Welcome to the W3C Cognitive AI Community Group

Hello Onn Kei,

Please introduce yourself and your interests in cognitive AI to others on this list.

The CogAI CG has previously worked on Chunks & Rules inspired by John Anderson’s ACT-R.  There is a specification, open source implementation and a suite of demos. Chunks & Rules can be applied to low-code descriptions of real-time behaviour for digital twins, see the demo for a bottling plant.

We’ve also worked on plausible reasoning with imperfect knowledge, i.e. everyday knowledge that is uncertain, imprecise, context sensitive, incomplete, inconsistent and subject to change. This was inspired by the work by Alan Collins and his colleagues in the 1980s.  There is a specification and a proof of concept demonstrator.

More recently, the focus has shifted to artificial neural networks and how to make the quantum leap from Generative AI to Sentient AI as the basis for agents with continual learning through continual prediction, continual reasoning, awareness of their environment, goals and performance, along with the means to remember and reflect on their past experiences.

Sentient AI is a more concrete objective than AGI, and sidesteps the question of what it means to be conscious. Moreover, Sentient AI systems only need to be as smart as their application requires, for example, a factory robot clearly doesn’t need to know the complete works of William Shakespeare!

The immediate challenge is to understand how to design neural networks that can mimic the characteristics of human memory, e.g. for working, short and long term memory, as well as for factual, episodic and procedural memory.  We also need to consider how to separate private memories from those that can be safely shared.

Best regards,

Dave Raggett <dsr@w3.org>