Published 2 December 2021 at 10.53
Home page. Getting computers to do calculations as energy-efficiently as our brains has long been a goal of research. Now, for the first time, a Swedish-led study has succeeded in combining a memory function with a calculation function in the same component. The discovery paves the way for more efficient technology in everything from mobile phones to self-driving cars.
Like the article p & aring; Facebook
Computers that can handle advanced cognitive tasks such as recognizing language and images – or who have superhuman chess skills – have made a big impact thanks to AI technologies in recent years.
At the same time, the human brain is still superior in terms of how efficiently and energy-efficiently the tasks can be performed.
– Finding new ways to make calculations that are similar to the brain's energy efficient processes has been a major research goal for decades. Cognitive tasks such as image and voice recognition require a lot of computer power, and more energy-efficient solutions are especially important in mobile applications such as mobile phones, drones or satellites, says Johan Åkerman, professor of applied spin electronics at the University of Gothenburg, in a mailing.
< p>In collaboration with a research group at Tohoku University in Japan, he has led a study that has now taken an important step along the way. In the study, which was published in the journal Nature Materials, the researchers have for the first time managed to connect the two main tools available for the advanced calculations: oscillator networks and memristers.
Oscillators can be described as oscillating circuits that can perform calculations, and that can be likened to human nerve cells. Memristors are programmable resistors that can also perform calculations and also have built-in memory, and which can thus be compared to memory cells. Now that researchers have been able to integrate the two, much has been gained.
– This is an important breakthrough because we show that it is possible to combine a memory function with a calculation function in the same component. These components function more like the energy-smart neural networks in the brain, and can therefore become important building blocks in future, more brain-like computers, says Johan Åkerman.
According to Johan Åkerman, the discovery opens up for faster, smoother and less energy-intensive technology in a number of areas. He emphasizes that it is a great advantage that the research group has succeeded in producing the components in extremely small formats: hundreds of components can be placed on a surface that corresponds to the size of a bacterium. This can be especially important for smaller applications such as mobile phones.
– Many functions in mobile phones would be facilitated with more energy-efficient computing capabilities. An example is digital assistants like Siri or Google. Today, all handling takes place via servers because the calculations require too much energy for the mobiles' small formats. If the calculations can instead be done locally, directly in the mobile, they can be done faster and easier when no connection to servers is needed, he says.
He mentions self-driving cars and drones as other examples where a more energy-efficient calculation capacity would lead development forward.
– The more energy efficient it is possible to make cognitive calculations, the more applications open up. Therefore, our study really has the potential to move the area forward, he says.