The Nature of the Beast
Our course began by examining the history and social impact of computing which led us into a deeper and deeper investigation of what computers are really all about. As we draw to a close we are ready to come full circle, looking at the history (and plausible future) of intelligent computing and its impact on people. In doing this, we must recognize that computers are very capable of performing millions or billions of computations during a single heartbeat. But does this awesome computational power lead us to the conclusion that the machine is intelligent?
With this question in mind, it is instructive to begin the discussion of artificial intelligence (AI) by first considering what we mean by natural intelligence. Consider, for example, the following questions:
After considering these questions, we are led naturally to ponder, "Is a computer intelligent? How do you know?" For the purposes of discussion, let's assume you say, "No. Computers are not intelligent." Then a second question appears. "What added capability would a computer have to possess before you might be persuaded to call it intelligent?"
Research - The First Fifty Years
The AI community's research interests have evolved into two main branches:
Researchers in this area aren't very concerned about the philosophical question, "Is the machine intelligent?" Instead, these researchers tend to focus on the development of software which embodies a knowledge base of an intelligent human. In other words, without necessary regard for how a human might formulate her solution to the problem, these systems attempt to simulate the same result. Two techniques are especially helpful in this regard, brute force computation and expert systems.
Brute force
In these systems, developers harness the awesome computational speed of the computer to allow it to behave "intelligently." The most famous example of this approach is the chess playing system named Deep Blue. IBM developed this system which beat Garry Kasparov, the world's chess champion, in a regulation match. Deep Blue does not solve the problem of playing chess in the same way that a human would. Deep Blue calculates the outcome of millions of possible moves and then selects the "optimal" move. A human chess player plays much more intuitively and only considers a small number of possible moves. Nonetheless, the world's greatest chess player was beaten by the system.
Expert systems
In these systems, developers attempt to embody the knowledge of a human expert into a system that will mimic the deductive reasoning chain used by the human to solve the problem. It has been said that an expert system could be easily constructed for any problem in which a person might consult an expert for advice over the phone. This may be a serious underestimate of the power of the technique, however. Some of the most succesful expert systems have been developed in the fields of medicine, chemistry, mathematics and investment management.
Researchers in this highly interdisciplinary field hope to understand how humans think. The primary goal is not to build a machine that does something as well or better than a person, but to build models which help us to better understand how humans work. In the process, of course, new tools will be developed which will have practical outcomes.
AI researcher Marvin Minsky has proposed that the brain works by having hundreds or thousands of expert systems, which he calls agencies which interact with one another to give us our consciousness and intelligence. Minsky has dubbed this cooperative arrangement the Society of Mind.
Neural network architectures learn to behave intelligently by modelling "the way the densely interconnected, parallel structure of the mammalian brain processes information."
A related topic is that of machines which are more human physically. For example, Cynthia Breazeal is conducting research at MIT towards the development of a sociable robot named Kismet. This robot uses facial expressions to indicate its "emotional" state. Meanwhile, Roz Picard's research group at MIT focuses on affective computing. In this work they are empowering the robot with the ability to sense and interpret emotional input from humans. One of the more ambitious projects is called COG, The Humanoid Robot. The fundamental premise of this research is described by MIT's Rodney Brooks: "The motivation behind creating COG is the hypothesis that humanoid intelligence requires humanoid interactions with the world."
It is interesting to note that the earliest proponents of AI, such as Alan Turing, envisioned systems which could interact so naturally with a human that the human would not recognize that they were dealing with a machine. Conversational systems such as ELIZA, PARRY, RACTER, and HeX might indicate that this truly is still a daunting task, even a half century after Turing's measure of computer intelligence was proposed. Sample conversations with ELIZA, PARRY and RACTER can be viewed at this link, while HeX transcripts can be obtained at the author's web page. One of the most interesting chatterbots is the system called ALICE which continues to grow as a collaborative project on the internet.
Other researchers note that creativity is a hallmark of human intelligence. Systems have been developed to explore machine creativity. For example, visual artist (and computer programmer) Harold Cohen has developed AARON which paints creatively following a set of rules engineered by Cohen. Inventor Ray Kurzweil has developed RKPC which composes poems following the style of human authors to which it has been exposed.
Research - The Next Fifty Years - Evolving human intelligence
 |
Futurists have always imagined a time when humanity will be freed to live life in happiness without any of our normal drudgery.
If machines could be so improved and multiplied, |
New futurists, such as Kurzweil, now imagine a time when our artifacts, the intelligent machines, are much more intelligent than mere humans (e.g., Will My PC Be Smarter Than I Am?). This potential outcome has led some writers to wonder whether our machines will one day be granted dignity (e.g., Isaac Asimov's novel Bicentennial Man or Roz Picard's essay Human and Machine Dignity). It gives one pause to wonder if these super-intelligent machines would even care since they might be too busy tending to one another's needs.
A nervous smile here;
it will adorn your love
it will be as the stars
Smile Here; a poem written by Ray Kurzweil's Cybernetic Poet after reading poems by Various Love Poets