PHOTONIC ENABLED PETASCALE IN-MEMORY COMPUTING WITH FEMTOJOULE ENERGY CONSUMPTION
Applications employing artificial intelligence (AI) are increasingly becoming part of our everyday life. AI already influences how we communicate, work, manufacture, learn, research or even consume online services. However, many AI applications rely on substantial computational resources and require demanding storage capacity to process large data volumes. Prime examples are internet application such as the Internet of Things and broadband services such as HD video on demand or social media.
Modern digital electronic technologies are struggling to satisfy this computational challenge and rapidly approach their physical limits.
By moving away from electronic towards photonic approaches, the PHOENICS project aims to develop disruptive methods for implementing next-generation compute platforms.
Concept of PHOENICS architecture
PHOENICS aims to
- demonstrate a neuromorphic compute platform with Petascale processing performance
- realize photonic processors with unprecedented energy efficiency
- explore novel types of materials for brain-inspired computing
- remove latency barriers of electronic hardware
- establish new paradigms for AI computing