‘Breakthroughs are bound to come’
Prof. Dr. Stuart Parkin – director of Germany’s Max Planck Institute of Microstructure Physics, IBM Fellow, among other things – is among the world’s pioneers in neuromorphic, or brain-like, computing. In an interview at the September 2016 MESA+ Meeting in Enschede, the Netherlands, we asked him to share some of his vision.
‘The brain is still too complicated for us,’ he says, tapping the side of his head and grinning. ‘The concept of neural networks has been around for twenty years, but we really have no understanding of it. We don’t even know where our brain stores memory, or how. But we can guess.’ He shifts forward to the edge of his chair. ‘We can look for principles. We can recognize the opportunity of taking a giant step forward in computing. Nature took several billion years to produce the human brain. But we don’t have to take as long. We can’t afford to. Conventional computing is pretty good, but with computing gaining so much importance and conventional systems requiring so much energy, we have to come up with something different soon. A system similar to the human brain in terms of capacity and low power use. Some completely new kind of device. Something better than nature.’ He smiles, his humility having made way for a confidence that is contagious. ‘With so many clever people all around the world working on the problem from so many angles, breakthroughs are bound to come.’
At least one major breakthrough in the increase of computer capacity already came from Parkin in 2008, when he found a way of storing digital data in the magnetic domain walls of nanowires – a technology with the potential to enable handheld devices to hold a few thousand movies, run for weeks at a time on a single battery and be practically unbreakable. Making no secret of the hurry he is in to forge ahead, Parkin called the new system ‘Racetrack Memory’. ‘the concept is that by storing many, many bits to main walls and magnetic nanowires, we can build a new type of device that could replace virtually all conventional means of storing digital data. It could replace disk drives; it could replace flash; it could replace most solid state memories. And it would enable much simpler computers in the future’, he explained in a 2009 Youtube video published by Physics World.
Racetrack technology uses spintronics in a compact form that does not require electricity to maintain the data. It stores digital bits as tiny magnetized regions on nanowires – the racetracks. Unlike other current storage disks and chips, the Racetrack Memory is a 3D device with the nanowires constructed as vertical columns rising like a forest from a silicon chip. The 3D nanostructures provide far more storage density than anything we’ve had before.
Q. When will our smartphones be equipped with Racetrack Memory, or how near is this technology to marketability? What is the main obstacle that has to be overcome?
‘The Racetrack Memory has evolved from a concept that depended on the movement of a series of magnetic domain walls via current impulses, which I proposed in 2002, to the validation of this concept in 2008. But in the meantime, we have discovered unforeseen physics that allow us to move these domain walls at speeds ten times higher than the fastest speed we could have hoped for in 2002. Moreover, we have recently demonstrated these very high domain wall speeds in racetracks in which the magnetic domains and the domain walls themselves have no net magnetization. The domain walls, if you like, are magnetically invisible! This means we can pack the domain walls very closely together to achieve extremely high memory capacities. Thus the materials and physics have combined to be very favourable for Racetrack Memory. The major challenge now is to find ways to build the three-dimensional Racetrack and the associated reading and writing devices.’
Q. How does Racetrack Memory fit into the bigger story of the quest for a working neuromorphic system, and a new era in computing?
‘The Racetrack Memory is distinct from neuromorphic computing systems. It is a novel form of, if you like, memory storage. In time it could replace not only magnetic disk drive storage – because it enables similar or even larger capacity storage – but also some forms of conventional solid-state memories. This is because it offers high performance and is both non-volatile and cheap! Racetrack Memory could support some in-memory computation, because it is certainly possible to imagine ways of carrying out computations using magnetic domain walls. But it’s difficult to imagine how Racetrack Memory could support ultra-low energy computation, which is one of the main thrusts behind neuromorphic computing.’
Another research area you are engaged in is cognitive, or bio-inspired materials. What is currently the most exciting development in this field, and what would you say is the main challenge? What kind of results are you hoping this research will produce?
‘The main challenge in building computing materials or devices, or arrays of devices, that could enable cognitive devices, is to realise systems that will consume much less energy than any of today’s computing systems. One very interesting challenge, in my mind, is to work out how to store information not in single devices but spread over many devices, as perhaps we store information in our own brains. This might mean, for example, that one bit of information is stored in a network of many thousands of connections or nodes.’
Could you give a glimpse of what the world might look like if and when the ambitions you are pursuing become reality? For instance, how will the breakthroughs you are pushing for affect our daily lives, science, and various industries, from health and manufacturing to environmental care and, say, space research?
‘If we could build new computing systems that can operate with energy needs a million times lower than today’s computing systems, then I think that this would further spur the widespread development and use of ‘thinking machines’ throughout the very fabric of our society. Certainly the concept of the Internet of Things is of billions of devices connected to the Internet – but maybe we would rather have billions of devices talking to and communicating through each other. It has always been difficult to understand how our world would be changed by dramatic advances in computing power and in access to data. Today, it is really the latter that has so dramatically impacted our lives. Perhaps in the future it would be the innate thinking capabilities of devices and machines – and, even more so, our nearby environment - that would greatly impact our lives – I think in a very positive way!
If you had to pass on the baton today to the next generation of scientists, what would your message to them be?
‘One should always “think differently” and not follow the madding crowd.’
As the director of the Max Planck Institute, what role do you believe collaboration between different scientific institutes can and should play in the pursuit of answers to the challenges of our time?
‘Collaboration is essential to the quest for knowledge and understanding of our world. Today it seems, more and more, that one needs to combine knowledge from so many different disciplines – engineering, the natural sciences, the biological sciences, mathematics, and beyond – to solve the greatest challenges of our times. Of course, these go beyond our direct needs at this time – for example, in terms of energy, sustainability and the environment – to important philosophical questions, even about the meaning and purpose of our lives. In this regard, the Max Planck Society is very well positioned to address many of these challenges, since its institutes have leading scientists in all these domains and the Society strongly encourages collaboration and interaction among its institutes to focus on the truly most difficult and impactful challenges that affect all of our lives.’
How might the kind of scientific developments you have described help us in our pursuit of answers to these existential questions?
‘I am not sure that I can add more, but I do believe that as machines become more and more capable and as they take over much of our thinking this will make us wonder about the purpose of our lives. Of course, this touches on religion and philosophy and beyond.’
By Stephen Teeuwen, for MESA+