European Project on the Human Brain : 3 Questions for Dr Grégoire Hinzelin, the Neurologist


What exactly is the European project on the human brain ?

The project on the human brain was launched in Europe at the end of 2013 for a period of 10 years. Its initial ambition was aimed at numerous research themes, at the point of convergence of several disciplines: neurology, medical computing, information and communication technology. In neurology, the aim is to establish an atlas as detailed as possible of the brain, to enable more accurate medical treatments. Better modelling of the operating mode of the brain is also expected: this is useful both for medical research and for the technologies which aim to emulate that operating mode in computer science.

One of the aims is for example to work on an Artificial Intelligence replicating the brain in the hope of achieving energy saving. It is known that the brain is far more efficient in terms of energy consumption than computers for the processing of information and for calculations. AlphaGo, by DeepMind (a subsidiary of Google) which in 2016 beat the world champion of Go, consumed 20,000 watts per day, whereas a human brain consumes between 20 and 40 watts! The human brain combines both the processing of information and storage in its network of neurons, whereas computers are designed by separating the processing units from the memory. This research to find energy-saving solutions would be applicable to numerous cognitive tasks: image and voice recognition, as well as for drones, satellites etc. In order to recognise a chat, a child requires 12 iterations, an AI machine currently requires some 15,000.

Another objective was to obtain the most complete as possible atlas of the brain, in 3D, to be made available to the research teams.

The project was granted 607 million euros of European funding, and the collaboration of more than 500 research scientists in 19 nations.

What are the achievements and advances of the project?

The detailed mapping of the brain has not modified our current understanding of its operation but it has enabled its considerable refinement. This mapping will enable for example more precise operations on brain tumours. Similarly for cerebral re-education which is dependent on the plasticity of the brain.

Measurement of consciousness may also be refined. Following serious brain lesions, the patient may be declared unconscious, but that diagnosis is not always correct, because certain patients may be conscious but incapable of showing it.

In France, research scientists have developed customised brain models of epileptic patients who do not respond to medicines. Such virtual brain models help to identify the zones of the brain where the convulsions appear.

At the intersection between neuroscience, robotics and computing, a project like SpiNNaker seeks to replicate the functioning of the human brain. The difference compared with super-computers, is that the brain does not have powerful calculation processors but small integrated processors like in mobile telephones interconnected between each other, in order to achieve the high degree of connectivity of the neurons within the brain. According to one of the directors of this project, Professor Steve Furber: “SpiNNaker allows its users to explore hypotheses and theories on the functioning of the brain. Because the way in which the brain functions as a processor of information is still a mystery for science, and it is one of the major challenges of neuroscience to attempt to start to use convincing explanations on the way in which the brain conducts its work. But until such explanations become available, science advances by putting forward theories and subsequently by testing those theories, and computer models are a good means of testing theories”.

What are the limits of this type of project?

The dream of modelling the thought process remains a utopia which cannot currently be achieved.

For the SpiNNaker project for example, according to its director, “Even with a million processors, we are a long way from achieving the scale of the complete human brain. Optimistically, we can model something like one percent of the human brain, or maybe 0.1 percent”.

Brain implants which are currently all the talk have been the subject of research for decades by units in Grenoble and Lausanne for example. Some results have been achieved but such implants require strict precautions: what about infections, durability, psychological effects of the implant?

One must also distinguish between the ability to repair and the ability to predict. We know the old dream of being able to detect lies, or to identify emotions by monitoring the brain. Current experiments, using powerful machines, to decipher the messages from the brain are conducted with the patient’s consent. It is not possible at the moment to read the brain against the will of the patient.

Furthermore, the operating mode of the brain makes considerable use of an ability to forget, not merely storage of information as in a super-computer. For a machine to approach the brain, it would require better understanding of how the brain forgets.

Finally, the question of sense remains a question intimately linked to humanity. In an all-digital society which could be coming, what would be the sense of these machines which would tell us what to do, what to choose, how to find one’s bearings, on roads and in life? Increasing the power of machines is achieved at the cost of a reduction in the human experience. The perfecting of machines must not be allowed to occult the singularity of each human being, whose brain is in fact unique.



Restez informé de nos dernières actualités

Articles récents