Adaptive Robots: Yet More Evidence for Evolution

BA: Thank you for the link about the Abel paper. It seems extremely interesting (like everything from Abel). For those interested in the full paper, here is the link: http://www.bentham.org/open/tocsj/articles/V004/14TOCSJ.pdf (BA, how can you be so efficient in finding all that stuff? :) ) gpuccio

Constraints vs. Controls - Dr David L. Abel Excerpt: Inanimacy is blind to and does not pursue utility. Constraints produce no integrative or organizational effects. Only the purposeful choice of constraints (e.g., an experimenter choosing the initial conditions of an experiment), not the constraints themselves, can generate bona fide controls. When constraints are deliberately chosen to steer physicodynamic causal chains towards pragmatic pursuits, however, those constraints then become legitimate controls at the moment of their purposeful selection. The end result of utility tends to confuse many investigators into thinking that the constraints themselves caused the formal function. They forget the essential role that choice contingency played in steering events and optimizing algorithms. Classic examples of the above confusion are found in the faulty-inference conclusions drawn from many so-called “directed evolution,” “evolutionary algorithm,” and computer-programmed “computational evolutionary” experimentation. All of this research is a form of artificial selection, not natural selection. Choice for potential function at decision nodes, prior to the realization of that function, is always artificial, never natural. http://www.scitopics.com/Constraints_vs_Controls.html bornagain77

As in, islands of function -- an image I believe I owe to you GP. kairosfocus

Mung: Yes, I am curious too to know why you think that it is an "utter misrepresentation". Well, maybe Beethoven is not necessarily the end point, nit I believe you probably mean something different. I think SCheesman has defined very well point 4). It is a very essential point, often misunderstood or utterly ignored by darwinists. I often refer to the same point as the necessity, for darwinism to work, to be able to "deconstruct" all complex functions into a succession of simpler, gradual ones. I have many reasons to believe that's impossible. To state it in SCheesman's brilliant words, I do believe that no "continuity of functional solutions in the search space" can be found in biology. Nor, as for that, in any truly complex functional space. And you, what do you think about that? gpuccio

Mung. "Claim", as used by Alex73 above, is likely the wrong word. "Implication" is probably more accurate. I expect that you would agree that in Darwinian evolution all the genetic changes necessary to take us from the last universal common ancestor to ourselves necessarily occur as the result of purely naturalistic processes (i.e. no teleology added). Moreover, that the vast majority of changes that actually occurred are those which stood a reasonable chance of occurring (for instance, I expect you don't believe that entire proteins popped into existence in a single generation). This limits "progressive" genetic mutation (that is the kind of change that actually introduces something truly new, as opposed to merely shuffling about existing information, like horizontal gene transfer, or sexual re-combination) to those changes which are actually likely to arise, from a biochemical, statistical standpoint. If evolution is/was inevitable, then it/was inevitable for good, biochemical and statistical reasons. There is only one more link in this chain, and that is that the probability of a single "creative" mutational event occurring decreases extremely rapidly with the amount of variation contained therein, so small changes are far more likely to succeed than large ones (consider the difference in the odds of a non-destructive single-point mutation vs. a non-destructive double-point mutation). You are, after all, altering reproduction code at each generation. If what I have just offered is a misrepresentation of the situation, please feel free to enlighten us with how "large steps" can occur. Maybe start off by defining your own terms. What do YOU mean by "small steps" and "large steps" in reference to single-generation changes in genetic code. Simple contradiction is not an argument. SCheesman

Indeed, it is an essential factor and a key claim of Darwinism, ie that you can get from bacteria to Beethoven by taking small steps at a time.

This is false and an utter mispresentation of Darwinism. Mung

SCheesman, Thanks for point 4. Indeed, it is an essential factor and a key claim of Darwinism, ie that you can get from bacteria to Beethoven by taking small steps at a time. Alex73

Alex73: If I might add one more to your list of three factors: 1. The size of the search space 2. The steepness of the fitness function 3. The efficiency of selection ... 4. The continuity of functional solutions in the search space. It is one of the primary assertions of ID that you can't always get from "A" to "B" in small steps. SCheesman

Ha! I like it! Clever engineers pulling the wool over the eyes of biologists! Good job guys! Of course, the one can simulate the effect of the wind on a building or a car if one builds a scaled down model and uses the appropriate liquids, speeds etc, then scales the results again with the appropriate numbers. The main problem with this paper is that there is no realistic relationship whatsoever between the model and nature in the following important aspects: 1. The size of the search space 2. The steepness of the fitness function 3. The efficiency of selection The experiment was successful, because these parameters were chosen using engineering insight so that the search would be successful. This is everyday practice in engineering when the optimum parameters cannot be found by calculations only. Similar searches are also done by genetical engineers when they try to modify the function of an enzyme. Of course, that sort of research was sold to my kids as "evolution used for our benefits". Again, the missing bit is to prove that nature has sufficient probabilistic resources and the selection between the immediate steps is strong enough to make the model valid for evolution. This is a nice demo of engineering and I am sure that the researchers had a lot of fun time, just like Dr Nefario in Despicable Me. (I liked his robots, by the way.) Alex73

