Shadows of the Mind
Every year, I try reading at least one "epic" work that challenges my thinking processes. Last year, it was "On Intelligence" by Jeff Hawkins. That was a great work emphasizing a new approach to understanding and perhaps simulating intelligence: rather than being the result of computational procedures, intelligence is actually based upon the retrieval of stored memories that are utilized to make forward-looking predictions.
This year, I've been working through "Shadows of the Mind" by Roger Penrose. Like Hawkins, Penrose has little faith in computation (either human or computer) in order to understand such attributes as human awareness, consciousness, and intelligence. The difference is that while Hawkins lends some hope for a computer simulation of intelligence, Penrose explicitly rules it out. In fact, he mathematically proves that there are cases where an algorithm can't be proven, but the human mind can comprehend its inner workings and end results (see pp 74-75). He demonstrates this with a procedure that points to an algorithm that allegedly never halts, and thus continues forever. What he shows is that there are cases where the procedure simultaneously indicates the algorithm will halt and will not halt, hence inducing a contradiction that can only be resolved by admitting it continues forever.
Penrose discusses the famed Gödel Incompleteness Theorem and it's implications for formal systems established for proving mathematical propositions. He states that Gödel showed that the consistency of the formal system can't be proven by the rules or axioms of the formal system itself; that in effect, the proof of this consistency is external to the formal system. This is different than other interpretations I've read about his Theorem, namely that there will be at least one general theorem, based upon those rules, that will be true but can't be proven. In any case, he discusses these implications for computer intelligence and robotics and concludes that they can only be programmed with human directives and can't "think for themselves".
Step by step, Penrose dismisses any linkage between algorithmic computation and human awareness. Regardless of whether the system is sound or unsound, or is known or unknown, he is steadfast in his conviction that it can't be simulated. I have to say that I have some reservations about his claims. I'm not convinced with the complete soundness of some of his arguments here (have I myself fallen prey to Gödel's Theorem?!) and feel that he might be missing some logical cases that would weaken those arguments; but I myself wouldn't be able to construct a counterargument to them. Nevertheless, I'm impressed with his general line of reasoning -- he's a brilliant thinker and theoretical mathematician.
Penrose classifies 4 types of views of human awareness, consciousness, and intelligence:
1) These attributes are the results of computational processes. This is the strong view of Artificial Intelligence (AI).
2) These attributes can be simulated by computational processes. This is the weak view of AI.
3) Science can describe these attributes, but they are neither the result of nor can be simulated by computational processes. This is Penrose's position.
4) Science can't even describe these attributes. This is the "mystical" position, and Penrose claims that Gödel himself falls in this camp. He argued that Gödel's Theorem might point to the existence of a "mind" distinct from the brain, outside of the formal system!
The second part of the book, which I haven't read yet, starts with the assumption that consciousness can't be described in classical terms. Rather, it's based upon the indeterminism of quantum mechanics. Consciousness is the result of the collapse of quantum wave functions and the superposition of these events! He argues that these events occur in cellular structures called microtubules.
Well, I'll have more to say when I'm finished with the book. But for now, I recommend you read this book. It's more difficult reading than "On Intelligence" because of the formal arguments and mathematical reasoning presented here. Feel free to comment on the subject matter.