[ExI] Some new angle about AI

[Stathis Papaioannou]

2009/12/30 The Avantguardian <avantguardian2020 at yahoo.com>:

> I appreciate that you accept that possibility that the brain may be uncomputable so I won't belabor that point. But randomness is not as cut and dry as you seem to think. The only measurable difference between randomness and predetermined chaos seems to be a priori knowledge of the system dynamics.

Yes, that's what I meant by saying it isn't possible to tell true
randomness apart from pseudorandomness.

Incidentally, it is possible to generate true randomness on a computer
by simulating something like the MWI of QM with a branching algorithm.
The true randomness exists relative to an observer embedded in the
program, while for an external observer (i.e. for the multiverse as a
whole) the program is completely deterministic.

> Furthermore chaos theory, essentially an uncomputable branch of classical physics as suggested by Serafino, blurs the distinction between randomness and undecidability by recursion.
>
> For example Conway's "Game of Life" is undecidable at high iterations despite using a very simple set of rules which could be considered to be a canon of physical law in the greatly simplified microcosm of the game. In other words, based on an initial configuation of pixels the only way to determine whether the resultant pattern will stabilize as an infinite loop or go "extinct" is to actually run the program and see what happens.

If the brain is computable it does not necessarily mean there will be
computational shortcuts in predicting human behaviour. You may just
have to simulate the human and let the program run to see what
happens.

> Similarly despite the fact that fluid turbulence is known to be governed by the Navier-Stokes equation, turbulent flow is similarly uncomputable.. This is the main reason why long-term weather prediction is not currently possible and why one cant predict the shapes of clouds or the dispersion patterns of fallen autumn leaves. These are all *deterministic* systems yet the mind is so boggled by the recursive complexity and sensitivity to initial conditions thereof that it hides its ignorance under the security blanket of *randomness*.

These chaotic systems *are* computable, strictly speaking, since given
initial parameters and the laws of physics you can predict exactly
what they will do. The problem is that the systems are so sensitive to
initial parameters that no matter how accurately you measured these
the model would diverge from the original after only a short time.
However, the model would also diverge from the original if you
attempted not a computer simulation but an atom for atom replica, so
the problem is not with computability per se.


-- 
Stathis Papaioannou