Has anybody heard of the "Chinese box" thought experiment? It does a pretty good job of illustrating the difference between a simulation and actual understanding.
Imagine a man with absolutely no understanding of Chinese is isolated in a box. Messages written in Chinese are inserted through a slot in the box. The box is equipped with many, many manuals which tell the man in the box what Chinese characters should be written in response to a particular message, without describing what any of those symbols actually mean. The man simply sees the message, looks it up in a manual, copies the response written in the manual, and slips his reply out of the box.
The result is a man who might appear to understand Chinese, but in fact doesn't have a clue what any of those symbols mean. He doesn't have the symbolic links to associate the Chinese word for 'dog' with a furry, amiable quadruped.
Writing a program that can simulate the thought process of a cat is no different than writing a more complete Chinese phrasebook for the man in the box. They might be making better simulations, but they aren't teaching the computer to think.
"The worst part about these reviews is they are [subjective]--and their scores often depend on how drunk you got the media at a Street Fighter event." — Mona Hamilton, Capcom Senior VP of Marketing
*Image indefinitely borrowed from BrainBoxLtd without his consent.
