#AI should not be sentient for the purpose of being able to control - LinkedIn

Chief AI Scientist at Meta, who has published a sort of blueprint for creating A Path Towards "autonomous machine intelligence'. Why I think they are important gaps in an otherwise good vision. In the meta-level architecture, components like "world-model", Does the "Critic" and "Actor" provide a "inner dialog", but would like to see more system 2 consciousness dianoetic modelling / workspace. If this is to be modeled on #DeepLearning NN models only, I think we need "new thinking" and innovation to go beyond that to enable closer neuroscience biologically plausible modelling (using the brain as a worked example). An integrated, DL-based, modular, cognitive architecture. Joint-Embedding Predictive Architecture (JEPA) as an architecture for world models that can handle uncertainty.