http://blog.wired.com/wiredscience/2009/01/retailrobots.html#previouspost

…higher productivity, expected energy savings, increased security (less pilferage of products).

There’s the future. In fact, I expect that in less than 2-3 decades, it will be possible to “virtual grocery shop,” with the products taken from a completely automated warehouse (incredibly densely packed, and dark, with no parking lot), and delivered to one’s door by a autonomous vehicle.

My personal contribution toward this goal was a low level module that transforms pixel based images to symbolic data similar to the function of the human retina. Although implemented in a security system the software was structured to allow use in a generic visual perceptual system.

One more item of interest, at least to me: I believe that new information is processed in the context of prior trusted knowledge. If that is true could we achieve intellectual diversity in the laboratory by loading different robots with different memories, even if they all have the same logic inference engine?

This is one more reason why anyone who is really interested in predicting future events should read a whole lot of what Ray Kurzweil has written.

For instance, how many mammograms does a world-class radiologist review in a career? Well, why not have a whole team of world-class radiologists judge a set of mammograms, then tie in their judgements with the image of the mammogram? This is all tremendously complicated, I’m sure.

But the end product is that one has a computer that has the knowledge of a whole team of world-class radiologists, and can read tens of thousands of mammograms an hour, and also learn and remember all the outcomes of all the cases that were related to those mammograms.

That eventually means anyone in the world has access to a computer “expert” that has read 10s of millions of mammograms. And never sleeps or blinks, or gets distracted. Amazing.

http://radiology.rsnajnls.org/cgi/reprint/191/2/315.pdf