What else could be more exciting than having intellectuals (us) creating intellect? Artificial Intelligece (AI) is a scientific study encompassing philosophy, computer science, mathematics, and even history, biology or engineering. Perhaps the most far-reaching goal in AI is to build an artificial human being. Unfortunately (or maybe fortunately) we have not nearly reached this level. Perhaps the most important purpose of AI is to increase human understanding of learning, reasoning, and other cognitive processes. One day we may be able to answer the important philosophical questions that were once unanswerable. What is intelligence? Are machines intelligent? Are machines capable of conciousness? emotion? or even aesthetics? and if so, how can we build machines that have these animal/human characteristics of intelligence? These considerations must be taken very seriously. Many people believe that only people and animals can possess intelligence; almost all science fiction movies portray computers as zombie-like objects that can only deal with 1s and 0s. Machines can be intelligent. The intelligence that a chess-playing program could exhibit may not be the same as a human chess-player, but if we were to say that that chess playing program did not possess intelligence in terms of playing chess it would be ludicrous. However, it is a not uncommon to believe that machines can only "crunch numbers". If they ever do anything that would exhibit human intelligence, it is always superficial and menial. This misconception usually lies in the paradox that although our computers can make calculations millions of time faster than any average human being, they are yet unable to perform many simple tasks that humans can accomplish. We can build chess-playing programs that will take even the highest of chess grand-masters down, but we have yet to create a machine that can talk as prociently as a human of average intelligence. There are expert systems that exhibit problem-solving skills and other aspects of human intelligence. MOLGEN, (created by M. Stefik at Stanford University in 1979), is a program that plans scientific experiments to help molecular geneticists. This program actually developed a method for producing insulin in bacteria. It works by manipulating its huge knowledge base of molecular genetics to generate sequences of small steps (all of which are constrained by certain rules) that constitute plans that would help it arrive to a solution. This is not to say that if something so mechanical and "dead" can possess intelligence, then human beings are no longer special. All machines are subservient to us, we are far from the point of creating machines that will surpass the status of human beings. (How do you feel? Please take time to cast your vote on this subject). Some AI programs can be very flexible. They are able to adapt to certain situations, as well as perform a variety of tasks, very much like human beings. Though, usually, most AI programs are not very flexible, specifically, Symbolic AI systems. These programs are usually created to accomplish specific tasks for the betterment of our lives. If we wish to build a machine that can emulate human thinking, a Symbolic AI program which operates under a rigid set of rules is definately not the right approach. Human reason cannot be likened done to just a simple set of conditions and subsequent if-then actions. Although in a top-down general fashion this may be true. Suppose that you want to go to the mall. However if it was raining, then you would change your mind. It turns out that its not raining so you do. A system that operates under a set of rules will have great difficulty creating totally different, "out of the blue" rules that may be neccessary for learning and adaptation. A story understander such as SAM (Script Applier Mechanism) that utilizes conceptual representation may be told by a script that JOHN is a PERSON. Yet, does it really know that JOHN is really a PERSON? With legs, arms, a brain etc.? This is a problem that had perplexed many AI researchers. The solution to this problem was a more flexible approach to reasoning: the connectionist approach. In this system, meanings are interconnected and linked. Each definition of something gives a definition to something else. JOHN could be a PERSON but he could also be a DOCTOR. Where do these ideas come from? The answers to machine intelligence lies not in the thinking of machines, but the way we think. Back in the 50s, AI researchers were never concerned with learning. We were very content with our chess-playing programs and theorem provers. As more research had been done on understanding, it was suddenly discovered that learning could become a very powerful tool for intelligent systems. Many types of systems present different approaches to learning (see aspects of artificial intelligent systems). Generally speaking, learning is very important and almost essential if we expect a machine to adapt to its ever changing environment. It is impossible to endow a machine with all the thoughts and concepts that humans possess. If we created a system that could learn these things instead- by themselves, then our task is made much simpler. Through simple observations, or experimenting through trial and error, our machines can learn.
|
