MSGID: 1:317/3 6274a49a   
   PID: hpt/lnx 1.9.0-cur 2019-01-08   
   TID: hpt/lnx 1.9.0-cur 2019-01-08   
    'Nanomagnetic' computing can provide low-energy AI    
      
     Date:   
         May 5, 2022   
     Source:   
         Imperial College London   
     Summary:   
         Researchers have shown it is possible to perform artificial   
         intelligence using tiny nanomagnets that interact like neurons in   
         the brain.   
      
      
      
   FULL STORY   
   ==========================================================================   
   Researchers have shown it is possible to perform artificial intelligence   
   using tiny nanomagnets that interact like neurons in the brain.   
      
      
   ==========================================================================   
   The new method, developed by a team led by Imperial College London   
   researchers, could slash the energy cost of artificial intelligence   
   (AI), which is currently doubling globally every 3.5 months.   
      
   In a paper published today in Nature Nanotechnology, the international   
   team have produced the first proof that networks of nanomagnets can be   
   used to perform AI-like processing. The researchers showed nanomagnets   
   can be used for 'time-series prediction' tasks, such as predicting and   
   regulating insulin levels in diabetic patients.   
      
   Artificial intelligence that uses 'neural networks' aims to replicate   
   the way parts of the brain work, where neurons talk to each other to   
   process and retain information. A lot of the maths used to power neural   
   networks was originally invented by physicists to describe the way magnets   
   interact, but at the time it was too difficult to use magnets directly   
   as researchers didn't know how to put data in and get information out.   
      
   Instead, software run on traditional silicon-based computers was used   
   to simulate the magnet interactions, in turn simulating the brain. Now,   
   the team have been able to use the magnets themselves to process and   
   store data - - cutting out the middleman of the software simulation and   
   potentially offering enormous energy savings.   
      
   Nanomagnetic states Nanomagnets can come in various 'states', depending   
   on their direction.   
      
   Applying a magnetic field to a network of nanomagnets changes the state   
   of the magnets based on the properties of the input field, but also on   
   the states of surrounding magnets.   
      
      
      
   ==========================================================================   
   The team, led by Imperial Department of Physics researchers, were then   
   able to design a technique to count the number of magnets in each state   
   once the field has passed through, giving the 'answer'.   
      
   Co-first author of the study Dr Jack Gartside said: "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. 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."  Co-first author Kilian Stenning   
   added: "How the magnets interact gives us all the information we need;   
   the laws of physics themselves become the computer."  Team leader Dr Will   
   Branford said: "It has been a long-term goal to realise computer hardware   
   inspired by the software algorithms of Sherrington and Kirkpatrick. 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."  Slashing energy cost AI is   
   now used in a range of contexts, from voice recognition to self-driving   
   cars. But training AI to do even relatively simple tasks can take huge   
   amounts of energy. For example, training AI to solve a Rubik's cube took   
   the energy equivalent of two nuclear power stations running for an hour.   
      
      
      
   ==========================================================================   
   Much of the energy used to achieve this in conventional, silicon-chip   
   computers is wasted in inefficient transport of electrons during   
   processing and memory storage. Nanomagnets however don't rely on the   
   physical transport of particles like electrons, but instead process and   
   transfer information in the form of a 'magnon' wave, where each magnet   
   affects the state of neighbouring magnets.   
      
   This means much less energy is lost, and that the processing and storage   
   of information can be done together, rather than being separate processes   
   as in conventional computers. This innovation could make nanomagnetic   
   computing up to 100,000 times more efficient than conventional computing.   
      
   AI at the edge The team will next teach the system using real-world   
   data, such as ECG signals, and hope to make it into a real computing   
   device. Eventually, magnetic systems could be integrated into conventional   
   computers to improve energy efficiency for intense processing tasks.   
      
   Their energy efficiency also means they could feasibly be powered by   
   renewable energy, and used to do 'AI at the edge' -- processing the data   
   where it is being collected, such as weather stations in Antarctica,   
   rather than sending it back to large data centres.   
      
   It also means they could be used on wearable devices to process biometric   
   data on the body, such as predicting and regulating insulin levels for   
   diabetic people or detecting abnormal heartbeats.   
      
      
   ==========================================================================   
   Story Source: Materials provided by Imperial_College_London. Original   
   written by Hayley Dunning. Note: Content may be edited for style and   
   length.   
      
      
   ==========================================================================   
   Journal Reference:   
      1. Jack C. Gartside, Kilian D. Stenning, Alex Vanstone, Holly   
      H. Holder,   
         Daan M. Arroo, Troy Dion, Francesco Caravelli, Hidekazu Kurebayashi,   
         Will R. Branford. Reconfigurable training and reservoir computing in   
         an artificial spin-vortex ice via spin-wave fingerprinting. Nature   
         Nanotechnology, 2022; DOI: 10.1038/s41565-022-01091-7   
   ==========================================================================   
      
   Link to news story:   
   https://www.sciencedaily.com/releases/2022/05/220505114646.htm   
      
   --- up 9 weeks, 3 days, 10 hours, 50 minutes   
    * Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)   
   SEEN-BY: 15/0 106/201 114/705 123/120 129/330 331 153/7715 218/700   
   SEEN-BY: 229/110 111 317 400 426 428 470 664 700 292/854 298/25 305/3   
   SEEN-BY: 317/3 320/219 396/45   
   PATH: 317/3 229/426