Thursday, September 4, 2008

Artificial Intelligence has more untapped potential for causing social change than any other area of information technology. Not only does the development of AI allow games to become more complex, and the situations found within more novel and sophisticated, pure AI could revolutionize our world, and our view of ourselves. AI has developed to the point that an intelligence has been developed to play on second life, where it reasons and learns as a four-year-old human would.[1]

The very first AI, developed even before the term "Artificial Intelligence" was coined, was the "Logic Theorist" which was written in 1955-56.[2] The Logic Theorist was a search-tree based AI that used heuristics to limit the exponential growth of brute-force calculations. The use of Logic Theorist was to prove (again) some of the mathematical theorems in Principia Mathematica. The program eventually did prove 38 of 52 theorems, and even found a more elegant solution for one of the theorems.

The Logic theorist worked by beginning at the initial hypothesis of the theorem it was working on, and then would branch out by applying a different rule of logic to the statement drawn at the previous "node" of the tree that is constructed by this method. it would then check each new node that branched off of the previous ones against the desired conclusion of the theorem. if it did not find the conclusion, it would iterate through again, creating another set of nodes, causing the tree to exponentially grow in size every step. The problem with using this brute-force method is that it takes a long time to process that much information, and exponential growth can be surprisingly burdensome. To counter this, the developers of Logic Theorist "taught" it a few tricks, how to recognize certain kinds of logical dead ends, and in this way the program was able to decrease its load to a more reasonable level. These tricks are called heuristics, which are often referred to as rules-of-thumb.

The Logic Theorist was met with moderate recognition in the scientific community, and a paper documenting the new proof of theorem 2.85 from Principia Mathematica was turned down on the grounds that a new proof for a fundamental theorem was not worthy of publication, despite the fact that the computer program itself was listed as an author to the paper, along with its programmers.

The Logic Theorist was the first indication that computers could be programmed to mimic reasoning and thought, and that computers have the capability to rival humans in areas other than massive calculations. It has been said that the success of the Logic Theorist is"proof positive that a machine could perform tasks heretofor considered intelligent, creative and uniquely human." The basic methods developed for the Logic Theorist have been used by AIs since then, notably Deep Blue, a chess-program who beat Gary Kasparov.[3] and most video game artificial intelligences are at least influenced by the Logic Theorist and its early use of heuristics. Most every video game AI is faced with a multitude of choices many times each second that the AI is not engaged in another action. developers of AIs have learned from the Logic Theorist, and have been developing better heuristics and methods of decision making for AIs ever since. The programming language developed for the Logic Theorist later influenced LISP, which is still used in artificial intelligence research to this day (We use it here at RPI in the Cogworks lab).

But the real importance of the Logic Theorist is not on the mechanics of how it accomplished what it did, or the programs, languages, games, and innovations that followed it, but rather the immense potential of the field of Artificial Intelligence, both realized and unrealized. Currently, AIs are used in making weather forcasts, air traffic control, security (face/fingerprint/handwriting recognition), stock trading, robotics[4], and manufacturing. Clearly AIs have a daily impact on society and our quality of life.
AIs also keep many sites such as Google running, without spidering AIs, humans could not possibly track and index a fraction of the links that are currently in Google's databases.
AIs are extremely important to the optimization of our transportation networks, which get all of our consumer products where they need to go, as well as our consumables such as food and gas.
If you dial 411 on your phone, an Artificial Intelligence handles the language parsing and speech recognition to hear and understand what you say, but also a searching AI to find what you want and return the information to you. The GPS you use to keep you headingwhere you want to be uses Artificial Intelligence. Relatively soon, some companies think that cars will drive themselves using rigorously tested AIs. All of these things are tangibly close, if not already here, but the most exciting potential developments in AI, originating all the way back with the Logic Theorist, are those that are predicted for the future.

When an Artificial Intelligence passes the Turing Test, AI will have developed sufficient reasoning ability, language parsing, and symbol manipulation skills that some will begin to question human intelligence, and how different machine and human intelligences really are. Human perceptions will be shattered if an AI ever becomes self-aware and also has the ability to tell us so. Some futurists have predicted the event that all human and machine intelligences merge into one, Singularity. These events, if they do happen, will do so because of a computer program from the mid-1950's that re-did a couple of math problems out of a book. The academic project that was the Logic Theorist is giving way to a new generation of human-simulating intelligences, like the Second-Life avatar mentioned at the beginning of this post, and social simulating language parcers like Facade. From just these two examples, its not difficult to imagine (at least not too difficult) a world where we interact daily with purely electronic intelligences, and provided that the physical end of robotics keeps up, artificial intelligences expressed through and carried in physical and mobile robots.

No comments: