Interview: Laurent Fesquet, research scientist, TIMA

“Rebooting Computing challenges the very foundations of today’s computers”

The ICRC conference, to be held in Grenoble in April, will provide insights into the Rebooting Computing initiative that began in the United States in 2016. What is it all about?

The idea is to rethink how computers are designed. Today, most machines are built on the von Neumann architecture. This architecture was invented in 1945 to perform exact computations, something that uses a lot of energy. We are now seeing other, more economical approaches like neural networks, approximate computing, and probabilistic and quantum architectures.

And technology is creating even more new alternatives…

Precisely. In 2018, Intel presented its Loihi neuromorphic chip, which contains a record 130,000 neurons. And advances in microelectronics are enabling very dense neural networks. Not to mention the first quantum computers and the advent of nanocomponents, which will bring computation closer to memory. And there is also the use of statistical distributions for probabilistic computing…

What are you working on at TIMA?

We are developing clockless, or asynchronous, circuits. An asynchronous neural network uses less energy than an equivalent synchronous network. This is because only those neurons used for a computation are activated. This is how Intel’s chip works. At ICRC I will present research that combines asynchronous technology and probabilistic computing with a view to designing a probabilistic computer. This is work we are doing in partnership with our colleagues at LIG.

*International Conference on Rebooting Computing

Contact: laurent.fesquet@univ-grenoble-alpes.fr

Facebook

Twitter