I have been around enough to remember a similar A.I. hype in the 1980s involving Japan, expert systems, and logic (Prolog) supercomputers. This time it is China

I have been around enough to remember a similar A.I. hype in the 1980s involving Japan, expert systems, and logic (Prolog) supercomputers. This time it is China, deep learning, and deep learning accelerator hardware. The 1980s was a commercial failure, but expert systems never really went away. One interesting project was Cyc a massive comon sense rule system, sort of like a thesarus; now into its 4th decade. Another is IBM' Watson which is a hybrid of rules and statistical large data. The difference this time is that deep learning appears to work in some cases and getting fast enough to operate in real time. One application is environmental object classification for self driving vehicles. The deep learning systems do as well as explicit algoritms and are easier to program. Secondly, I saw some papers at the recent SIGGRAPH where they replace procedural portions of graphics flows with deep learning and obtain good results. Whether deep learning can solve every problem under the Sun as some hypers claim is undetermined.