Dan has many good points as to why this kind of device is much
farther off than the writer thinks.
We at Smith-Kettlewell have been working on parts of this kind of
thing for years mostly using computer vision techniques.
The successes, too put it mildly have been lacking.
With all the power of a modern fast laptop and a
very clever algorythm, a computerized vision system takes 15 to
thirty seconds to tell if there is a print sign in a still photo
containing a clearly visible sign.
A recent test in which I participated as a subject was of a
system to try to find a cross walk at an intersection. This was
looking for the kind of cross walk with large crosswise stripes,
not just parallel edge lines.
The test was a complete failure, the thing was neither consistant
or accurate.
computers, no matter with what kinds of sensors, just don't do
well loking at the real world. They only do well dealing with
carefully devised set of specific circumstances from which they
must choose.
The one area that seems to be the exception to this conclusion is
speech recognition, that seems to be getting pretty darned good.
It would be very interesting to know, in the robotic car
challenge, just how many of the obstacles that faced the cars
were defined in the protocol and how many things were completely
unexpected.
How many pedestrians jumpped out in front of those babies when
they shouldn't have.
How many cars were where they shouldn't have been etc.
To summarize, I've heard this line of 'reasoning' for over 40
years now and I am not impressed with it.
It's a very long way off of reality
I remember a science fiction story from many years ago wherein
they had little such helpers that sat on your shoulder.
They helped you keep your schedules, did your phone calling etc.
Pretty soon somebody got the idea to have them hooked into your
blood stream to deliver appropriate psycho active medications in
case you got too excited, too rebelious, too sleepy etc.
If you dcan't see the ending of that one, you're too optomistic!
<GRIN>
Not holding my breath.
Tom Fowle
Smith-Kettlewell Rehab Engineering Research Center
Net-Tamer V 1.13 Beta - Registered
