A U.S. nuclear engineer and a Russian scientist say their firm will be the first to create the technology that finally succeeds at mimicking human sight.

That seems like a long shot for their Los Altos, Calif.-based startup Pixlogic, as well as for the myriad industry, government and academic researchers who have been pounding away at this technical challenge for decades. 'People have not been able to crack this problem,' admits Pixlogic founder and Chief Executive Joseph Santucci.

Indeed, computer vision was one of the original promises of artificial intelligence when it first was talked about decades ago. Nevertheless, efforts to achieve that far-fetched feat have resulted in some exciting new technology in the realm often referred to as visual search, or image understanding.

Pixlogic is backed by the Central Intelligence Agency's nonprofit venture capital arm In-Q-Tel, whose mission is to nurture technology that serves U.S. security interests--and to make money for taxpayers while they're at it. Santucci and his Russian-born chief scientist, Dr. Shelia Guberman--who invented the handwriting recognition technology used in Microsoft's Windows CE program--say that the CIA is now testing their software. In their words, the technology can be 'visually programmed' to monitor video feeds in real time to search for certain events or elements.

The software sees objects in a picture or video frame and makes a mathematical formulation to describe them. The formulations are stored in a searchable database and compared to the formulations for objects that are already filed. This is nowhere near mimicking human sight, but it's a start.

[...]

[CLB: And it's the first step in ongoing video surveillance tied in to identity recognition and activity recognition. What's this? Imagine you are at an ATM. The video monitor recognizes your face, perhaps as part of a larger identification and authorization process. As you go through your transaction, the same system watches your actions. Do you reach into your purse? Do you drop your card after removing if from the system? is another person with you? Are your lips moving - perhaps as you sing in the privacy of the ATM booth? Etc. What sort of information can be gleaned from your physical motion? And then associated with you for future purposes? In fact, could the way you walk in to the booth be used to uniquely determine who you in the first place? Visual recognition combined with searchable keyword correlations expands the privacy and security maze of ethics.]