Project Presentation

The Josephine project aims to realize a novel class of high-temperature Josephson junctions (JJs) that behave as artificial neurons and synapses. These will enable a new neuromorphic computing paradigm: 100x faster, 10x more compact, orders-of-magnitude more energy-efficient, and with key novel capabilities -such as optical sensitivity. This will dramatically enhance the impact of neuromorphics and multiply its projected applications (from artificial intelligence to autonomous vehicles and medical devices), which constitutes the long-term vision for the science we propose. Ultimately, we will make high-T_C quantum electronics a game-changer in the new generation of computing, communication, and sensing technologies.

Jospehine is based on the idea that JJs will have a ground-breaking impact in a different computing architecture: neuromorphics. Neuromorphic computing is based on bio-inspired, non-Boolean computing schemes that mimic the brain’s architecture to solve the Von Neuman bottleneck.

Neuromorphic “software” that emulates neural networks already underlies current artificial intelligence applications (e.g., machine learning), but is slow and highly energy-intensive. This is why many efforts in fields that span from optics to solid-state physics are devoted to developing neuromorphic “hardware”. This uses analogue circuit elements that replicate the behaviour of synapses and neurons. Currently, the most advanced approaches combine CMOS technology for neurons and memristors as synapses. This non-monolithic approach suffers from the shortcomings of CMOS (energy expenditure and 2D footprint), and the mixing of digital neurons with analogue synapses, which requires analogue-to-digital and digital-to-analogue conversion at each neurone.

JOSEPHINE will unleash the intrinsic potential of Josephson physics, yielding a qualitative and quantitative leap forward with respect to the most advanced non-superconducting approaches