When one mentions artificial intelligence to a layman, he or she may conjure up a Cylon "skin-job" battling on the planet New Caprica, or HAL in the film 2001: A Space Odyssey. If he or she is more knowledgeable, he may envision the February 10, 1996, chess match between the computer "Deep Blue" and the reigning world chess champion, Garry Kasparov. Or the avatar "Baley" featured in the "First Law" episode of NUMB3RS may come to mind, that of a computer that can pass the Turing Test, and maybe even murder its creator.

Such flights of the popular imagination showcase both the benefits and the failings of artificial intelligence ("AI"). AI can put on a pretty face, as in Baley. It can engage in conversations and run a space station as in 2001. In the popular imagination, AI machines can "frak" with your body and your mind. This brave new world of conversants and rivals, lovers and enemies, and noble opponents does not yet exist, and may never exist.

And yet, in the past two decades mankind has made immense leaps in what should be called Artificial Intelligence. This article posits the view that AI should not be views as imbuing machines with human like intelligence. Rather, AI should be view as imbuing humans with machine-like powers of pattern recognition, action and capability. Computer AI gives the human creature the ability to be aware of information and the potential to act in ways never before thought imaginable. Computer AI gives us the ability to manipulate machines with nuance and indirection.

In this article, I hope to highlight some of the truly revolutionary advances in automated pattern recognition, and demonstrate how these advances amount to an evolutionary advance in human intelligence and capability. This article will explore advances in computer pattern recognition and their implications for advances in the human condition.

BIG BROTHER IS WATCHING

AI is now routinely applied to crime-solving. At busy intersections and toll booths cameras routinely capture license data and photos of occupants. These images are fed to computers with AI pattern recognition that far exceeds human ocular capacity to translate the images into actual license plate numbers and traffic flow data. Such photo proof was recently used in a murder case to (1) capture the murder,and (2) defeat the murder defendant's alibi.

Now some would say this is not AI. In fact, the human programmer is the one who supplied the parameters to the computer program to help spot in the photo: (1) what a license looks like, (2) where that license plate can be found, (3) what numbers and letters look like, (4) how to identify the state, and (5) how to use fuzzy logic to identify the likely owner of that license plate. And yet, this is an artificial advance in human intelligence. For the computer can handle this process in a millisecond and can see with far greater resolution than the human eye.

One may debate whether this is good or bad for society. The good in crime solving is the flip side of the bad in a loss of privacy and freedom to travel (anonymously) of the individual. However, one cannot deny the "artificial" advance in human capability at detection and deterrence.

BIG BROTHER IS LISTENING

AI is now routinely applied to review of telephone and internet traffic. Ask yourself how the Department of Homeland Security has foiled recent plots against the United States and its citizens. Did the DHS have undercover agents who revealed those plots in time to prevent them? Most likely, the answers was electronic and internet surveillance. And how were the key recordings and communications identified. Some may have been part of wiretaps approved by the FSA courts. More likely, the initial leads were part of "web crawlers" and other "bots" that were sent out on a mission with search parameters defined by humans.

These "bots" were the real undercover agents. They infiltrated the domain of the "enemy" and revealed its secrets. They searched for "code words" using, you guessed it AI pattern recognition to identify matches and correlate them to produce reports for further analysis and follow-up by humans. The enhanced capabilities artificially enhanced the aural and oral capabilities of humans to detect potential crimes in advance of their commission.

Again, one may debate whether this is good or bad for a free society like the United States. The good lies in the prevention of terrorist acts and apprehension of terrorists before they commit crimes. On the other hand, is this AI giving us a future like that envisioned in The Minority Report where we apprehend and incarcerate people for "thinking" about the crime before their actions even rise to the level of conspiracy to commit.

LEGALZOOM, BLUE FLAG AND THE AUTOMATED PRACTICE OF LAW

AI is now routinely used in the delivery of legal services. At one end of the spectrum, Linklaters, a leading British law firm uses Deal Builder, an expert drafting system, to guide financial services company in creation of term sheets, loan documentation, and regulatory rules. At the other end of the spectrum, LegalZoom delivers legal documents to individual and small business owners that are created following the completion of an online questionnaire that is then processed internally through HotDocs.

I am expert in both DealBuilder and HotDocs, the document assembly platforms used by Linklaters and LegalZoom respectively. These tools are not "AI" in the sense that they "think" for you. Rather, both contain what is known as a relevance engine that determines from a list of potentially thousands of questions, which questions, documents, and clauses are RELEVANT, based on the answers by a human to prior questions. These systems "learn" in the sense that they can calculate, from human markup syntax, what a lawyer would think is the the next relevant question and what a lawyer would think is the relevant response.

FACES AND SPACES: EVOLUTION OF THE POINT-AND-SHOOT CAMERA

Pattern recognition is also making inroads in consumer products. Modern "point and shoot" cameras work really hard to make it easy for the casual photographer to capture great shots.

One feature most people have heard of is *face recognition*, which is (not surprisingly) the camera's ability to discern human faces in the scene. The camera will attempt to adjust lighting and focus to render faces clearly, balancing these settings if more than one face is in view.

Incidentally, this feature shows a subtle bias that's typical of products with "dumbed down" AI. In this case, the assumption is that people are always the most important of the scene. While this is usually what the photographer has in mind, it can lead to disappointing results if she's more interested in capturing the details of the Roman ruins than a nearby tourist eating gelato.

Face recognition has other uses that are less obvious. Some cameras now offer the ability to take a photo when a new face enters the scene, which is a clever alternative to the traditional timer when a photographer wants to join a group portrait.

Newer Canon cameras have a second pattern-recognition feature that helps to fulfill the promise of "point and shoot." The iSAPS technology ( Intelligent Scene Analysis based on Photographic Space), evaluates the overall scene and tailors the camera's automatic settings (exposure, focus, and white balance) to take a pleasing photograph.

© 2009 Canon U.S.A., Inc.Used only with express written permission

A diagram of the "Photographic Space" database behind Canon's iSAPS technology.

These choices are made based on simple comparison against a large database of past photographers' decisions, rather than any kind of modeling of aesthetic or physical principles. Still, this "dumb" system is very quick, makes better decisions than most amateurs, and may have some benefits as a learning too for the aspiring photographer.

AN ONGOING STUDY

This is a start at catologing examples of AI already in use, already providing services artificially enhancing the human capabilties. I welcome subission of other examples to include in this article.

Notes