On Fri, 15 Dec 2000, Steve Nichols wrote:
> > James Rogers <jamesr@best.com>
> >
> > More precisely, you can calculate the limits of predictability for
> > finite state machines, given any certain amount of memory to work
> > with. All finite state machines are predictable, but very complex
> > ones may have high predictive error rates on current hardware. With
> > sufficient memory ("sufficient" being a miniscule fraction of the
> > amount of memory required to map the entire state space), one can
> > predict what any finite state machine will do with a fairly high
> > degree of certainty.
>
> Exactly ..... whereas evolvable circuitry machines (silicon or biological)
> can be or approach infinite-state.
It is either finite state or not. "Approaching infinite-state" is
meaningless as a category.
> The mammalian brain is infinite-state
> in a way that a simple thermostat, or even a massive Turing machine,
> cannot.
I would say otherwise, and the evidence is not with you. Many analyses
have been done on streams of "random" data generated by the human brain.
Every single such test that I am aware of has demonstrated that a computer
(using predictor algorithms) can predict the next piece of "random" data
that a human will generate with a predictive error rate *much* lower than
what would be expected mathematically if the data source was truly random.
In fact, this is a usable mathematical definition of "finite-state
machine". So, while this is by no means proof that the human brain is
a finite-state machine, no one is able to produce evidence to suggest
otherwise and plenty that suggests that the mind *is* a finite state
machine.
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:50:37 MDT