Fair enough Tortoise. But a few comments. First, he is an indisputably brilliant man, and a very talented mathematician (his work in twistor theory is now being adopted by many in string theory, and may be a precursor and foundational building block to string theory). I have read his two books (the second of which was written to refute his critics) on this topic, and found his arguments (though I am NOT an expert in the field) to be compelling. I know that he has been attacked incessantly by those in the artificial intelligence field, because there it is politically incorrect to assert that computers won't lead to consciousness. So, I'm left with what I do know about Penrose vs. what you tell me you know. If you have good cites relating to refutations of his work in this area(and even better, to unbiased and professional syntheses of pro and con), please Freep-mail them to me. I'd be interested in reading them. You say "he was writing stuff.. that had already been rigously refuted by the experts in his field..." Well, a prior refutation of the argument that algorithmic calculation cannot be foundational for consciousness would be something like a proof that it can be. To my knowledge, such does not exist. Lastly, though this is not my area, I freely acknowledge, I have personally always been skeptical of the idea that computational algorithms can give rise to consciousness - and like Penrose, I do believe that there will be a quantum mechanical component involved (for reasons outside of what Penrose was talking about). Notably, I have read of many AI people over the years who claim that we just need bigger computing machines, better algorithms and more complex neural networks before consciousness spontaneously arises in machines. I personally think this is hogwash, and note that now that we have those powerful machines, nothing of the sort has happened. One of us is wrong about Penrose's assertion. Time will tell who is right!
I don't dispute this.
I know that he has been attacked incessantly by those in the artificial intelligence field, because there it is politically incorrect to assert that computers won't lead to consciousness. So, I'm left with what I do know about Penrose vs. what you tell me you know.
There are a number of very specific problems with his ideas on quantum consciousness and the non-computability of consciousness. Some of the original dialogs regarding this are probably not preserved online due to their age. Two problems in his core assumptions that he has been soundly criticized for:
The idea that he proved the non-computability of consciousness with Godel's theorem is false. First, his "proof" was flawed and you can find rigorous mathematical deconstructions of his proof on the web using Google. Second, implemented universal state machine designs exist that do not have a property that he was assuming for his theory. However, I don't hold this part against him because he wouldn't have known about such things at the time he wrote his book; it was an esoteric "hard problem" of computer science that was recently solved.
Quantum behaviors in the brain cannot be a meaningful factor in consciousness/intelligence. This was first pointed out a long time ago by physicists/engineers familiar with neurophysiology. If you do the calculations, you find out that any quantum mechanical behavior falls below the noise floor of the brain by orders of magnitude; thermal noise and other factors are sufficiently high that coherent information could not be stored in the quantum mechanical substrate. Other points worth making is that quantum computation can do nothing that current silicon can't do; all it does is change the computational complexity of certain classes of problems from exponential to linear and so forth (very important, but not revolutionary). I've seen numerous qualities commonly ascribed to quantum computation that simply aren't true.
Penrose didn't come up with the non-computability idea until after the problems with his quantum consciousness idea had been pointed out to him by others with intellectual clout. The problem with his quantum consciousness idea is that he didn't do his research before writing it. The non-computability argument was later taken apart after he published the book, but he was outside his area of expertise on that count. The problem most AI (and non-AI) folk have with his work in this area is that Penrose really wants to believe in an almost mystical view of consciousness and intelligence, to the extent that he stretches his credibility.
Notably, I have read of many AI people over the years who claim that we just need bigger computing machines, better algorithms and more complex neural networks before consciousness spontaneously arises in machines. I personally think this is hogwash, and note that now that we have those powerful machines, nothing of the sort has happened.
I agree with you on this point; I consider the above to be nothing more than excuses in general principle (the "why" part is lengthier than I want to go into now). There is a lot of weird religion in AI research, but there are also some very sharp people working in it as well. Nonetheless, the cranks probably outnumber the genuine article in that line of research.
One thing to realize is that although the problem of AI has been worked on for half a century, there has been no mathematical foundation for virtually all the work that has been done. In computer science, this essentially means that the research being done had little or no chance of amounting to anything. The first part of the problem is nailing down the definitions; people use terms like "consciousness" without really knowing what they mean. After you nail down the definitions, the second part is tying the definitions into the relevant mathematics. Still, most of the work being done has essentially been a "shotgun" approach to AI, throwing pellets in lots of directions hoping that you hit something.
In the last decade or so, and particularly in the last five years, the mathematics has finally started to catch up with the software design. It turns out that everyone was pretty much looking in the wrong direction, and we can now prove with a great deal of mathematical rigor why all the attempts to date never panned out. As a matter of bad luck or something, the solutions and mathematical definitions of the problem emerged out of an area of mathematics that has been largely ignored by computer science except in very narrow and mundane applications. The first direct overarching mathematics paper to outline the character of the problem (there exists a universal super-algorithm which can provably express all forms of intelligence, ironically in an algorithm space that has not really been explored in CS) was published around 2000 (although I know for a fact that the core mathematical concept was discovered a few years prior to that publication). So after almost half a century of wandering aimlessly, the AI community finally has a specific set of comp.sci problems to solve that will provably generate the results they are looking for. For better or worse, nobody is willing to publish much on algorithm development in this area, since the first real gold rush in AI has just started, though most people aren't aware of it. I come into contact with AI folks a lot, and am deeply involved in very relevant mathematics, and there is a bit of excitement in many circles that the first general AI technology is imminent; from their perspective, having a mathematically rigorous target has taken a great weight off their shoulders. Knowing what I know of the mathematics, I am a lot more optimistic about the outcome than I used to be (and early work in this area is looking far more promising than any past direction according to the people in the trenches).
You can find more information on various aspects on Google, though information on some things is a bit sparse on the web. I am very, very busy and really don't have the time to elucidate more or write a dissertation on the myriad of matters touched upon.