Artificial Intelligence and Marxism
I suggest those interested in these issues pick up Roger Penrose's two books on consciousness/computing, etc. Penrose raises some of the problems with current "connectionist" thinking. He also points in the direction of quantum computing as a possible explanation as to why the human brain is able to solve problems that now existing computer can. There are several books out since that time that carry quantum mechanical computing farther than Penrose did, building on the construction in laboratories of primitive quantum computers.
Quantum computers, taking advantage of some side effects of the uncertainty principle, are able to compute in a nonalgorithmic way. It is possible that quantum computers are actually making use of the interactions of matter/energy extended into dimensions that we are unable to reach, that is, if super-string theory is correct and there are really 10 or more dimension in this universe. By one popular theory of quantum mechanics, Everett's Many Worlds Theory, quantum computing could actually be interacting with matter in a multitude of other universes.
The Turing Machine, on the other hand is limited by the confines of this one universe, and by the fact that it must make its computations in chronological order, one after the other.
What all this means is that the goal of creating an artificial intelligence with the current digital technologies may be impossible.
People with deep streaks of unfortunate mysticism, like Penrose for instance, use that fact to try to make the claim that there is something special about the human mind.
In fact, however, quantum computers are already being created -- albeit in a primitive form. The dream of AI is not impossible, as many supernaturalists might have you believe. But the dream of creating AI with the current model of computing is impossible.
That only means we need to turn our attention to quantum computing for the real answers to AI, or to some other model altogether.
What is obvious is that using only the simple methods of natural selection and complexity dynamics the human brain was able to evolve on its own within just a few billion years. With scientific method and intellegently directed selection, we should be able to beat nature at this game in short order -- not in just a few years or decades, but certainly it won't take us 4 billion years either!
Also, for those who are interested in this subject, take a look at the work of Stuart Hammeroff of the University of Arizona. Hammeroff's work is a central to Penrose's second book on consciousness. Hammeroff is an anesthesiologist who has studied the question of consciousness from the perspective of one who puts peoples' consciousness out for a living.
He noticed that the electrons in the cytoskeleton of brain cells (neurons) become frozen in place during total unconsciousness. From that observation he has done a lot of work on the idea that the movement of electrons in the cytoskeleton of the cells and the quantum states they occupy play a role in the brains information processing processes.
Hammeroff also points out that, while the connectionists contend that the neurons act like single switches in a parallel information processor, the neurons are actually most likely a lot closer to the action of a complete information processor, carrying out magnitudes of order higher numbers of decisions that the connectionist model can account for.
Consider, Hammeroff asks of us, that a single-celled creature seems to be able to make complex decisions, like moving toward food and away from danger. If the massive colony of brain cells is the direct descendant of the single cell creature, which undoubtedly it is, then it makes sense to assume that the single brain cell has at least as much computing capacity as does the single-celled organism.
If that is the case, then the dominant paradigm of brain science, the connectionist model, is wrong.
Since the science of quantum computing is so new, we don't really know yet if that theory can shed some light on how the human brain is able to solve so many complex problems that current computers are unable to solve. But I think we can be relatively safe in saying that the Turing Machine is way too primative to explain every way that every problem is solved. The theory claims that it should be able to solve any problem given enough time. The problem is, many of the problems would take an infinite amount of time to solve that way, and we don't have an infinite amount of time. In fact, we must think on our feet. Quantum computing may explain how problems can be solved without taking an infinite amount of time.
In solidarity, Chris Driscoll, P.S. see you at the Labor Party convention in Pittsburgh if you are going.