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.
I followed your link to this statement:
“OK, here’s the proposed “test” of whether Weasel programs do better than random search.”
My question is, where did I ever claim that Weasel programs do no better than random search or offer to wager that Weasel programs do no better than random search?
I responded here:
Your response:
You want to wager over something that was never in dispute?
Why not offer a meaningful wager?
Joe F, what odds are you willing to offer?
Richardthughes, what odds are you willing to offer?
The buy in is $10,000.00
Put up or shut up.
Bump
Bump
Mungs metamorphosis into Joe G continues. I’ll take Joe F’s bet at $100. I’m but a humble cashier, more money won’t make me more right.
Wow the last two got borked
Mungs metamorphosis into Joe G continues. I’ll take Joe F’s bet at $100. I’m but a humble cashier, more money won’t make me more right.
poor dickie
Still dodging the bet? Look at your original claims.
Please don’t Guano any of Mung’s posts. I’d like them nice and visible. Thanks.
You’re wiling to put up $10,000.00 or you aren’t.
Mung,
Your willing to take the original bet or you aren’t
10 grand dickie. Or bow out. Money where your mouth is.
Original bet Mung. Or bow out. money where your mouth is.
You have put up nothing. You’re willing to put up or not.The buy in is $10,00.00. This is how we separate the players from the pretenders. You’re a pretender. Joe F. is a pretender.
$10,000.00
You have it or you don’t. You’re willing to risk it or you’re not willing to risk it.
My bet offered even odds that a Weasel program, mutually agreed upon by Mung and me, having 10 offspring per parent, one adult, and mutation rate 0.04 per character would reach the target sequence before 10,000 generations. That was clear in my comment.
Wager would be $5 per run, and we would do 20 runs in all. The runs would differ only in the past number generator seeds.
Joe Felsenstein,
Mung will do anything but take the bet.
Joe F, the wager I am offering is for $10,000.00. What odds are you offering? You’re offering even odds for what? Why should I accept the odds you are offering?
original bet Mung. Or bow out. money where your mouth is.
Ten Thousand Dollars.
Richardthughes cowers.
The bet is for $10,000.00. Put up or shut up.
A nice public Mung Meltdown.
keiths can’t meet the bet. $10,000.00. It separates the cowards and pretenders.
Ten Grand.
Ten Thousand Dollars.
$10,000.00
Put up or shut up.
So Mung won’t take a 100 dollar bet but will take a 10,000 dollar bet for the exact same thing?
Hmm, I guess we’ll have to find a genius to work out what Mung’s motivations for this behaviour are….
Or not!
What is your bet, precisely? I see Joe has detailed his bet quite clearly. Can’t imagine anyone putting up ten grand without you making clear what you are betting for or against, who would adjudicate, hold the stake etc.
BTW, dollar signs open LaTeX which why some comments look odd.
Well, that’s the problem. It’s difficult to be precise in 10 word drive by comments.
As far as I can tell Mung is willing to take Joe’s bet but 10,000 instead of 100. Mung, that about right?
If so, I’ll happily chip in some ££.
I’m happy to hold everyone’s stake. 😉
I’m in for £100 then, pending clarification.
Mung has not said what he is proposing to bet on. I have said very clearly what I am proposing to bet on.
And Mung also then said (here)
My original offer of a wager still stands.
Sorry, but Lizzie has rules and I have to report myself to her when I don’t enforce them.
It turns out that Mung’s offer of a bet was exactly as expected, a bait and switch.
It’s funny how these people never want to go on record one way or another.
Ah, I see. When Mung was calling repeatedly for a test suite, it was because Mung was unhappy with people for being critical of natural selection. Which Mung calls “intelligence”.
As in Darwin’s book “The Origin of Species by Means of Intelligence”.
Mung, it turns out, is a big supporter of the Weasel program.
A disappointing amount of weaselling.
Essentially, any digital evolutionary process operates in a space comprised of all possible strings, given parameters of maximum length and alphabet. A subset of possible strings from that space forms a population, which is mutated and bred from. Because of finite resource, the fraction of space occupied by the current population does not continue to grow – ie as well as birth, there is death.
If the question is: “does this demonstrate the power of cumulative selection?”, the answer at this stage must be “no”. There is no selection. Or at least, none that makes any difference. All genotypes are of equal fitness, the process can only drift. But the selective differential between genotypes still takes a value. That value happens to be zero.
So, how to implement a differential between genotypes? It appears, looking at the rearguard action conducted by Mung, to be impossible. One cannot demonstrate ‘the power of selection’, cumulative or otherwise, because any process that differentiates between genotypes – that sets a nonzero selective differential – ‘guides’ the process away from the space occupied by the less fit towards the space occupied by the more fit. Selection is, in many ways, indistinguishable from guiding. This is Blind Watchmaker‘s central theme, so it’s hardly a major ‘win’.
In the case of Weasel, a single member of the space is taken as a reference, and arithmetic distance from this can be computed. One could take any subset of the space, of any size and do something equivalent. The bigger this subset, the less like a search it becomes. But there is no need to measure fitness by reference to any subset of the space. It’s a feature of Weasel, rather than a feature of GAs, and its use in Weasel largely a rhetorical device, to key into the ‘infinite monkeys’ idea that precedes it. All the phrase actually does is provide a differential between existent genotypes at a moment in the population’s history. There are many ways one could introduce a differential between genotypes without reference to any existent genotype. It’s the reddest of herrings to see GA as universally a search, or Weasel as ‘guided’ because its fitness differential happens to be determined by reference to a single genotype which it ultimately hits.