Richard Dawkins has made his programs from The Blind Watchmaker and Climbing Mount Improbable available via a web browser interface.
Sadly for our friends at Uncommon Descent, the Weasel is not included.
189 thoughts on “Richard Dawkins’ Software”
Leave a Reply
You must be logged in to post a comment.
But it’s still only HTML.
No tests for “the power of cumulative selection” either.
What a blow to Patrick’s empiricism.
Mung:
Poor Mung. Apparently it still stings.
maybe this will help the learning challenged: http://theappleandthefinch.com/2016/05/22/finding-dawkins-weasels/
Mung,
Sour grapes much?
I often wonder why anti-evolutionists still make a fuss about Dawkins’ Weasel program and yet never mention his later biomorphs – or perhaps I don’t.
ETA Good fun playing the environment. Thanks for the link, Patrick.
They don’t understand either, but there is a snappy thing to say about weasel.
Your lack of any empirical demonstration doesn’t sting at all.
What empirical test do you propose?
From your link:
LoL! Talk about rewriting history. Doh!
Just an observation. Have you seen mention of biomorphs by ID proponents?
I’ll mention them now. They are selected for by an intelligent agent. And their selection has nothing to do with whether or not they actually function in any environment. Nor is their selection based on their reproductive advantage.
Empirical Tests. You either have them, or you don’t.
Way to missssssssssssssssss the point.
And nice to admit you are actually an ID supporter at last Mung. Tell me, what’s the particular reason you support ID?
Sorry, Rich, but that’s a poor explanation of what Dawkins was getting at with his monkey/Shakespeare model of cumulative selection. The root problem, I think, is that we — definitely I — lost touch with The Blind Watchmaker as we went back and forth, year after year after year, in online arguments about the so-called Weasel program. Rereading the book, back in December, I was appalled to see how much I had been missing. (I went on to read a lot of other stuff related to ID, some of it for the first time, and suffered the embarrassing realization that I’d had no gestalt comprehension of the ID movement. What I ended up believing was important to say was uncannily like what Wesley Elsberry emphasized shortly thereafter, in his “Open Letter to the United Methodist Church on ‘Intelligent Design’.”) Dawkins explains in the preface what he regards to be the conceptual obstacles to understanding evolution. The first half or so of the book is clearly designed to help readers overcome those obstacles.
The Weasel program is an intermediate step from Hoyle’s “tornado in a junkyard” fallacy to biomorphs. (It serves other pedagogical purposes. The notion that an abstract mechanism could be realized by a computer program was much harder to get across in 1986 than it is today. Skipping the Weasel, and moving directly into Biomorphs Land, would have been too much for most readers.) That is, Dawkins eliminates only one of two fundamental errors. He leaves in place the erroneous assumption of a prespecified target, and illustrates the consequence of eliminating the erroneous assumption that the target must be hit in a single step. He immediately emphasizes that cumulative natural selection has no target.
Dawkins draws several analogies in his presentation of the Weasel, consistently using scare quotes when he does. He refrains from analogizing the characters in phrases to DNA/genes, and for very good reason: the “computer monkey” selects on the basis of what it sees, and copies what it sees (with occasional errors). To say that the “program used ‘critters’ whose DNA were strings of characters” is to do violence to Dawkins’s presentation. The morphogenesis (brilliantly done) in Biomorphs Land contrasts with the complete absence of a genes-to-critter transformation in the Weasel program.
By the way, to appreciate fully what Dawkins was doing with his references to clouds, you need to know about Fred Hoyle’s science fiction novel, The Black Cloud.
Mostly because of the sort of people it seems to upset the most.
Which is to say, ID is SOL without arguments from improbability. The silence of the leading ID activists on biomorphs is deafening. It’s patently clear that they decided in the early Nineties, with law professor Phillip Johnson very much in charge, to make much of targets. That was essentially a commitment to misrepresentation of the role of the Weasel in Dawkins’s presentation, and suppression of references to Biomorphs Land.
The first reference to biomorphs at Evolution News and Views came last year. It was due to Paul Nelson (whom I’ve seen argue whether Adam had a navel — can’t recall which side he took). Then BA77, bless his little heart, took to quoting Nelson in comments at UD. When last I googled, there had never been a reference to biomorphs in a post at UD.
Put simply, the Weasel program eliminates only one of two errors in Hoyle’s fallacy, and thus leaves activists in the ID sociopolitical movement something to exploit. The only appropriate characterization I can find of how the leading activists have exploited it is intellectual psychopathy.
So what do you make of the evidence in “Parameter Dependence in Cumulative Selection,” by Christian apologist David Glass? (By the way, I think well of Glass, though not of his paper.)
Here’s your chance. Assess a highly readable research paper addressing a topic that you have addressed many times, representing yourself as competent. Glass’s Weasel-on-steroids model is just what you’ve been demanding. It features recombination, as well as fitness-proportionate selection of parents. Let’s see you commit unambiguously to a position on this research.
