News

Future of computing: Intel studies neurochips that promise energy efficiency

After releasing a study on how bee honey can be a critical element for supercomputers and neuromorphic chips, now the Intel points out how neurochips could bring energy efficiency considerable contribution to computing and, more than that, they can bring improvements to tasks done with artificial intelligence (AI) – such as, for example, speech or gesture recognition.

The manufacturer has been working on the Loihi — neuromorphic chips, that is, they mimic the functioning of a human brain in terms of speed and efficiency.

Despite not yet having specific products being developed with the technology, the company believes that the future will be made of chips of the type that will open doors to perform AI tasks, bringing energy efficiency, as well as improving the functioning of the cloud.

“Right now, with Loihi, we’re at that point where we think we’re on the right track, but we don’t actually have any product plans yet. We’re a little further along this workflow,” said Intel Labs leader Rich Uhlig at a roundtable last month with journalists.

The manufacturer’s laboratory works with research and development precisely to find new ways to improve the technology without necessarily needing a commercially viable invention in a short time. The Loihi are within this panorama, as well as other innovations with a focus on microarchitecture, system architecture and even software and compilers.

Loihi chips, in their second generation, are smaller than the size of a finger and giant in efficiency – Image: publicity/Intel

Inside Intel’s Neurochips and Their Efficiency

Neuromorphic chips have been one of the relatively recent technologies that have attracted attention and investment from IT giants like Intel — but not only. Samsung, for example, is another big name behind the studies on neuromorphic chips in the market, with application in AI.

In the case of the Loihi, presented for the first time in 2017, the dynamics of functioning brings together pin-like structures that aim to mimic the functioning of a human brain, consuming electricity only when distributing data. Traditional processors, on the other hand, keep working all the time, regardless of whether they are processing information – so the energy efficiency of these new “brain chips” draws attention.

In practical terms, Loihi chips can perform an image and likeness search 24 times faster than a traditional search, and using 30 times less energy, to name just one example.

Researchers at Sandia National Laboratories, in the United States, also point to the use of Loihi chips in scenarios that require high-performance computing, such as tracking financial market movements, the spread of diseases in a population and data flows on social networks.

The possibilities are still being studied, but they could bring a huge advance to the future of computing.

Via: The Register

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *