart of the criticism, that the project manager, Henry Markram, is not doing that (or at least not well). The letter calls for a more transparent and open process of setting priorities for the project.

The leaders of the project have also recently issued a public response. They begin:

The members of the HBP are saddened by the open letter posted on neurofuture.eu on 7 July 2014, as we feel that it divides rather than unifies our efforts to understand the brain. However, we recognize that the signatories have important concerns about the project. Here we try to clarify some of the main issues they touch on. We also invite the signatories to discuss their concerns in a direct scientific exchange with scientists leading the HBP and its subprojects.

They also emphasize that in January 2015 the project will undergo its next annual review. This review is likely to get a lot of attention.

Conclusion

The Human Brain Project is certainly very exciting. The recent hubbub is unremarkable in that such conflicts over funding and priority are common, it is simply magnified by the amount of funding of one specific project. The discussion seems healthy, however, addressing reasonable questions and concerns, and calling for transparency and dialogue.

I would not minimize it as “just” a fight over funding – questions of funding are largely scientific questions.

It is also worth pointing out that the overall vision of the project is a good one. The next 50 years or so are likely to be a time of remarkable progress toward this very goal – simulating the human brain in a computer. It seems to me that the effort to map and “reverse engineer” the brain is working symbiotically with the effort to simulate the brain. Each project informs the other.

Of course we need knowledge of the brain in order to simulate it, so the IT end needs the neuroscience. But also, simulating the brain, or even specific components of the brain, is a great way to ask questions about how the brain works.

In the end, it’s all good.

Share

10 responses so far

10 Responses to “European Commission Human Brain Project Hubbub”

  1. backgroundNoiseon 17 Jul 2014 at 10:35 am

    Will super brain want to learn how to play the guitar?

  2. FacelessManon 17 Jul 2014 at 12:41 pm

    Why start with the human brain project and not the fruit fly or at least the mouse brain project. Considering a bottom-up aproach I am sure that even the fruit fly brain would be quite a challenge to simulate. It should be much more feasible than simulating a human brain, because it is simpler and we can probably get much more information from lab experiments that simply can´t be done on humans. This should give a solid foundation to move on to more complex organisms and eventualy humans.

    Neuroscience isn´t my field so perhaps I am underestimating our capabilities, but to me this project sound like we are sending a man to the moon before making certain that we can even send a satelite in orbit.

  3. Steven Novellaon 17 Jul 2014 at 12:56 pm

    FM – the project may, in fact, take that path. I strongly suspect we will simulate a mouse brain prior to doing a human brain. For one thing, it will require much less computer power.

  4. Karl Withakayon 17 Jul 2014 at 4:32 pm

    As with genome sequencing, they should start with the 302 neuron nervous system of C.elegans

  5. The Other John Mcon 18 Jul 2014 at 7:29 am

    Karl, C.elegans has been fully mapped (completed connectome) for quite some time now:
    http://en.wikipedia.org/wiki/Connectome

    But I can’t figure out if anyone’s bothered to actually simulate its function with the computational model of it.

  6. FacelessManon 18 Jul 2014 at 7:33 am

    Dr. Novella, thank you for your reply, that would make sense.

    I agree that C. elegans would be a good start. I also thought in terms of genome sequencing, although I would suspect that the step from smaller to bigger brains will be harder than from smaller to bigger genomes.

  7. mumadaddon 18 Jul 2014 at 7:35 am

    Is there a target date for completion of this project? ie. when they expect to have a fully functional artificial brain?

  8. Mr Qwertyon 18 Jul 2014 at 11:23 am

    > ie. when they expect to have a fully functional artificial brain?

    “fully functional artificial brain” – nothing less, eh? :P

    If you think about it, that is neither their stated goal, or a well defined one.

    To begin with, whose brain (what animal), as people already commented? “fully functional” also covers broad spectrum – does it have to simulate “just” the neural network or also hormone-based chemistry and modifiers or any other of the known (and unknown) interconnected systems?

    I don’t know nearly enough to judge well spent this money will be, so I’ll trust the experts (which is why I read this blog, heh), but I’ll be more than happy even if all it yields are some completely unexpected discoveries or applications in the fields of AI or computation.

    For example, I want a house cleaning, Roomba-style, spider-like robot that can climb all over the room interior, controlled by the AI somewhere in the “cloud”. What could possibly go wrong with that? :D

    But seriously, AI quality is the limiting factor in automation in many applications, industrial or other.

  9. hardnoseon 21 Jul 2014 at 7:11 pm

    They should start with something that doesn’t even have a brain, like a bacteria. That would be challenging enough.

  10. The Other John Mcon 23 Jul 2014 at 3:13 pm

    There is a special issue about “general artificial intelligence” hot off the presses, from the Journal of Experimental and Theoretical Artificial Intelligence:
    http://www.tandfonline.com/toc/teta20/26/3#.U9ADxWYpCPQ

    Looks like some very interesting articles, relevant to this discussion.

Trackback URI | Comments RSS

Leave a Reply

You must be logged in to post a comment.