Inseparability of Mind and Body

Hubert L. Dreyfus's On the Internet describes the chief problem of building human-like artificial intelligence: the human mind is inexorably linked to the body. (Dreyfus) Isolating the mind from the body is a fundamentally flawed approach to this kind of AI development because it ignores the aspect of the mind that develops solely because humans have bodies. As Dreyfus writes, humans have a perception of what it is to feel relaxed, or to feel pain, or to feel temperature, because of interaction between the body, the mind, and the environment. A piece of software, since it is physically formless and has no way of interacting with the environment, can only form abstractions about these feelings, most likely based on interpretations of numerical values (i.e. 90 degrees is hot, 30 degrees is cold). (Dreyfus)

Chatbots such as those used in the Loebner Prize and search engine helper bots operate despite this inability to interact with the environment. Instead, chatbots and helper bots must have information coded in to help them deal with the "feeling" words given to them by humans in order to gain a modicum of the common sense humans all share by having bodies. The most common approach is through symbolic programming, where a certain word or phrase represents a value the computer can use in its response algorithms. Textual search engines simply compare the search phrase with text in web pages, and image search engines commonly base their results on textual tag information attached to the image rather than the image itself. Chatbots operate in a similar manner, picking up on words used in a conversation and looking up a response based on the word's usage. (Dreyfus)

While this model of operation is sufficient for general searches or chats, the flaws in using this approach to build an intellectual equal to a human are apparent: every piece of "feeling" information, no matter how insignificant, must be given to the computer. Jack Copeland writes:

"A moment's reflection reveals that even the simplest activities and transactions presuppose a mass of trivial-seeming knowledge: to get to a place one should (on the whole) move in its direction; one can pass by an object by moving first towards it and then away from it; one can pull with a string, but not push; pushing something usually affects its position; an object resting on a pushed object usually but not always moves with the pushed object; water flows downhill; city dwellers do not usually go outside undressed; causes generally precede their effects; time constantly passes and future events become past events ... and so on and so on. A computer that is to get along intelligently in the real world must somehow be given access to millions of such facts." (Copeland)

An ongoing project, called CYC, has programmers hand-coding assertions about physical existence into an AI system. The goal is to have a system capable of using these assertions to make complex inferences, such as concluding that a person is wet because he ran a race and therefore was sweating. However, its creators estimate that it will take about 2 centuries of work before the system has the 100 million assertions needed to be useful. (Copeland)