A team of researchers at Imperial College London has demonstrated how it’s possible to perform artificial intelligence (AI) with tiny nanomagnets that interact like the brain’s neurons.
This new method of “nanomagnetic” computing could cut energy costs related to AI. This is crucial given how AI energy costs are doubling globally every 3.5 months.
The research was published in the journal Nanotechnology.
Achieving AI-Like Processing With Nanomagnets
In the research paper, the team demonstrated the first proof that networks of nanomagnets can achieve AI-like processing. They also showed how these nanomagnets can be used for ‘time-series prediction’ tasks, which include things like predicting insulin levels for diabetic patients.
Classic neural networks are based on the way the human brain works, with neurons communicating with each other for the processing of information. However, it has been difficult to use magnets directly in this process, with researchers not knowing how to put data in or extract information.
To simulate magnet interactions, experts usually rely on software run on traditional silicon-based computers, which helps simulate the brain. The current advancement witnessed the team using magnets themselves to process and store data, which gets rid of the need for software simulation.
Nanomagnets are not all the same. Instead, they come in various ‘states’ that depend on their direction. By applying a magnetic field to a network of nanomagnets, the state of the magnets can change based on the properties of the input field and the states of surrounding magnets.
Designing the New Technique
The team was able to take this and design a technique to count the number of magnets in each state after the field passed through.
Dr. Jack Gartside is co-first author of the study.
“We’ve been trying to crack the problem of how to input data, ask a question, and get an answer out of magnetic computing for a long time,” Dr. Gartside said. “Now we’ve proven it can be done, it paves the way for getting rid of the computer software that does the energy-intensive simulation.”
Killian Stenning is co-first author of the paper.
“How the magnets interact gives us all the information we need; the laws of physics themselves become the computer,” Stenning said.
Dr. Will Branford is team leader.
“It has been a long-term goal to realize computer hardware inspired by the software algorithms of Sherrington and Kirkpatrick,” Dr. Branford said. “It was not possible using the spins on atoms in conventional magnets, but by scaling up the spins into nanopatterned arrays we have been able to achieve the necessary control and readout.”
Reducing Energy Waste
A lot of the energy used for AI in conventional, silicon-chip computers is wasted due to inefficient transport of electrons during processing and memory storage. On the other hand, nanomagnets don’t require the physical transport of particles like electrons. They process and transfer information with a ‘magnon’ wave, with each magnet affecting the state of others around it.
This process results in less energy waste. The processing and storage of information is done together instead of separately, such as the case in traditional computers. With all of these advancements, nanomagnetic computing could be up to 100,000 times more efficient than conventional computing.
- The Black Box Problem in LLMs: Challenges and Emerging Solutions
- Alex Ratner, CEO & Co-Founder of Snorkel AI – Interview Series
- Circleboom Review: The Best AI-Powered Social Media Tool?
- Stable Video Diffusion: Latent Video Diffusion Models to Large Datasets
- Donny White, CEO & Co-Founder of Satisfi Labs – Interview Series