Achieving real-time response to complex, ambiguous, high-bandwidth data is impractical with conventional programming. Only the narrow class of compressible input-output maps can be specified with feasibly sized programs. Present computing concepts enforce formalisms that are arbitrary from the perspective of the physics underlying their implementation. Efficient physical realizations are embarrassed by the need to implement the rigidly specified instructions requisite for programmable systems. The conventional paradigm of erecting strong constraints and potential barriers that narrowly prescribe structure and precisely control system state needs to be complemented with a new approach that relinquishes detailed control and reckons with autonomous building blocks. Brittle prescriptive control will need to be replaced with resilient self-organisation to approach the robustness and efficency afforded by natural systems. Structure-function self-consistency will be key to the spontaneous generation of functional architectures that can harness novel molecular and nano materials in an effective way for increased computational power.
展开▼
