I was reading through a creation/evolution discussion board recerntly and the website creationsafaris.com came up. Creationsafaris is the baby of David Coppedge, a computer scientist employed by NASA, who also happens to be a rather adept disinformation peddlar for Christ.
Now, I've read some of Coppedge's stuff before, but I'd not visited his site for some time, so I decided to pay a visit and I was reminded of why I hadn't been there in a while - to describe the site as inflammatory is putting it mildly. I searched his site for some issues of interest to me, and no surprise, I was soon rewarded with a number of highly misinforming/disinforming smug dismissals of legitimate science and that characteristic cherry-picking of quotes to employ as 'look how little evolutinists know!' fodder.
And let us not forget his condescending name calling and the like.
Take a look at this masterpiece of flim flammery, sleight of hand, and inflammatory disinformation, my comments interspersed:The Hopeless Task of Building Evolutionary Trees
07/25/2002
A paper posted in the online early addition July 25 of the Proceedings of
the National Academy of Sciences starts out with an optimistic subtitle: “An
efficient solution for the problem of large phylogeny estimation,” but then
opens with a tone of despair:
Optimality criterion-based phylogeny inference is a
notoriously difficult endeavor because the number of solutions increases
explosively with the number of taxa. Indeed, the total number of possible
unrooted, bifurcating tree topologies among T-terminal taxa ... [corresponds] to
nearly 32 billion different trees for 14 taxa and 3 X 1084 trees (i.e., more
than the number of atoms in the known universe) for 55 taxa. ... As most
mathematicians expect that no such algorithm [i.e., polynomial time solution]
exists, one is forced to admit that no future civilization will ever build a
computer capable of solving the problem while guaranteeing that the optimal
solution has been found.
Instead of number-crunching the impossible, the authors propose a heuristic
approach. Heuristic approaches sacrifice the goal of getting an optimal tree in
hopes of getting one faster that has maximum likelihood (ML).
Um, no... I would think that a taxpayer-funded NASA "scientist" like David Coppedge would at least try to understand the material he is bashing, but I guess that is too much to ask of a creationist. Even a NASA employee. The whole point of a heuristic search is still to find the optimal tree, but to do so without having to examine every single possible arrangement of taxa.
Can a 'computer geek' like Coppedge really not know this?
As this applies to phylogeny searches, look at it this way (A spectacularly simplified example) - say you have DNA sequences from 4 taxa and you want to do a phylogenetic analysis. Their sequences are:
1. AGGGCCCCAAAATTTTT
2. AGGGCCTCAAAATTTTT
3. AGGCCCTCAAATTTTTT
4. AGGCCTAGAAGTTTAAA
We can note that sequences 1 and 2 differ by only one substitution; 1 and 3 by two; 1 and 4 by 8; etc., and we know that taxon 4 is not closely related to the other three. If we choose to root our tree (that is, if we recognize that one of the taxa in our analysis will be the most distantly related to all the others), there are 15 possible arrangements that could be produced (we can quickly see which taxon this is going to be, but we would want to test each one to be sure). Now, we could simply draw out all 15 possible arrangements, count up the number of nucleotide changes needed to 'describe' each tree, then pick the tree with the fewest changes as our optimal tree. But that would take, if we use 5 minutes drawing and calculating each arrangement for our example, 75 minutes. Now, if we employ a heuristic search algorithem, we do not draw and calculate every possible tree, because we can immediately discard 'bad' trees/arrangements without having to draw and calculate them.
For example, a quick comparison of each pair of sequences tells us what I wrote above ("We can note that sequences 1 and 2 differ by only one substitution...") and we can use this information, which can be generated in much less than 5 minutes, to see which groupings are the most likely. In our little example here, we can see that taxa 1 and 2 are the closest, with taxon 3 joining the group of 1 and 2, with taxon 4 as out outgroup. So we know that taxon 4 will be the outgroup, therefore, we only have to draw and calculate how taxa 1, 2, and 3 should group. And as we established that taxa 1 and 2 are the closest, there is only one tree that will do. This can be calculated in maybe 10 minutes, if we do it by hand, and use the "heuristic" I just described, which is essentially what computer analysis programs do (albeit, obviously, with much more rigor).
Coppedge is just trying to use the classic creationist "argument via big numbers" and relying on the fact that most of his acolytic readers will not know any different.
Or care.
