Artificial Intelligence (AI) may be an abstract concept, but, in reality, AI is specific applications that are executed on physical computers and consume physical resources. AI applications are compute-intensive, and for this reason, they are commonly executed in datacenters. This approach is called Cloud AI and has led to a massive demand for datacenters, where the most powerful ones can consume as much energy as the city of Oslo. A better alternative is to perform AI tasks on the lower-energy devices that AI systems use to collect input data, so-called Edge AI, but this is technologically challenging. Norway is however in an ideal position to address these challenges through its strong and internationally oriented microelectronics industry and leading academic research groups – and this is exactly what SFI NICE aims to do.

The key challenge when moving AI applications from the Cloud to the Edge is to deliver sufficient computing performance while respecting stringent energy and power consumption requirements and minimizing system cost. The Research Pillars (RPs) of NICE focus on the fundamental subsystems of Edge AI platforms, i.e., sensing, processing, communication, and system software. NICE further includes selected Applied Partner Themes (APTs) in which the technological advances of the RPs are used to address real-world industrial and societal challenges. We foresee that our work in the APTs will expose new challenges that we can feed back into the RPs, thereby creating a self-reinforcing cycle of research and innovation.