You are here : Home > Enabling Artificial Intelligence at the Edge

News

Enabling Artificial Intelligence at the Edge


​Until recently, development of artificial intelligence (AI) solutions has been based on software bricks installed in networks and computer servers (the Cloud). However, the future is now opening up on a local level. 

Published on 4 September 2019

​Growth in connected objects (Internet of Things - IoT) now requires new technologies to run AI on the edge. Greater reactivity of components or intelligent connected objects is now needed —e.g. to activate nearly instantaneously an autonomous vehicle's brakes— along with cybersecurity enhanced with locally processed data, which no longer need to be exported to a central network to feed an AI system. The new AI revolution therefore depends on developing dedicated electronic components such as sensors or specifically designed chips.

Technological solutions, based on very low consumption computing systems incorporating AI rapid computation modules, need to be deployed in our everyday objects. CEA-Leti and its advanced electronic component partners are now addressing this new challenge involving technological breakthroughs and market shares. The nano-electronics industry is directly impacted throughout its value chain (from components to systems) because many industries are experiencing the digital revolution: car manufacturing, aeronautics, defense and security, health and medicine, etc.

Artificial Intelligence and Trusted Intelligence

Can we trust artificial intelligence to drive an autonomous vehicle, if its computations are performed in the Cloud? One can't be sure. The slightest transmission bug and a decision will be made too late. This is why CEA-Leti is now orienting its research towards "decentralized" AI, meaning directly built into the integrated circuit and no longer remote. The institute researchers already possess multiple technological bricks: algorithms, low consumption, photonics, imagers, memories, architectures, etc. 

Top page