Re: Human design and natural "design"

dkomo wrote:
> >>Of course that's what I meant. That was clear from the context. And
> >>there are some major differences between genetic algorithms and
> >>biological evolution such as the fact that in nature the fitness
> >>evaluation supplied by the environment is non-constant. In other words,
> >>selection varies.
> >
> >
> EKurtz99@xxxxxxx wrote:
> > This is also true for some GA's; for example in the checkers playing GA
> > system of Chellapilla and Fogel, which gets a brief mention in
> > Dembski's NFL.
> > http://www.natural-selection.com/Library/2001/IEEE-TEVC.pdf
> >
> > "Evolving an Expert Checkers Playing Program Without Using Human
> > Expertise
> > Kumar Chellapilla and David B. Fogel
> > Abstract-An evolutionary algorithm has taught itself how to play the
> > game of checkers without using features that would normally require
> > human expertise. Using only the raw positions of pieces on the board
> > and the piece differential, the evolutionary program optimized
> > artificial neural networks to evaluate alternative positions in the
> > game. Over the course of several hundred generations, the program
> > taught itself to play at a level that is competitive with human experts
> > (one level below human masters). This was verified by playing the best
> > evolved neural network against 165 human players on an Internet gaming
> > zone. The neural network's performance earned a rating that was
> > better than 99.61% of all registered players at the website. Control
> > experiments between the best evolved neural network and a program that
> > relies on material advantage indicate the superiority of the neural
> > network both at equal levels of look ahead and CPU time. The results
> > suggest that the principles of Darwinian evolution may be usefully
> > applied to solving problems that have not yet been solved by human
> > expertise."
> >
> > Since the candidate neural nets play against each other and the best
> > survive to the next round of play, the competive environment is
> > constantly changing - ie there is no fixed "fitness landscape" which
> > the GA is searching for an optimum solution.
> >
>
> Nope, I disagree. The key is in the phrase "the evolutionary program
> optimized artificial neural networks to evaluate alternative positions
> in the game." The GA had a fixed set of built-in fitness functions
> which it used to optimize the neural networks.

Don't understand; what is the "fixed set of built-in fitness functions"
you are talking about? The equations that determine the output of a
node in the NN based on its inputs? These cannot be called fitness
functions.

> The improving chess
> skill was stored in the weights of the NN's, which constantly changed.

Actually checkers, but the principle is the same.

> Incidentally, Samuels had a chess playing program back in the early 50's
> which could play against itself and learn how to play better chess. It
> ran on computers which probably were no faster than a $5.00 calculator
> you can buy at Walmart today. I wonder if the evolutionary programs
> above are any better than the Samuels' program. I heard it could play a
> pretty mean game of chess.

Don't you mean checkers? If so, this appears to be an urban legend.
Samuels program was apparently successful against an inferior player,
but this limited success was talked up until everyone believed it was a
world-beater. See "Blondie24: Playing at the Edge of AI" by David B.
Fogel.

> > More importantly, they challenge the notion that the combination of
> > random processes and competitive selection cannot generate complex
> > specified information.
> >
>
> How can you generate something you can't define? AFAIK Dembski is the
> only man in the world who understands what complex specified information
> is. :>)

That doesn't matter in the case of Chellapilla and Fogel's system,
since he concedes (in NFL) that a checkers-playing algorithm that can
equal human expert play must contain CSI. However, he obfuscates when
it comes to explaining where the CSI came from (given that no checkers
CSI appears to have been "smuggled" into the GA NN system).

"But did the evolutionary checker program of Chellapilla and Fogel
achieve its superior play without commensurate input from prior
intelligence? If one looks at how Chellapilla and Fogel actually
programmed their evolutionary algorithm, one finds that they instituted
a rating system (like the one used to rank chess players) that
continually tracked how well a given neural network (i.e., candidate
solution) was doing.(In place of a fixed fitness function, Chellapilla
and Fogel therefore defined what might be called a "floating fitness
function," or what Stuart Kauffman calls a coevolving fitness
landscape. But the mathematics of evolutionary algorithms is the same
whether the fitness functions are fixed or floating (see section 5.10
of my forthcoming _No Free Lunch_)."

"The important thing to note about these ratings is that they are fine
grained and specify very precisely how well a candidate solution is
doing with respect to other possible solutions. It's not as though
there are only two or three discrete categories for ranking solutions.
Instead there is a whole series of numbers ranging from 0 to 2400 and
above in which higher numbers correspond to superior skill and
expert-level play corresponds to between 2000 and 2199 (master play is
ranked 2200 and above). Consequently, finding an optimal solution here
is like the old Easter egg hunt game in which one is told either
"hotter" or "colder" depending on whether one is getting closer to or
farther away from the hidden prize. There is an incredible amount of
complex specified information packed into a fitness function (whether
it's fixed or floating) that for every pair of elements in a solution
space can tell you which is superior. What's more, any evolutionary
algorithm capable of precisely implementing such a fitness function by
preserving only the superior and weeding out all the inferior is making
full use of that information (Chellapilla and Fogel's algorithm did
just that; note that natural selection in biology operates with nowhere
near this precision). Again, there is no free lunch here -- complex
specified information has not been generated for free. "

None of the above seems to me to be relevant to the issue. The weeding
out is simply the result of competition; what else would checkers
programs do other than play checkers? How does a scheme for playing one
checkers program against another and assessing the result embed
checkers CSI? The same scheme could be used for any other game that
involved 2 players - chess, go, backgammon - is CSI for all these games
"smuggled" into the program as well?
If Dembski really understood what CSI is and could compute it, he
should be able to show mathematically that the CSI in the output of the
C&F system is implicit in the way the system is constructed; instead he
takes refuge in a weak hand-waving verbal argument.

.

Relevant Pages

  • Re: Evolution article in BioScience Encyclopedia
    ... The only way to empirically refute Darwinian evolution (which remains the ... omitted from Neo Darwinism fitness "Total Darwinian Fitness". ... i.e. "natural selection" cannot operate directly on each genomic ...
    (sci.bio.evolution)
  • Re: This snag EPTly dissolved for you - once again! (was:Hamiltons Rule In The Mirror)
    ... > The above can only argue that because all of life has persisted life ... > broken up into contesting forms each with their own independent fitness. ... > all empirical fitnesses remain epistatic, every selectee must compete by ... particular intellectual (evolution theorizing) AEVASIVE style. ...
    (sci.bio.evolution)
  • Re: Hard science of evolution
    ... known for making pronouncements about the theory of evolution without ... they look at the math carefully and notice the unrealistic ... And show that fitness really is heritable. ... Heritability of fitness seems to be pretty small. ...
    (talk.origins)
  • Re: Stem cells and Human evolution
    ... Darwinian fitness as I have objectively ... Darwinian evolution by natural selection be halted. ... For Darwinian fitness a highly fecund tic and a lowly ...
    (sci.bio.evolution)
  • Re: Publishing scientific information
    ... selective act by an establishment peer review process. ... a highly fitness mutualised species like our own. ... This also applies to the history of science. ... cultural evolution unlike organic evolution, ...
    (sci.bio.evolution)