> Whoa. If we use "we do not have a full understanding of the laws of
> physics" as a rationale, then the substrate of an AI is equally suspect
> of being a non-formal system.
Not really. The point is, we don't know how neurons work. We don't know
what aspects of their function are important. Conceivably even unknown
subatomic quantum details could affect their gross macroscopic behavior.
This is not true of computers. They are designed to have known principles
of operation, to work in a strictly defined, algorithmic way. Like any
machine, they may be imperfect, but their imperfections are not fundamental
to their operations, rather the designers do their best to avoid them.
An AI which runs on a computer has a well understood, formal, algorithmic
system at its base. A mind running in a brain does not.
> Penrose's argument strikes me as exceedingly disengenious. It is the
> interconnections between neurons (rather than the neurons themselves)
> that result in the emergent information processing power of the brain.
This is one theory, but it is not known what the truth is. Neurons are
generally considered relatively simplistic nonlinear summation devices,
but there could be more going on than we are aware of.
> Those interconnections operate via molecular-level signaling, not QM.
> At some level quantum uncertainty affects everything, including a
> "deterministic" AI substrate. But for all practical purposes, a
> computer is a formal system, even if quantum fluctuations will cause it
> to flip a memory bit every few thousand years. Does Penrose present any
> compelling reason to think of the brain differently?
"Compelling," no. However he suggests some possibilities which can't
be ruled out. You have to assume (I think) that neurons carry more
of the burden of computing than in conventional theories. Within the
neurons there may be quantum systems which manage to stay coherent.
Some people point to the insides of microtubules as an area protected from
the environment; others suggest that water molecules near surfaces may
have a quasi-solid structure which could support quantum superpositions.
It's all quite speculative, especially combined with Penrose's theories
about quantum gravity solving the measurement problem. Nobody would
pay it much attention except that Penrose's Godelian argument seems to
be surprisingly well accepted (as is Searle's Chinese Room argument).
If you accept either of these, you have to either believe in zombies
(beings which act conscious but aren't), or else you have to find some
unknown physics which allows brains to behave non-algorithmically (I
would say magically).
Hal