Yes, the advances in neural machine learning are real but in substance they have been around for decades and are now emerging having been submerged in the noise. For instance, read Jürgen Schmidhuber's timeline of "deep learning <http://people.idsia.ch/~juergen/firstdeeplearner.html>" going back to 1991. Although I had the highest performing neural image processor as of 1990 (3e9 connections per second multisource image segmentation, Neural Engines Corp of La Jolla presented at the IJCNN in San Diego) my important contribution was circa 2005 <http://prize.hutter1.net/#history> setting up an AI competition based on Schmidhuber's colleague, Marcus Hutter's notion of universal artificial intelligence driven by Ockham's Razor (Kolmogorov Complexity). Hutter's criterion has now been adopted by the "singularity" folks <https://www.youtube.com/watch?v=V6umr1OP8uo>.
Having said that, there are still important advances that are submerged in the noise -- not the least of which is Relation Arithmetic <http://www.boundary.org/bi/articles/Relation-arithmetic_Revived_v3.pdf> (which, btw, subsumes quantum theory hence quantum information systems) and even in neural systems, there is insufficient attention being paid to Hecht-Nielsen's Confabulation Theory <http://r.ucsd.edu/Cogent%20Confabulation.pdf>. On Tue, Jan 27, 2015 at 4:31 PM, Jed Rothwell <jedrothw...@gmail.com> wrote: > James Bowery <jabow...@gmail.com> wrote: > > >> For many day-to-day operations the responsiveness of systems like MS >> Windows has actually decreased. >> > > That is because they keep adding bells and whistles. It is feature-itus > run amok. When I installed recent versions of Windows I went through and > turned off a bunch of features and it went back to working quickly again. > The features that slow down my computer the most are ones that display all > kinds of useless clutter on the screen. I think Windows was going slowly > because of a bottleneck between the CPU and the screen display, rather than > bad programming *per se*. It may be a problem on my computer in > particular, because it has an old, 2-port screen display card attached to > two large screens. > > > >> There are real advances in software but they're generally buried in the >> noise. >> > > I think there must have been some astounding advances in software > recently, judging by the results at places like Google and IBM. I mean, for > example: > > * Self driving cars. Many experts predicted this would take another 20 > years, but here they are, and they are reportedly safer than human drivers. > > * Google's uncanny ability to recognize faces, which is beginning to > exceed the human ability. > > * Google's ability to translate documents. This is still way behind human > abilities, but it is far ahead of where the technology was 10 years ago. > > * The Watson computer and its superhuman ability to win at Jeopardy and > diagnose diseases. > > They could not have done these things with hardware alone. Nor do I think > they could do them by brute force methods. Watson and the Google-Plex are > MPP computers, so however difficult it is to write MPP software, apparently > they are making progress in doing it. > > Google has published papers describing its MPP techniques. > > - Jed > >
