Researchers have shown that it is possible to create artificial intelligence using small nanomagnets that interact like brain neurons (Calculation of artificial intelligence with small nanomagnet networks).
Developed by a team of researchers at Imperial College London, the new technology (Calculation of artificial intelligence with small nanomagnet networks) has been able to reduce the energy costs of artificial intelligence (AI), which are currently doubling worldwide every 3.5 months.
In an article published today (May 5, 2022) in the journal Nature Nanotechnology, the international team provided the first evidence that nanomagnetic networks can be used to perform AI-like processing. Researchers have shown that nanomagnets can be used for “time series prediction” tasks, such as predicting and regulating insulin levels in diabetic patients. Artificial intelligence using “neural networks” aims to replicate the way parts of the brain work where neurons communicate with each other to process and store information. Much of the mathematics used to operate neural networks was originally invented by physicists to describe how magnets interact, but at the time it was very difficult to use magnets directly because researchers did not know how to store data and retrieve information.
However, software powered by traditional silicon-based computers is used to mimic magnetic interactions, to mimic the brain. The team is now able to independently use magnets to process and store data – reducing software intermediation simulation and offering potentially enormous energy savings.
Nanomagnetic state
Nanomagnets can be in different “states” depending on their direction. The application of a magnetic field to a network of nanomagnets changes the state of the magnets based on the properties of the input field, but also the state of the surrounding magnets.
The team, led by researchers from the Imperial Department of Physics, was able to design a technique for counting the number of magnets in each state as they pass through the field, giving an “answer.”
Co-author of the study Dr. Jack Gartside said: “We have been trying for a long time to solve the problem of entering data, querying and getting an answer from a magnetic computer. Now that we have proved that this can be done, it provides a way to remove computer software that performs high-speed simulations.
Co-author Kilian Stenning added: “As magnets interact, it gives us all the information we need; the laws of physics themselves have become computers.”
Group leader Dr. Will Branford said: “It is a long-term goal to achieve computer hardware inspired by the software algorithms of Sherrington and Kirkpatrick. It is not possible to use atomic rotations on conventional magnets, but by scaling rotations on nanopatter fields, they have achieved the required control and readability.
Reduce energy costs
AI is now used in a variety of contexts, from voice recognition to self-driving cars. But training artificial intelligence to do relatively simple tasks can take a lot of energy. For example, training artificial intelligence to solve the Rubik’s Cube required energy equivalent to two nuclear power plants in operation per hour.
Much of the energy used to achieve this goal in conventional silicon chip computers is wasted on bad bearing electrons during processing and storage. However, nanomagnets do not rely on the physical transport of particles, such as electrons, but process and process information in the form of a “magnon” wave, with each magnet affecting the state of neighboring magnets.
This means that less energy is lost and that information processing and storage can be combined instead of separate processes, such as conventional computers. This innovation will make a nanomagnetic computer up to 100,000 times more efficient than a conventional computer.
