I submit that using neural techniques to solve a problem is not AI. Neural programming is a different solution type - more like comparing writing a program in a sequential language like C or Fortran compared to a data driven language like Labview that is fundamentally multi-threaded. If it were AI, the machine would be able to understand what had been said, ascribe context to it, and be able to integrate it into its database for application in a completely different domain. Not only would it have translated what it read, but it would have learned something about the psychology of what had been said - not just how to better convey it in a different language.
We are really far away from this type of AI concept learning, concept incorporation into it own intelligence, and application of the concepts to problem solving in a completely different domain. This still requires invention. Comparing machine to human intelligence for Futurist predictions not only must presume indefinite timing of inventions to move it forward, but must also account for a sliding scale in our measure of human performance. There are recent reports that the speed of cognitive processing in the brain is much faster in (if I can coin a term) "cog-nits/second" than they had previously estimated. On Thu, Mar 16, 2017 at 9:09 AM, Jed Rothwell <jedrothw...@gmail.com> wrote: > Neural network improvements to Google translate are described here. Look > at the sample sentence. > > https://blog.google/products/translate/found-translation- > more-accurate-fluent-sentences-google-translate/ > > See also: > > https://www.nytimes.com/2016/12/14/magazine/the-great-ai-awakening.html > > https://www.cnet.com/news/google-translate-machine- > learning-neural-networks/ > > - Jed >