Artificial thinking, fast and slow
ML luminary's AI roadmap promotes popular dichotomy from psychology
At his NeurIPS keynote, Yoshua Bengio (one of the Turing Award winners for research in deep learning) gave a keynote to a massive crowd. I’m in there somewhere.
In it, he lays down a roadmap for tackling consciousness. Bengio cited the “System 1 and System 2” dichotomy introduced by Daniel Kahneman in his book Thinking, Fast and Slow. System 1 refers to what current deep learning is very good at — intuitive, fast, automatic, anchored in sensory perception. System 2 meanwhile represents rational, sequential, slow, logical, conscious, and expressible with language.
Important milestones on this roadmap include more forms of compositionality and systematic generalization. These terms generally refer to the ability to dynamically combine modeling artifacts into new artifacts in a way that provides exponential performance gains. He points to human language as an example.
Bengio believes that deep learning will be the primary driver of this transition from system 1 to system 2, mainly using attention mechanisms.