Further, maximum likelihood is a
search criterion,
not an intrinsic value.
Coppedge is clueless. Perhaps for real.
Lemmon and Milinkovitch wrote a computer program that converges quicker on an ML model with larger number of taxa. They call theirs the “metapopulation genetic
algorithm.” It is a quasi-Darwinian model that tries to optimize trees based on
mutations and selection, and it can incorporate rate heterogeneity estimates
into the model. The authors try their program on real and imaginary populations
and compare their results with other heuristic methods.
The article in question
came out in 2002. Coppedge writes as if such an algorithem is a new thing. To put this in perspective, I entered graduate school in 1999, and I started out using a computer analysis package that utilized maximum likelihood methods (as well as about 4 others) that was written nearly 10 years earlier. Below is Coppedge's snarky disinformative commentary:
Were you ever told in biology class that generating a phylogenetic tree from the
raw data was mathematically impossible, and that no future civilization would
ever overcome this barrier? Probably not, yet textbooks are replete with neat,
authoritative-looking phylogenetic trees. So how do they determine them? By
heuristic methods, which by translation, means guesswork, inference,
trial-and-error, hunches and hope. Their model incorporates a number of
optimization parameters, such as rate heterogeneity, which means that not all
genes mutate at the same rate, and branch length, the presumed evolutionary
distance between taxa.
Wow... Where to begin? No, Coppedge, heuristic methods do NOT at all mean "guesswork" etc. And for a NASA employee to denigrate trial and error??!?!?? Is he for real? One of my interests is spaceflight, and I have seen many programs about the space program, and NASA's history is little BUT trial and error! And the shock of a program incorporating rate heterogeneity! Imagine, incorporating what the evidence indicates! I have little doubt that one can find on Coppedge's site some snarky little essays decrying any notion that all genes mutate at the same rate. And as far as branch length goes - the programs FIND the branch length as part of the analysis! How can it incorporate something that is one of the outputs of the algorithem!
This guy is CLUELESS!
The tweak space is enormous, and they already have a mental picture of what they
want, so this whole approach is based on circular reasoning. If the program
outputs a tree that agrees with the evolutionary assumptions, is scores high;
otherwise, it is rejected.
Amazing. Coppedge - who should at least have a basic understanding of the terminology (since things like "heuristic searches" are common computer science terms) , is simply spouting nonsense. I challenge anyone here to read the
actual paper and point out where the authors state that they already 'have a mental picture of what they want' and somehow force the program to spit out this desired result. The entire purpose of the paper in question was to outline the authors' procedure for producing phylogenetic trees using large datasets with many, many taxa more quickly than can be done with existing methods. Coppedge acts as if they are merely trying to find a way to make a program that produces what they want it to (he must be confusing real science with baraminology).
Of course, what they set out to do - to make a computer analysis program that can analyze large numbers of taxa in large datasets more quickly and just as accurately as existing but slower models all but requires them to 'know' what they want to get - that is how you test your model! Does Coppedgfe REALLY not understand this?
Does this provide any confidence that evolution is being confirmed empirically?
Is this how scientists in our universities should be spending their time,
playing Darwinian computer games?Instead of explaining how mutation and natural
selection could produce a Monarch butterfly or a finch or a peppered moth in the
first place, scientific papers on evolution seem obsessed with trying to uncover
phylogenetic relationships that are impossible to calculate objectively or
verify independently without begging the question whether common ancestry is
even true.
Is this how a taxpayer-funded NASA scientist should be spending his time -writing disinformation filled drivel to prop up his religious beliefs?
And the coup de grace - Coppedge at his sleight of hand best. We see this silly sentiment in many creationist rants - this whole 'those evos are ASSUMING evolution is true when they do their experimets!' as if that is a bad thing to do.
Do creationists assume creation is true when they do their experiments (I mean, if they did any)?
Do physicists assume that gravity is a constant throughout the universe when they plot spacecraft flight paths?
Of course they do - why, I'd even bet that NASA computer technicians like Copppedge assume certain things about the software writing programs they utilize and the models they make.
I would like to point out that this distortion-laden diatribe was one of the shortest I could find on Coppedge's site on this subject. I can only assume that longer rants have even more distortions and embellishments. But David Coppedge need not be concerned -his like-minded readers will believe everything he writes - he is a YEC AND he works at NASA, after all - and will think his smug insults are all well deserved by those evilutionists.
Sad
