Notes by RBJ on

The Emperor's New Mind

by Roger Penrose

I first looked at this book not long after it appeared as a paperback in 1990. At this time I was interested in popular works on physics and astronomy, and bought the book primarily to discover what Penrose had to say about physics. I made the mistake of starting at the beginning, and was sufficiently unimpressed by what Penrose had to say about AI, logic and mathematics that I never reached the chapters on physics.

I looked at the book again some years later when my interest in philosophy and AI was rekindled. On this occasion I was interested to know what Penrose's position on AI was. This was mainly because it does crop up on newsgroups which I read. Knowing a little more exactly what his position was had become of interest.

On this second reading I found that both in AI and in physics Penrose seemed to me to be raising some issues which are of interest. I am tempted to think that it is precisely because Penrose is not a philosopher that he is able to do this. The other notorious alien intervention into AI, was by the philosopher Searl who was sufficiently well trained to raise a completely uninteresting question, and still think that he had a substantive point to make.

Unfortunately, my circumstances when looking again at Penrose did not permit me to make notes, and I suspect that Penrose will slip down the priority stack at the same time as I become able in principle to make notes. Real notes may be a long time coming.

Meanwhile, here are the chapter headings.

  1. Can a Computer have a Mind?
  2. Algorithms and Turing Machines
  3. Mathematics and Reality
  4. Truth Proof and Insight
  5. The Classical World
  6. Quantum Magic and Quantum Mystery
  7. Cosmology and the Arrow of Time
  8. In Search of Quantum Gravity
  9. Real Brains and Model Brains
  10. Where Lies the Physics of Brain

Can a Computer have a Mind?

This chapter sets the agenda, which is to argue against the claims of AI and of strong AI. The descriptions of AI, and of strong AI, given by Penrose suffer from the weakness that they appear to be compilations of views of many different workers in the field, but are presented as though all these workers were unanimous on all these matters. Penrose could easily be construed as putting up a paper tiger.

However, the argument which follows (in the rest of the book) is clearly intended to dispute the claim that intelligent behaviour is Turing Computable, so Penrose does not exploit all the weaknesses in his own description of AI. Penrose appears to believe that he is attacking strong AI as defined by Searl, but as far as I can see he has not fully understood Searl's definition or his position. In fact, rather than hypothetically accepting the feasibility of AI (taking this to be the claim that intelligent behaviour is turing computable) to strengthen the argument against strong AI (viz the claim that a turing machine could not only exhibit intelligent behaviour but also actually be intelligent). Penrose's arguments are stronger than is necessary merely to attack strong AI, providing a basis for an attack on AI in general.

UP HOME © RBJ created 95/5/10 modified 97/2/3