CEA-Leta finds inspiration for edge AI in insect’s nervous systems

CEA-Leti, a France research institute, has received a $3 million grant from the European Research Council (ERC) to build a first-of-a-kind intelligent edge AI system inspired by the insect’s nervous system. The nanoscale system uses novel technologies to boost the performance and energy-efficiency of edge AI. Targeted applications include robotics, such as fruit-picking and for rescuing, medical implants and wearable electronics.

Although classical computing architectures as used in electronics such as the PC, the data center and mobile devices today are ubiquitous, a problem is that most of the energy is actually used for moving data rather than processing data. That’s why researchers over time have tried to find more optimal approaches, especially given the rise of AI over the last decade. One such alternative is called in-memory computing, where memory is used both for storage and processing. 

Buggy inspiration for edge AI

However, the issue with in-memory computing is that it requires a fast, non-volatile memory with high endurance. Such a memory currently doesn’t exist since DRAM is volatile. To circumvent this problem and reduce the memory requirements, CEA-Leti scientists have found inspirations in the insect nervous system. 

Senior scientist Elisa Vianello has received a $3 million grant from the ERC to use novel nanoscale memory technologies that mimic the biological mechanisms from insects to create silicon-based, energy-efficient nanoscale systems for edge AI.

“My project is to take inspiration from insects’ nervous systems to relax hardware requirements in terms of memory density and reliability, and to build the new nanosystems we need to enable learning from a very limited volume of noisy data,” said Vianello.

“Crickets make accurate decisions based on sluggish, imprecise, and unreliable neurons and synapses in order to escape their predators. Looking closely at their biology, we identified a diversity of memory-like functions at play in their sensory and nervous systems. By combining these different functions, the cricket’s internal computing system achieves amazing performance and energy efficiency.”

The primary goal is to create devices that enable learning from a limited amount of noisy data. Such data could come from sensors such as video cameras, radar, ECG, EMG, bio-impedance streams and brain signals. To that end, Vianello has discovered that various functions of the insect’s nervous system closely resemble those of different memory technologies that are being worked on by CEA-Leti. Hence, the scientists aim to create a “hybrid synapse” that co-integrates these different memory technologies.

The concept of a heterogeneous distributed computing system shows various processing units and sensors that are connected through both spiking and non-spiking links.