As much as Glass’s paper is up what you claim to be your alley, as much as it does what you have said should be done, and as much as the results comport with your expressed beliefs, you should devote an opening post to it. I insist. Put up, or shut up. I’ll respond with a post of my own.
P.S.–Let’s make this a bit more appealing to Mung. I am already working on nontechnical response to Glass’s paper. So I’ll promise up front not to respond to Mung’s post. We’ll simply have a (stark) contrast of our responses to the paper. I’ll request also that our posts be featured for a while. It’s fine by me if Mung gets top billing.
That’s funny. Mung was active in discussions of the Weasel here recently (for example here, here, and here) and never seems to have raised the issue of empirical evidence for the power of cumulative selection. Back then Mung was concerned whether genetic drift would make natural selection ineffective. So Mung did think then that these simulations did say something about the effectiveness of natural selection.
Well, I guess people change their mind.
I’m sure you have better things to do than read every post of mine here at TSZ.
But if you really don’t have anything better to do you could start here.
And I’ve brought it up since then as well. keiths has no tests. We’re supposed to take his word. I lack the faith for that.
For Joe:
If stuff like this your raison d’être, then you’d be better off vesting faith in raisins.
Is this “empirical tests” stuff something carried over from another thread? Because I’m still not sure what it is Mung is demanding be tested empirically.
Yeah, he wants some sort of unit/functional tests handed to him to “falsify”/verify his hilarious algos. Pretty amusing
The claim was that the Dawkins WEASEL program or it’s latest iteration here at TSZ as the keiths WEASEL program, “demonstrates the power of cumulative selection.”
Yet no one seems to be able to devise any unit or functional tests that demonstrate that the program actually does as advertised. Instead it’s just snark and mocking and ridicule. The usual fare.
So from where I sit, there’s no objective empirical evidence that has been offered to support the claim. And what is offered without evidence can be dismissed … blah… blah…blah. (Said in my most Patrick-like voice.)
Devise!? I’ve tested various implementations, and also some theory, against the theoretical predictions at DiEbLog. Do a hundred thousand runs per combination of mutation rate and population size, and the empirical results should be very close to the predicted results.
(When I can put together modules from a program that is not a Weasel to get a Weasel, I do it. That was PART of the testing of my implementation of the Glass algorithm.)
So, Mung, have you read the Glass paper? It’s not only simple, but brief. Doesn’t he provide empirical evidence that cumulative selection works well only with fine-tuning of parameters?
ETA: I should have told you that his results are correct.
Mung,
I’m honestly trying to detach you from a silly, silly belief that you would be embarrassed to hold if you had an inkling of how silly it is. I’ve put you on the spot, but it’s ultimately not meanness. And it’s not all about you. I reckon that whatever persuades you persuades a lot of other people who bought into Dembski. My own family is much on my mind.
Are you aware that Stephen Meyer and Douglas Axe not only have used scare quotes when referring to evolutionary “search,” but also have explained why? The ID movement will be moving away from Dembski, whether you like it or not. He’d reached the point where he would have made a very poor expert witness in court. If nothing else, the LCI switcheroo — going from specified complexity to active information (loosely the opposite) without ever commenting on the change — made him a liability.
Put the pieces together. How long could Dembski keep asserting that “Darwinian evolution is inherently teleological,” when Meyer and Axe insisted otherwise? I’m not claiming to know why Dembski retired. But I am telling you that Dembski was wrong on this particular point, and that Meyer and Axe are right. And disagreement on such a huge point was not a good thing for the DI.
(Marks is 65, by the way — no idea how long he’ll stay in the saddle. I doubt that the DI’s summer camp will ever again feed him grad students. When Ewert isn’t actively annoying the hell out of me, I recognize what a sad case he is.)
Rumraket:
dazz:
Poor Mung. His knowledge of software testing is as limited as his knowledge of genetic algorithms.
Funny really. And quite representative of the whole ID mindset.
Nobody has done what I’m asking!
Yes they have!
Crickets…..
Meanwhile on a different forum:
Nobody has done what I’m asking!
If Mung mentioned “empirical tests” in his own thread on the Weasel, I should have noted it. But I looked at the one he linked to, and it starts out by admitting the “power of cumulative selection” and then arguing about where the power comes from. So it is precisely *not* asking for tests to verify that power. It is admitting the “power of cumulative selection”. But now Mung seems to be calling for such tests and not to be willing to admit the power anymore, absent those tests.
The “empirical tests” Mung asks for are, if I understand, tests of the Weasel algorithm (not real-world tests using organisms).
In my thread on “Wright, Fisher, and the Weasel” I showed some theory for a Weasel with $N = 1$ survivors, and a selection coefficient $s$. By taking $s$ to infinity, you end up with a classical Dawkins weasel.
From the formulas I have given there, you can then see that cumulative selection does hugely better than random wandering unguided by selection. Furthermore it is theory, which is equivalent to infinite numbers of runs of the program.
