dsmith@GELAC.ARPA (Mr. David Smith) (04/17/88)
I have an interesting piece of software which has caused me to think at various levels about self-awareness. First, I will describe the software. It is an aid for learning to type correctly. It types a word randomly selected from a large dictionary, and waits for the user to type the word, checking that each letter is correct and requiring the use of the backspace key to correct any errors. So far, a useful tool. Unbeknownst to the user, it is also measuring his proficiency in typing each letter combination and adapting its dictionary selections to emphasize those he is weak at typing. The first question is this: "Is this AI?" to which the most likely answer might be: "You have not given me enough information." The second question then arises: "What would you need to know about this program to decide whether it is AI or not? " Its size? language? The method for determining proficiency? The adaptation technique? The size of the dictionary? Where the 'expertise' came from? The third question: "Why are we in the AI community always asking the first two questions?" Can you measure the "AI-ness" of something by its behavior, its structure or its source of wisdom? Should not AI as a scientific discipline be content, as the manufacturers of pencils, compilers etc. are, to become an integral useful part of engineering or computer science? David Smith. David Smith: dsmith@gelac.arpa