RE: <nettime> The Age of Spiritual Machines (Review)

["Hari Sundaram" <sundaram@ctr.columbia.edu>]

felix's comments on the kurzweil book are fascinating. I have not read the
book, but, as someone who has grappled with signal processing/pattern
recognition problems in the last ten years, I'd like to add that Mr.
kurzweil's optimism has largely bypassed the AI community. AI has evolved
into something that is very different, from what was intended in the
original Turing project.

The solutions that we have today in the AI community, work very well in
highly constrained environments where lighting, sound are all carefully
controlled. We are light years away from solving pattern recognition
problems in the real world, i.e. the highly unconstrained environment
found outside our house. Speech recognition works well, for example, when
*one* speaker trains the recognition system, in carefully controlled
environment. Try IBM via voice in the subway station, and you'll see an
immediate performance degradation. Speech recognition that is speaker
independent, works for multiple languages, and that works well in noisy
environment, is at the moment, a nice, pleasant dream.

 Secondly, the mere act of recognition is not the same as understanding.