The Human Brain Project
The Human Brain Project has been granted funding under the EU’s multi-billion euro competition, Future and Emerging Technolgies (FET). The project will receive funds of over one billion euros for a period of ten years and endeavour to simulate the human brain using newly developed supercomputers.
On 28th January 2013, Neeli Kroes, Vice President of the European Commission and Commissioner for the Digital Agenda, announced the two winners of the EU’s Flagship initiative (FET-Flagship: “Future and Emerging Technologies Flagship”). The two projects will receive the highest ever level of EU funding for scientific research. The winning flagship projects stand to receive around one billion euros each to deliver “ten years of world-beating science at the crossroads of science and technology” and reinforce “Europe’s position as a knowledge superpower”. The winning projects competed against 19 other projects. Their selection is the result of more than three years of preparation and a rigorous evaluation by a large panel of independent scientists: “Graphene” will investigate and exploit the unique properties of a revolutionary carbon-based material described in 2004; the “Human Brain Project” will simulate the human brain using newly developed supercomputers.
The Human Brain Project (HBP), which involves 80 institutions from 19 European countries, the USA, Israel, Argentina, Japan and China, was developed over a period of three years. It was driven by neuroscientist Prof. Dr. Henry Markram from the École Polytechnique Fédérale de Lausanne (EPFL), who is coordinating the project, and Prof. Dr. Karlheinz Meier from the Kirchhoff Institute of Physics at the University of Heidelberg. Meier is the co-director of the project and in charge of the scientific and physics-oriented part of the project.
The HBP brings together 256 leading scientists and their co-workers in the following fields of research: molecular and cellular neurosciences, cognitive and theoretical neurosciences and also six large information and communication technology (ICT) platforms, which are central to the project. The HBP will also launch a major Ethics and Society programme and establish a department for managing and coordinating the project. The project will develop ICT platforms for (i) neuroinformatics, (ii) brain simulation, (iii) medical informatics, (iv) high performance computing, (v) neuromorphic computing, and (vi) neurorobotics. The project involves 55 research groups from Germany, some of whom have a pivotal role to play: the Kirchhoff Institute is coordinating the Neuromorphic Computing platform, the Jülich Research Centre is coordinating the High Performance Computing platform and Munich Technical University, the Neurorobotics platform.
The “Blue Brain Project” developed by Henry Markram at the EPFL is regarded as the precursor of the Human Brain Project. The project involves the simulation of a rat neocortical column, which is the smallest functional unit of the neocortex, and consists of only 10,000 neurons. The processes are driven by experimental data and are carried out automatically on an overarching supercomputer consisting of eight thousand IBM processors (“Big Blue”). The model already displays behaviour that corresponds to neuroscientific experiments with rat brains. These building blocks will be used in the Human Brain Project as a basis for developing large-scale simulation models and eventually reconstructing the human brain in a supercomputer.
“It is a vision, but it is a realistic one,” said Markram. The following figures show why Markram is so confident: the rat brain has around 100,000 cortical columns the size of a pinhead; this corresponds to a billion interconnected neurons. The figures reach dazzling dimensions in humans whose cortex consists of around two million cortical columns with around 100,000 neurons each. Humans therefore have as many as 200 billion neurons.
The group of researchers led by Karlheinz Meier at the Kirchhoff Institute is pursuing a different approach in order to simulate highly complex neural networks. They are constructing electronic circuits that function like neurons and their synaptic connections. Using modern chip technologies, the researchers are able to produce electronic circuits that correspond to neural networks. The first “Heidelberg Network Chip” was 5 x 5 mm in size and included 384 chips, each of which implemented around 100,000 synapses. The BrainScaleS project has revolutionised the production of chips: the BrainScaleS researchers use silicon discs (wafers) that contain an array of identical, tightly connected chips. These can be grouped together to form simulated neurons, resulting in a “neuromorphic” system with a total of around 200,000 neurons and 50 million synapses per wafer. It is already displaying a behaviour similar to the neural networks used as biological models: it has the ability to learn, tolerate mistakes and a plasticity that allows it to organise itself (at least partially). Meier’s goal is to increase the performance of the system several thousand-fold as part of the HPB’s Neuromorphic Computing platform. The eventual goal is to build the largest neuromorphic system in the world with a performance that corresponds to that of the brain of a medium-sized mammal. The system will also be made available to other partners for experimental purposes.
Developments like these require the ability to process a huge amount of data and to understand the communication of the millions of artificial neurons. The Neuromorphic Computing platform is therefore closely linked with the High Performance Computing platform, which focuses on the installation of the required supercomputers at the Jülich Research Centre.
Financing and HBP infrastructure
However, the financing (around 80 million euros) of the new EU Framework Programme (“Horizon 2020”), and hence of the two FET Flagships, from 2014 to 2020 is not yet secured. Due to the financial crisis, European heads of state have not yet agreed on the overall budget of 1.19 billion euros for a period of 10 years, of which 643 million are to be provided by Horizon 2020 and 547 million euros by the project partners. Negotiations are ongoing, with the HBP partners – scientific institutions, industrial partners and national institutions – already agreeing to provide 300 million euros. In addition to 27 small- and medium-sized companies from twelve European countries and the USA, large groups of companies in the medical, pharmaceutical and biotechnological as well as information and communication technology sectors have already signalled their willingness to support the project.
During the ramp-up phase of around two and a half years, the project will be able to use the existing structures of project partners. Meier assumes that this phase will cost around 54 million euros. However, expensive infrastructure will have to be implemented after the ramp-up phase if ambitious project goals are to be reached. The goal is to establish a new supercomputing infrastructure at the Jülich Research Centre as part of the “Partnership for Advanced Computing in Europe” (PRACE), which was established to create a pan-European, high-performance computing service for research. This will cost around 130 million euros. The Baden-Württemberg government will establish a new research building for the Neuromorphic Computing platform; the planning phase has already been concluded. The Swiss government will provide 160 million euros for the Brain Simulation platform led by Henry Markram. In addition, “Neuropolis”, a building dedicated to in silico life sciences and centred on the Human Brain Project will be erected in Lausanne. The project organisers describe the Human Project as “CERN for the Brain”, comparing the project with the Large Hadron Collider in terms of scope and ambition.
However, the initial goal will be to bring together the HBP partners and their subprojects with the aim of starting the joint work. Details of the ramp-up phase were described in the project proposal. The project will be evaluated in about 18 months’ time, when the EU will decide whether the project will be funded until 2023 with up to 100 million euros per year.
When so much money is involved, there are also always voices of dissent. Criticism not only comes from the scientists who lost out in the distribution of the funds. The larger the consortia, the more money is required for management and bureaucracy. Professor Meier will also have to spend most of his working time in meetings rather than at the Kirchhoff Institute.
Some experts doubt whether the network of millions of synthetic neurons will enable the researchers to understand how the human brain works and how thoughts are created and expressed in complex behaviour. However, nobody disputes that the HBP will provide a lot of information, including information about the functional principle of neural information processing and the development of new computer and robotics technologies. Huge scientific progress is always associated with risks and “Europe’s position as a knowledge superpower depends on thinking the unthinkable and exploiting the best ideas”, as EU Vice President Neelie Kroes put it.