SHOWDOWN

07.10.14

The Science Community’s Fight Over an Artificial Brain

The European Union just granted 1.2 billion euro to the Human Brain Project—a sort of Human Genome Project for the brain. But some scientists are boycotting.

The Human Brain Project (HBP) is a massive neuroscience research endeavor funded in large part by the European Union. Among its primary goals is to simulate a human brain using a vast network of connected supercomputers. Led by Henry Markram at the Swiss Federal Institute for Technology in Lausanne, it recently received a giant 1.2 billion euro grant from the E.U.

Markram insists that his research will bring insights into the human brain and reap untold benefits for neuroscientists. An ever-growing number of European neuroscientists, however, are boycotting his project. As of this writing, 379 researchers have signed a petition denouncing the HBP.

No one can accuse Markram of failing to dream big: He plans to collate all neuroscientific data in one place. The HBP, he argues, could revolutionize computer science, which will surely profit from insights into how the brain works. Perhaps the cornerstone of his project includes building a computer simulation of the entire human brain, which he says will reap huge social and economic benefits when neuroscientists and medical professionals can study it and learn to treat brain disorders such as Alzheimer’s, depression, and autism.

You might think other neuroscience researchers in the E.U. would heartily endorse such a potentially fruitful project, or at least be elbowing each other out of the way to get a piece of the 1.2 billion euro action. Why, then, would there be a public boycott of such a large project?

The first and most basic objection is that the boycotters do not believe that the HBP can possibly deliver on its enormous promises.

Neuroscientists can approach studying the brain on many levels. For example, one might study the brain at the molecular level, the neuronal level, or the cognitive/behavioral level, among others. Neuroscientists can take a top-down approach—that is, they can take a given behavior or thought pattern and try to explain it in terms of computations or neurons or molecules. They can also take a bottom-up approach by examining changes at, say, the molecular level, and correlate it with changes at the neuronal or cognitive/behavioral level. At this point in time, no one is really sure how changes in the molecular or neuronal levels of the brain can actually cause changes in cognition and behavior. Only correlations can be seen.

The cornerstone of his project includes building a computer simulation of the entire human brain, which he says will reap huge benefits when neuroscientists can study it and learn to treat brain disorders such as Alzheimer’s, depression, and autism.

But Markram has entirely eschewed the top-down approach. In fact, he has axed the sub-group of the project that was slated to work on the cognitive level. He plans to build his artificial brain from entirely the bottom up, starting at the molecular level.

“Those inside and those outside criticizing widely share the belief that computational approaches—modeling, simulation—are vital to making progress in understanding the brain. We just don’t agree that a massive bottom-up simulation is how to approach it,” explains Zach Mainen, a principal investigator at the Champalimaud Neuroscience Programme in Portugal.

The detail afforded by such a bottom-up approach might well be either useless or distracting to explanations of cognitive or behavioral phenomena.

For example, consider an old-fashioned wind-up watch. If someone unfamiliar with it picked it up and wanted to figure out what it was and how it worked, a detailed molecular diagram of the watch wouldn’t be necessary or even helpful. The function and structure of the watch only needs to be explained in terms of springs, gears, and wheels.

Because we know so little about how changes in one level of the brain cause changes in another level, there’s no way to even begin to build a bottom-up simulation of the brain, suggests Peter Dayan, professor at University College London. He tells The Daily Beast that while we could perhaps build an artificial brain that functioned like a human brain if we actually had a perfect description of the brain, “…it is simply ludicrous to think that one could get such a description at present. Single neurons, of which we have billions, are fantastically complicated devices, living in a rich and complex 3D environment.”

Dayan further points out that the human brain is plastic. Each of our brains is changed by the experiences we have, and of course we don’t have all the same experiences.

“Without understanding how this works, we don’t know what aspects of the fantastic complexity we need to capture; we don’t know how to combine the results of multiple experiments on different individuals (since they’ll all be at different states); and we won’t have captured the nature of how the brain is shaped to the world. To suggest that one could do this for a human brain, for which most experiments are precluded is just absurd,” he says.

Mainen agrees. He suggests that since scientists have not even successfully simulated the brain of a nematode, which has about 300 neurons, the E.U. should not pour 1 billion euros into trying to simulate a human brain, which has billions of neurons.

Boycotters also note that they are uncertain if Markram’s previous brain simulation project, The Blue Brain Project, performs any of the functions that a brain does.

Markram responded to his critics in a message sent to Gary Stix, a blogger at Scientific American: “This is such an exciting direction that can bring everyone together to take on this grand challenge. Just so sad that it gets torn apart by scientists that don’t want to understand, that believe second-hand rumors and just want money for their next experiment. For the first time in my career as a neuroscientist, I lose hope of neuroscience ever answering any real questions about how the brain works and its many diseases.”

The researchers protesting the HBP are indeed concerned about the amount of funding the project is receiving and the possible reduction in funds for other projects. “The HBP is one of, if not the, largest current science project in biology/medicine. It is being used to direct funding policies in neuroscience,” says Mainen. “We do not want to kill the HBP, but we cannot sit by and let this monster derail European neuroscience. The community is sending a strong message that this is not a project to unite us all. It is a non-responsive beast that is unwilling to compromise and can be of little or no service in its present form.”

Yet another reason for the boycott is the lack of transparency and oversight over such an enormous grant. The petition states, “It is stated that the review [of the HBP] must address the excellence, impact as well as the quality and efficiency of implementation. We believe that a review will show that there are substantial failures to meet these criteria, especially concerning the quality of the governance demonstrated and the lack of flexibility and openness of the consortium.”

Although the name of the Human Brain Project recalls the massive and influential Human Genome Project, skeptics do not believe that the two projects are comparable. Almost all of a person’s cells have identical genetic information, and it is easy to collect a sample of a person’s genome. By contrast, brain states are far more heterogeneous, likely due to interactions between a person and his or her environment.

“[The HBP] is radical stuff indeed and very much the idiosyncratic vision of an ambitious group. We have to step back and rethink the process that led to this if we are to prevent science from taking large steps backwards from ill-conceived mega-projects sold by false analogies with the human genome,” Mainen says. “[It] is ludicrous to make such a risky and unproven view the centerpiece of European science policy.”

Henry Markram did not return a request for comment at the time of publishing.