The brain is one of the best, most energy efficient computers in existence, but it's also one of the least understood.
It's a system able to handle complex data from five senses, understand the world in 3D, and reflect on its own experiences -- and in order to understand just how it works, one ambitious project is aiming to build its own version of the brain, in computer form.
The Human Brain Project was set up in 2013 with a budget north of €1bn and a ten-year remit to help further humans' understanding of brain biology through technology.
"Understanding the human brain is one of the greatest challenges facing 21st century science. If we can rise to the challenge, we can gain fundamental insights into what it means to be human, develop new treatments for brain diseases, and build revolutionary new information and communications technologies," said a report prepared for the European Commission ahead of the project's launch.
The Human Brain Project is intended to boost Europe's potential for innovation, growth, and jobs, as well as to address some of the major societal challenges facing the continent.
"The idea was that very large projects, which are very innovative and also risky, have powerful support and a very long funding period of 10 years. This was something new in Europe -- usually European grants last for three to five years, and maybe go into a second round. It was quite new to have a project with a horizon of 10 years. It was exciting: you can approach completely different research questions when you're aware it is such a long project. It broadened the horizons in terms of what can be approached, and how challenging we can formulate our aims," Katrin Amunts, a neuroscientist and scientific research director at the Human Brain Project, told ZDNet.
That means the Human Brain Project can afford to be fearsomely ambitious, seeking to create a working simulation of the brain before its 10-year timescale is out. It's hoped by creating this, researchers will be able to understand some of the processes at work in our grey matter that have so far remained opaque to science.
"Crucial and central to our understanding of the brain is that it is organised on different spatial scales, from molecules to cells to circuits to networks to the whole organ, and at each level we try to create a productive loop between experiments, theory, and modelling simulation. Then we try to bridge the gaps and to verify results of simulation against empirical experiments. Simulation is a crucial tool, but our aim is not to simulate the brain, it's to decode it -- to understand it by decoding," Amunts added.
The Human Brain Project is divided into 12 separate sub-projects, almost evenly split between those trying to progress our understanding of neuroscience through scientific research, and those trying to do the same through technology. The HBP's technology sub-projects include work on neuroinformatics, high-performance computing, neuromorphic computing, and neurorobotics.
While the Human Brain Project's 10-year timeframe is an age in technology years -- 10 years ago the iPhone hadn't even been released and the bestselling phone was a Nokia candybar -- the project is working with tech companies on their upcoming product lines to make sure its future demands will be met. "For supercomputing, these people are used to thinking in such large periods. They have a good idea how fast the processors will be, how many petaflops can be achieved, and from here they have a very clear schedule where they want to go in five years, or seven years. We can just estimate how much space or memory we need in order to represent a full human brain at cellular resolution," Amunts said.
The project's research is underpinned by a high-performance computing infrastructure based on two prototype supercomputing systems, one from Cray and one from Intel. "These are new systems which are not sold yet; there are new technologies inside them. We wanted to collaborate with the companies to develop a future supercomputer which can be used to address future brain research, simulation, data analytics, and so on," she added.
The mix of tools, including hardware, software, datasets and programming interfaces, are at different stages of maturity -- some fully functional and some still in need of some development work.
"Neuroscientists like me, who have been developing methods for a long time, we haven't developed methods to such a level of maturity that we can easily work with them. Also, we don't have the resources to do it. We are neuroscientists and it's expected that we publish nice papers, but we cannot think so much about making a website user-friendly and easy to handle -- it's a technological question and universities usually can't do both. Here we have a unique situation that we can care about it, we can take care of the service aspects, there are hotlines [on which] people help you to use these tools that are developed."
The hope is that using technology to recreate elements of the brain will help both fields make their own breakthroughs. Take the field of neuromorphic computing, for example: by drawing inspiration from the low-energy, high-power computing infrastructure of the brain, it's hoped a less-energy intensive successor to today's von Neumann architecture can be devised. Likewise, by simulating areas of the brain on existing neuromorphic platforms (such as SpiNNaker), more will be understood about its workings. Neuromorphic computing could, for example, ultimately learn about how the brain is able to adapt to any number of different situations.
"We think that modular computing is very important development which is relevant for us -- we see already now that some of our jobs run very well on CPUs, while others are much better running on GPUs. If you think about neuromorphic computing... it is specialised for certain types of applications. You could think about transferring very time-sensitive applications like learning and plasticity -- learning over long time ranges - to run very nicely on neuromorphic chips," Amunts said.
Perhaps the Human Brain Project's combination of both neuroscience and technology research could ultimately shed light on one of the biggest mysteries of the brain: consciousness. While there are various theories on how consciousness works, very little about its mechanics is truly known.
It's an area of neuroscience that will become particularly pertinent to technology in future: as artificial intelligence continues to develop, and perhaps a true general intelligence emerges, there will be questions around whether the machine is conscious, and what it means if it is -- particularly if those artificial intelligences are embedded in any kind of robotic body.
"Our robotics friends are convinced that when we have a very good simulation of certain cognitive processes and we can translate that to a robot, then we have the possibility to test these robots to behave in the certain way we expect from experiments," she said.
"We try to understand the organisational principles and many of us are convinced if we better understand how the brain is organised and what are the rules that allow us to speak, understand, or calculate, or how information is organised, then we have a better concept of what does consciousness mean," Amunts added.