This article was published in our newsletter. Sign up here.
The computer wins 4:1. That was the outcome of a dual which saw one of the world’s best Go players competing in March 2016 against software by the name of AlphaGo. Previously, the board game Go was considered to be too complex for a computer on account of the huge variety of possible moves. The fact that the software beat Lee Sedol, a professional Go player, in five matches was therefore celebrated as a milestone in artificial intelligence. Lee Sedol came out on top in only one of the five matches.
Computer versus brain
The software triumphed over the top player because it has access to a database of millions of moves from the world’s best players – and because it can learn autonomously. This latter ability is something which the human brain, with its billions of nerve cells and interconnections, does every second of every day for an entire lifetime – and even achieves with a comparatively low consumption of energy.
Brain researchers and computer experts learn from one another
Because computer experts and brain researchers can learn a lot from one another, they have been working together in the international Human Brain Project since 2013. Their goal is to construct a virtual brain capable of simulating the complex interactions between cell networks on a computer. Among the participating scientists from 23 countries is Professor Markus Diesmann, an expert in neurophysics, and his group at Forschungszentrum Jülich.
The researchers are exploring the uniquely dynamic way in which information is processed in the brain cells. To this end, they analyse the structure of wafer-thin sections of brain tissue and then reproduce them in mathematical models. The idea is to achieve an exemplified and simplified depiction of individual nerve cells and their interconnections, while making the model as realistic as possible. For the brain’s activity – right down to the responses of individual molecules – then to be simulated, tailor-made software is required. This is being developed by the Jülich brain researchers in collaboration with computer experts in the global NEST network.
Supercomputers and super-brains
The volumes of data generated by the brain researchers are huge. They need extremely powerful computers to analyse and visualize them, and to simulate the brain’s activity. One such supercomputer is to be found at the Jülich Supercomputing Center. Boasting more than 130,000 processors, JUQEEN, as the machine is called, can perform over 1,000 trillion computer operations per second, making it one of the ten fastest computers in the world.
The Jülich computer experts are already working on a new generation for the Human Brain Project. The new supercomputer will be even quicker but will probably need the energy output of a small power station to do so. The efficiency of the human super-brain remains unrivalled: it only requires the energy of a 60 watt lightbulb to process highly complex information.