Thanks for sharing this kind of talks, it is very interesting. So we are entering into a new cognitive era, building our most potential tool - the mind.
Climate change, weather patterns... I'd think this is something perfect for this type of system to chew on. Anomaly detection. We have a worldwide network of weather stations, constantly streaming data. Plug that stream into an encoder and let it rip.
Great talk. However I found it awkward to say “mathematical vs memory-based” approach. This is saying the “mathematical” approach (which I guess covers most of the "main-stream" AI) does not give rise to a memory. This is simply not true. And this is also admitting that his memory-based approach has no (or much less) math, which is hardly a merit.
Huayan Wang They aren't saying that the system cannot be described mathematically - they are saying that there aren't mathematical formulas in the brain which constitute the actual physical functioning. Currently they are working on the math which describes the system.
+Huayan Wang It's a merit because when you are trying to catch a ball the brain isn't doing calculus. The brain's relying on emperical data and pervious experiences exclusively