A DARPA challenge for machine vision?

So DARPA wants intelligent cameras (story on Wired and also Popular Science.) Well don’t we all?

I bring this to your attention for two reasons. First, I’m surprised by the naivety shown by this announcement. The gulf between processing an image to extract information and actually understanding the content of an image is so vast that I cannot conceive of it being bridged in the next decade, maybe longer.

Secondly, while I don’t buy in to conspiracy theories, this does sound like a government-slash-military-industrial complex-driven step towards a total surveillance society. “It’s for your own good!” I can hear them saying, and “It’s our job to protect you!”

Fortunately, while the technology exists to automatically detect human figures in images, we’re a very long way from a computer being able to tell what the human is doing. At the moment that can only be done by having people look at images.

On the other hand, this announcement from DARPA sounds like a terrific boondoggle for those who research and develop machine vision software. Throw some of that money my way!