[PREV - AUTO]    [TOP]


                                             January  5, 2004
                                              (or so)
"The Emperor's New Mind"
Roger Penrose

A book grouping toward a proof of a
notion that artificial intelligence
may be harder than we think it is,
because human consciousness may not
be quite as simple as we think it is.

Consciousness might be deterministic
but non-computable, an entity not
translatable into an algorithm
on a von Neumann machine.
And/or it might use quantum phenomena
in some fundamental way.

   A lot of AI skeptics (such as
   myself) have had at least a few          AI enthusiasts are
   vague thoughts in this direction,        tireless in sneering at
   (with a hand-wave at Goedel,             this point of view:
   typically).  Penrose does his best       clearly we are all
   to put it all on a firmer basis.         irrational vitalists who
                                            feel threatened at the
Penrose does not succeed in arriving        thought of competition to
at a proof of any of this, but he           human intelligence...
does do a great job of holding the          "Inherit the Wind", redux.
door open to the possibility that AI
may not be possible (at least not in           They don't recognize that
the form commonly imagined).                   their own deep convinctions
                                               on the subject are
But one thing that he does succeed at,         similarly irrational --
more than any other work I can think           they've jumped to a conclusion
of is he *never* *ever* overstates             without proof, and smile
his case.  He never tries to conceal           condescendingly at anyone
a problem with rhetoric, when he               who questions it.
waves his hands he waves them in
plain sight, always careful to                    Particularly interesting
distinguish between points of proof,              is the poor track
and items that are merely suggestive.             record of predictions
                                                  of AI.
Would that a few more writers could
summon up this resounding honesty.                   Skeptics point to this
                                                     as reasons to doubt
I haven't gone looking for rebuttals                 the assumptions in
yet, but I suspect that they're                      play.
simply redundant, because Penrose
himself has already indicated the                    Enthusiasts see nothing
holes in his argument.                               that even needs
                                                     explaining away.

Also, in order to make his case, he                  Okay, we were wrong
presents a concise summary of just                   *last* time, but
about all of the current state of                    *next* time for *sure*!
physics, and much of the fundamentals
of computer science.  He never shys                     NEXT_TIME_FOR_SURE
away from equations, but never gets
too heavily mathematical for a
general audience.

This alone is a pretty amazing
achievement... and at some points I
was wondering if this was in fact one
of the main motivations for writing
the book for him, to attempt to
spread some of his notions about the
future development of quantum theory.

By the time you get to the ultimate
chapter where the final case is made,
it's a bit of a surprise how
lightweight it seems... but I think
this is simply because Penrose has
touched on all of his points many
times... nothing has been held back
for the conclusion.  This is a large
dramatic flaw, and it probably
contributes to the vague feeling in
the air about this book, a sense that
it's very good, though of course it's
completely un-persuasive.

My favorite quote:

   The question of freewill is
   addressed directly in Chapter 10,        FREEWILL
   and there the reader will
   doubtless be disappointed by what
   I have to contribute.

     (page 170, in Ch 5, "The Classical World")

                                          (( I should really
                                             run through my tabs,
                                             and see if I can do
                                             a better job of
                                             summarizing the