Mathgirl: Yes, indeed. That animation just about sums it up. The mutations operate at the level of code, and the effects must be beneficial at the level of the environment. That is exactly what I meant in my comments above. The problem is, the larger and more sophisticated the programme is, the more likely it is that random mutations in the underlying programmatic units will render it inoperable (just like in real life with DNA). Granted, those mutations that are not actually harmful may possibly result in small-scale optimization. They may be able to tweak existing functionality enough to move around the local functionality "solution space" and locate its maximum fitness. In effect, it is able to traverse all possible functional solutions that are separated from each other by a single, or possibly two point mutations, and given enough time it can sample them all, just as water will eventually find its way into every crack. But to create new functionality requires multiple, simultaneous changes on several layers in any programme, (as any programmer will tell you) and that is why you will never get to the flying robot you see at the end of the animation you pointed to. There is no step-wise continuous path, consisting of single or even double-mutational steps that will take you there. And that is the true limitation in neo-Darwinian evolution - the landscape of solutions, at the code level is not continuous, but stars of isolated functionality in a galaxy of mostly empty function space. The flying robot needs wings, engines, and all the other parts necessary for flight, each of which are sophisticated systems on their own. It is possible, with some ingenuity, to turn one word into another by changing one letter at a time. It is much harder to change one sentence into another with a substantially different meaning, one letter at a time, with all the intermediates still making perfect sense. I would submit that it is essentially impossible to do the same with a paragraph, let alone chapters or a full novel. SCheesman

SCheesman,

A true test would be to start with Robots with a rudimentary ability to navigate throught the maze. Now, randomly mutate and recompile the naviagational source code, selecting, of course, only those copies which actually compile, and then select robots, using the new code, that have improved navigational ability.

You mean like this? (Hey, gpuccio, I'm still scrounging the time to respond in the other thread.) MathGrrl

A true test would be to start with Robots with a rudimentary ability to navigate throught the maze. Now, randomly mutate and recompile the naviagational source code, selecting, of course, only those copies which actually compile, and then select robots, using the new code, that have improved navigational ability. SCheesman

Well, another example of what intelligent selection (=design) can do. From the paper: "Experimental selection was conducted in three independent populations each consisting of 80 individuals [18]. The performance of each robot was evaluated with a fitness function describing the ability of the robot to efficiently move in the maze. Over the first few generations, the robots rapidly improved their ability to move without collisions in the looping maze and, within less than 100 generations, most of them exhibited collision-free navigation." Think, they selected for collision free navigation and they obtained collision free navigation! So now we know that RV + Intelligent Selection can find simple functional combinations. Amazing... gpuccio

quotes from Marks in Pinball Wizard: * From an interview with Robert Crowther of the pro-intelligent design Discovery Institute (March 03, 2010). "Darwin as the Pinball Wizard: Talking Probability with Robert Marks,". Retrieved on 2010-05-03. * [Computer] programs to demonstrate Darwinian evolution are akin to a pinball machine. The steel ball bounces around differently every time but eventually falls down the little hole behind the flippers. * It's a lot easier to play pinball than it is to make a pinball machine. * Computer programs, including all of the models of Darwinian evolution of which I am aware, perform the way their programmers intended. Doing so requires the programmer infuse information about the program's goal. You can't write a good program without [doing so]. * Your chances of winning the lottery are about the same whether or not you buy a ticket. It's better ... if you give your money to me and I'll decide whether or not to give it back. * From the viewpoint of computer simulation, our universe does not contain the probabilistic resources to get a meaningful result for even a moderately sized unassisted [Darwinian] search. In fact, if you take ten to the one thousand of our universes in what is sometimes referred to as the multiverse, the probabilistic resources don't exist there either. * Let's abandon labels and pursue the truth no matter where it leads. Don't entrench yourself in a paradigm and claim a corner on truth. Many who have done so in history have been shown to be foolish. http://en.wikiquote.org/wiki/Robert_J._Marks_II further note: The Genius Behind the Ingenious - Evolutionary Computing Excerpt: The field dedicated to this undertaking is known as evolutionary computing, and the results are not altogether encouraging for evolutionary biology. http://biologicinstitute.org/2008/10/17/the-genius-behind-the-ingenious/ Signature In The Cell - Review Excerpt: There is absolutely nothing surprising about the results of these (evolutionary) algorithms. The computer is programmed from the outset to converge on the solution. The programmer designed to do that. What would be surprising is if the program didn't converge on the solution. That would reflect badly on the skill of the programmer. Everything interesting in the output of the program came as a result of the programmer's skill-the information input. There are no mysterious outputs. Software Engineer - quoted to Stephen Meyer http://www.scribd.com/full/29346507?access_key=key-1ysrgwzxhb18zn6dtju0 bornagain77

Thanks Dr. Hunter, In computer science we recognize the algorithmic principle described by Darwin - the linear accumulation of small changes through random variation - as hill climbing, more specifically random mutation hill climbing. However, we also recognize that hill climbing is the simplest possible form of optimization and is known to work well only on a limited class of problems. Watson R.A. - 2006 - Compositional Evolution - MIT Press - Pg. 272 In the following podcast, Robert Marks gives a very informative talk as to the strict limits we can expect from any evolutionary computer program (evolutionary algorithm): Darwin as the Pinball Wizard: Talking Probability with Robert Marks - podcast http://www.idthefuture.com/2010/03/darwin_as_the_pinball_wizard_t.html bornagain77