Why Perfect Surveillance isn't Perfect Law Enforcement
24474
Science and Technology
Jun 16, 2007
Why Perfect Surveillance isn't Perfect Law Enforcement
Technology helps in many areas in police work. We have everything plugged in on the national Amber Alert system, which has proven its worth. We have databases everywhere in law enforcement, tracking known offenders of every stripe. We have the house arrest system now, where an arrestee is tracked with an ankle bracelet from a global positioning satellite system. We can punch in a license plate and pull up its motor vehicles record. But these are the obvious solutions. When we try to create "robocop", we run aground of the inherent limitations to artificial intelligence. So far, our most successful method of using automation to control crime is the traffic-ticket cameras, which automatically record a license plate when it sees a car breaking the speed limit or blowing through a red light. A system like this is costly to implement and riddled with bugs, never as reliable as an officer sitting at the intersection. But at least it's partially effective, and we're getting the kinks worked out. Another attempt at automated policing is even more troubled. Face recognition software is still in its infancy, and it's likely to stay that way. There are simply too many variables involved in matching a picture of an offender with the same offender passing in front of a camera. Even though recognizing a face is something that we humans do very easily - even at the age of a few days - it turns out that our brains are biologically programmed for just such a task. This is why so many "sightings" of holy figures occur in such mundane places as a wall and a slice of cheese toast, why we see a "man" in the moon, and even why smilies [ :-) , :{| , 8^D ] work. The example of a smiley is one particularly good example. Given that you could tell the computer that the ASCII bit patterns of a colon, hyphen, and right parenthesis equals a smiley face, could you then ask the computer to guess if the other infinite variations were smilies or not and have it be reliably correct? No, because nowhere in computer circuitry is there a natural idea of what a human face is. We have to explain everything to it, and no matter how thoroughly we explain the basic parts of a human face to it and which characters represent the basic parts, some wise guy will always come up with a way to hack the system: (-: Gotcha! So, with a camera image, a human face is a matter of many pixels of data. Edge detection is the first place where we have problems in automated recognition. It is very far-reaching to teach a computer to separate an object from the background in a 2-dimensional image. Then it has to find the face in two sample images in order to compare them. Now it has to determine a match, based on the facial characteristics of the two images. But the two images, even if of the same person, may have numerous details making it impossible to be certain about the match. The pictures will almost certainly have different lighting and angles. The suspect may have a different hair style, or be wearing a different shade of mascara, or have grown or shaved off a beard, or may be wearing sunglasses. These subtle differences seem very simple to us, but the human brain is actually thousands of times more sophisticated at interpreting visual data than the best computer we can build today. The only course we can pursue with computers is "fuzzy logic" matching, whereby we teach the computer to look for a "close enough", "almost", "similar to" conclusion instead of a perfect match. Fuzzy logic is referred to in artificial intelligence research as a "kludge", because to a computer, a bit is either 1 or 0 with no in-between. We have to tell it to report a statistical analysis of the matching bits and return a percentage indicating confidence instead. The ground-breaking science fiction movie "2001" has as a main character the HAL 9000 computer, which is almost a personified "wish list" of what we hope to someday make computers do. In the movie, the HAL 9000 is able to recognize another astronaut from a drawing, comment on the aesthetic values of that drawing, read lips well enough to reconstruct a conversation, and read emotions from a human face. The man-vs-computer scene towards the end even has HAL 9000 guessing astronaut Dave's emotions and trying to talk him out of shutting it down. "Dave, I can see you're really upset about this." Now, that's artificial intelligence! Other advances on the front of technology-aided law enforcement show some different directions we might take. Recent research has shown that we can "read" quite a bit of what the human brain is thinking by scanning it and imaging it. Of course, this is test laboratory conditions we're talking about - unobtrusive surveillance of human thought is still miles away. However, technologies such as MRI have been able to tell distinct differences in brain activity depending on what the test subject is thinking or feeling. Ironically, the day may come when it's easier to have a computer read a mind than it is to read a face! Conclusion: don't hold your breath waiting for perfect computer surveillance to become a reality. We might just eliminate the social causes of crime itself before we perfect technology to prevent crime from happening. You can see this as job security if you want to. Just remember that at the end of the day, computers are only as smart as we humans are capable of programming them!
Freelance writer for over eleven years. Security Shirts Formal Wear Uniform Scrubs
Write a Review
Add to My Favorite
Refer it to Friend
Report Article
Other links at Science and Technology
This article talks about some cameras that you might be thinking of selling.
Category:
Calibration is one process that is done in almost all industries, be it publishing, medical, or commercial. Unfortunately, calibrating is such a technical term that not all people know that it is actually being done even in their own homes.
Category:
Fax has certainly been a great blessing. Read on to find out why everyone swears by the fax on the Internet.
Category:
Site Sponsor
Directory Statistics
Articles: 68285
Categories: 501
Yahoo Entertainment
