Theory and Practice
Some time ago I was flying in an aisle seat when a tiny stewardess was having a tough time closing the overhead compartment directly above me. I happened to be reading a book on Theory of Vibrations. The stewardess would grab the compartment door in both hands and do a massive two-handed slam. From my vantage point directly underneath, I could see that the force of her attempts was generating rather large waves in the compartment door.
After three or more tries she stopped to catch her breath or perhaps to contemplate her choice of career, and I reached up, gently pulled down the door, and pushed it shut with one finger. By this time her antics had attracted the attention of several nearby passengers who burst into laughter. I gave her a smile, and she said something like "I'll get you," and went back to her duties.
My point here is that our actions and decisions are guided by a theory of the world. Partly because of the book I was reading, I could form the idea that her energetic attempts were counter productive. Of course my smart alecky method might well have failed if conditions had been slightly different.
Teaching computer to reason in general contexts has not had the kind of success that that they have had in teaching Go, Chess, and even mathematical integration. Computers have had outstanding success in areas where knowledge is tightly constrained, including important tasks like radiology. In more open ended areas we often say computers lack "common sense."
I think that it might be more precise to say that they lack a theory of the world. Much of our common sense reasoning is really reasoning from our theory of the world. Much of the education of doctors consisting of constructing for them a theory of the human body - anatomy, physiology, and more arcane details of function. Similarly, mechanics need a theory of the automobile engine and other key components.
The most complex system most of us deal with is the mind of another person. We mostly seem to have some built in hardware tuned to help us develop a theory of mind, but some persons seem to lack it.
After three or more tries she stopped to catch her breath or perhaps to contemplate her choice of career, and I reached up, gently pulled down the door, and pushed it shut with one finger. By this time her antics had attracted the attention of several nearby passengers who burst into laughter. I gave her a smile, and she said something like "I'll get you," and went back to her duties.
My point here is that our actions and decisions are guided by a theory of the world. Partly because of the book I was reading, I could form the idea that her energetic attempts were counter productive. Of course my smart alecky method might well have failed if conditions had been slightly different.
Teaching computer to reason in general contexts has not had the kind of success that that they have had in teaching Go, Chess, and even mathematical integration. Computers have had outstanding success in areas where knowledge is tightly constrained, including important tasks like radiology. In more open ended areas we often say computers lack "common sense."
I think that it might be more precise to say that they lack a theory of the world. Much of our common sense reasoning is really reasoning from our theory of the world. Much of the education of doctors consisting of constructing for them a theory of the human body - anatomy, physiology, and more arcane details of function. Similarly, mechanics need a theory of the automobile engine and other key components.
The most complex system most of us deal with is the mind of another person. We mostly seem to have some built in hardware tuned to help us develop a theory of mind, but some persons seem to lack it.
Lovely blog you havve here
ReplyDelete