So exactly what are the empirical tests supposed to test for?
Sorry, correction. The theory I posted approaches a classical Dawkins weasel that has an infinite number of offspring, as $s$ gets large. It does not say what happens with a finite number $M$ of offspring.
But in any case we don’t know what Mung wants us to test.
Is this an admission that you cannot write a test for “the power of cumulative selection?”
Why not just say so?
The WEASEL program is software. The claim is that it demonstrates the power of cumulative selection. When writing software we write tests to demonstrate that the software functions as expected and that it meets the requirements.
I want someone to tell me how to write software tests (or provide the tests that they wrote) to validate that the WEASEL software actually does demonstrate the power of cumulative selection.
Something empirical. Something more than handwaving. Else I have no reason to believe the claim.
Is there some dispute over my claim that WEASEL is supposed to demonstrate the power of cumulative selection?
D:\projects\weasel\test>ruby test_weasel.rb
Run options: –seed 37000
# Running:
F
Finished in 0.001503s, 665.3360 runs/s, 665.3360 assertions/s.
1) Failure:
TestWeasel#test_houston_we_have_a_weasel [test_weasel.rb:6]:
Houston, it’s not a Weasel!
1 runs, 1 assertions, 1 failures, 0 errors, 0 skips
That is correct. We’re talking about software and automated software tests. Have you ever written a unit test? How do you test your software programs?
I asked for tests in the original thread. I provided examples and links above (@22).
Who or what provides the guidance?
Are you saying the WEASEL program uses guided evolution? Are you trying to give people here a heart attack?
You need to take a long break from the Internet, and try to regain perspective.
“power of cumulative selection” — 43 hits in Google Scholar, at least half of them related to evolutionary computation
“Joe Felsenstein” — 1070 hits in Google Scholar
“power of cumulative selection” — 2200 Google hits
“Joe Felsenstein” — 12,400 Google hits
What is this silly, silly, belief that you think I hold?
So Tom, do you also put the words “sort” in scare quotes when writing about sort algorithms, because, you know, it’s not the algorithm or the computer that is actually doing the sorting it is humans who are doing the sorting?
A sorting process (running program) can output justifications for all the steps it takes. A black-box “search” process can justify only its termination. The fundamental reason is that only inductive inferences about unobserved values of the fitness function are possible on the basis of observed values. There is no justification for inductive inference. Induction requires bias.
The silliest thing coming from Dembski and Marks is the notion that intelligence must create information to supply the bias. All that’s required is to do something very simple that a black-box “search” does not: look inside the “box” (process the problem description).
Read my entire comment.
Are you referring to this?
Are you referring to this?
Because I am talking about algorithms and software programs and you’re talking about biological evolution (or so it seems).
Do you think I hold the opinion that Darwinian evolution is a search for something?
Do you think I hold the opinion that Darwinian evolution is inherently teleological?
Out with it Tom. Let’s get this “sorted” out. Because, you know, natural selection doesn’t really sort things either.
Still with this nonsense Mung?
What’s the point of providing an automated test for an algo you don’t even understand? If you don’t understand the specification (Joe’s Wright-Fisher model) and how the algo implements it, how would you know if a functional test actually tests whether the algo implements the specification correctly? Let me guess, you would demand a test for the test right?
And why are you so obsessed with automated tests? It’s not like we’re in a continuous integration environment here. All you need to do to test the program is to check if it produces the results that the model predicts. Remember when some folks here plotted the long term average fitness for each value of the selection coefficient and compared that to the predicted value in the model? Yes, that’s a functional test any way you look at it.
Now that we know the algo implements the model accurately, how do we know if it demonstrates the “power of cumulative selection”? Well, you just run the algo with s=0 (no selection, pure randomness) and compare the resulting fitness to other runs using increasing values of s. Simple as that. Even you should be able to notice that the fitness values increase with s
“evolutionary search” About 93,100 results (0.30 seconds)
A few more than “Joe Felsenstein” returns.
What’s up with that?
Either you have tests or you don’t.
Automated Tests as Documentation
I posted the test, why are you ignoring it?
“god doesn’t exist” About 61,700,000 results (0.47 seconds)
“god exists” About 12,200,000 results (0.49 seconds)
Check mate
Mung:
dazz:
Because he’s desperate. All of his attacks on Weasel and the drift-Weasel have failed. His own version of Weasel was an embarrassment, demonstrating nothing but his incomprehension of the requirements.
Since he can’t find a flaw in the Weasels, he’s reduced to pretending that the only legitimate tests are automated tests.
Competent programmers know better.
Mung April 22, 2016 at 7:34 am
Deny that evolution is a search algorithm.
Deny that a search algorithm is designed to solve problems.
Deny that evolution has found by means of search the solution to any problem.
Deny that the size of sequence space matters.
Assert that programs using evolutionary algorithms have refuted the central claims of ID.
Deny that you wrote this.
HT: RichardTHughes.