
Prof Damien Thompson. Image: University of Limerick
University of Limerick’s Damien Thompson explains the research behind a brain-inspired analogue computing platform that is made of completely different material to the silicon used in traditional computers.
In September of last year, a research team at University of Limerick (UL) made a what it described as a “major discovery” by designing molecules that could revolutionise computing.
The team at the university’s Bernal Institute was led by Damien Thompson, professor of molecular modelling at UL and director of SSPC, the Research Ireland centre for pharmaceuticals.
Speaking to SiliconRepublic.com, Thompson said the team essentially made a new kind of material for computing. “We believe that this new discovery will lead to innovative solutions to societal grand challenges in health, energy and the environment, because of its potential to reimagine computing.”
Traditional digital computers – which includes just about every computer in the world from desktops to smartphones – are made from silicon.
This silicon can store and process data, but this is limited to just two states, on and off. “No grey areas,” according to Thompson.
“That’s clean, it has worked very well as we know, it’s got us to where we are now in our digital age, but it is a fundamental limitation because one chip does one job at a time,” he said.
“Here, we’ve developed a brain-inspired analogue computing platform that is a completely different material than silicon. It does the same job of storing and moving numbers around, but it is a molecular film that is capable of storing and processing data in thousands of conductance states.”
Wiggling and jiggling
Thompson said the breakthrough was a team effort achieved through an international collaboration with scientists at Indian Institute of Science and Texas A&M University. Through this collaboration, the team took inspiration from the human brain, using the natural ‘wiggling and jiggling’ of atoms to process and store information.
“We’re finally at a stage where we can design and build using the same material that makes life possible, makes living creatures like us possible, which is molecules. They are the smallest building block, a group of atoms bonded together,” he said.
“I recently traumatised my young children by describing molecules as ‘the tiny little worms that we’ve made of’, and we’ve decided we don’t like that and we’re going to call them instead ‘little squiggly squashy things’. Here in our work, we found a way to actually track all that wiggling and jiggling, all those minute molecular movements.”
The researchers tracked these tiny, molecular movements using precisely timed voltage pulses that capture each pose. They then mapped each pose or state to a distinct electrical signal, which created an extensive ‘molecular tour diary’ of different distinct memory state. Additionally, Thompson said each state stores and processes data within the same location, similar to the human brain, avoiding the latency or wait time that comes when information is moved across the network.
“Here, it’s a one-stop shop, all done in one tiny chip, that’s what gives you huge savings in space and energy. If you want massively parallel high-density computing in a data centre up in Meath, or you want a real-time interactive digital map that doesn’t kill your phone’s battery as you try desperately to find your way to Croke Park, or if you want really immersive online gaming, then these smaller, faster, cheaper processors are what you need.”
Real-world applications
Thompson said the new material extends neuromorphic computing beyond niche applications and could potentially “unleash the long-heralded transformative benefits of artificial intelligence”.
“By precisely controlling the vast array of available molecular kinetic states, we created the most accurate, 14-bit, fully functional neuromorphic accelerator integrated into a circuit board that can handle signal processing, AI and machine learning workloads such as artificial neural networks, auto-encoders, and generative adversarial networks. Most significantly, leveraging the high precision of the accelerators, we can train neural networks on the edge, addressing one of the most pressing challenges in AI hardware.”
Beyond data and image processing, Thompson said these types of neuromorphic chips could be used to design artificial neural systems, such as vision systems, auditory processors and autonomous robots. These robots’ physical architecture and design principles would be based on humans’ own evolved biological nervous systems.
“I see great potential for a convergence, a confluence, in quantum computing and these types of biomimetic molecular electronics materials. It makes perfect sense to me to exploit chemistry and molecular systems for quantum computing.”
In the near-term future, Thompson sees these highly efficient chips creating “big societal benefits”, including sensors for leaks in pipes and voice-controlled self-driving cars which have the potential to make life a bit easier and safer.
However, given the link to the human brain, he also spoke about it taking us one step closer to brain-computer interfaces, which involves using an implant that can pick up signals from the brain for a variety of purposes.
“My personal opinion is that, given the immaturity of the technology, it is clear that brain chip implants in humans must be limited to those for whom there are no other options to treat severe, debilitating conditions,” he said.
“Until we have a full fundamental understanding of the underlying science and large datasets on its effects, then the risks outweigh the benefits for all but the most severe cases. There is also a moral and legal obligation to obtain consent. We need clear regulation, akin to that developed for nuclear and biological weapons and, most recently, AI, to protect against misuse.”
From an educational standpoint, Thompson also pondered the possibility of students having a ‘smart patch’ implanted to get more points in an exam. “We have a hard enough time watching out for answers written on pencil cases, we really don’t want to be filtering out augmented brainwaves inside the exam hall as well. That’s still very much science fiction, thankfully.”
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.