Neurobiology and Artificial Intelligence: Part One and a Half - News from the Blue Brain Project

    While writing the second part, I came across the news and decided that it would be interesting to many in this hub.

    So there is such a great project sponsored by IBM called the Blue Brain Project . The main idea of ​​this enterprise is to take and digitize the brain. Up to the molecular level, in order to study all the processes occurring in it in silico. Well, there is a corresponding computer - Blue Gene Supercomputer.

    As of August of this year, they can simulate about 100 cortical columns consisting of 1 million neurons and a billion synapses.

    So the news is that they threw about 10,000 neurons into the simulation in random order in 3D space (taking into account the total density and local morphology of neurons), and tracked how they will build connections between themselves, and then compared with data on real the brain of a mammal. As a result, they got a 75-95% similarity in positioning synapses.

    So what does this mean for us?

    Here, in the previous topic, comrade eresik asked by what principle neurons establish connections, and so, VVR give us the answer: neurons grow as independent of each other as it is physically possible and establish synapses in places where they just collided.
    But, as always, there are exceptions that demonstrate the possibility of chemical signals that change the statistics of compounds. Given these exceptions, it is possible to predict the position of all synapses of the neurochain.

    Generally speaking, this makes sense. If cognitive functions depended on a detailed pattern of interneuronal connections, any trauma would be a catastrophe, and neuroplasticity would be impossible (taking functions of a damaged part of the brain by neighboring regions). It is precisely such possibilities that give reason to think that in order to simulate consciousness one does not need to go deeper to the level of molecules, and even more so, to the quantum level.

    They also note that changes in density, positions, and orientations did not affect the distribution of synapse formation positions. But the change in morphology turned out to be significant for reproducibility of the results.

    Well, the accompanying video:


    Also popular now: