Wesley R. Elsberry
Posts: 4991
Joined: May 2002
|
| Quote (GaryGaulin @ July 22 2013,08:46) | | Quote (Wesley R. Elsberry @ July 21 2013,09:58) | | Can Gary program a system that will itself program a controller for a robot agent in 2D that would yield a *provably* optimal strategy in finding high resource concentrations? |
That's what the Intelligence Design Lab already does, as well as can be expected for such an "intelligence". |
Gary knows the drill. We'll need the file name and line numbers in the PSC VB code where Gary claims this is happening.
Myself, I see no hint that a controller program is generated by the PSC VB code. Maybe Gary's reading comprehension is bad enough that he thinks that a controller program is just the same thing as a controller program generator. But that would be ridiculous, wouldn't it? My paper reported results from the generation of hundreds of thousands, if not millions, of actual programs.
Nor does Gary's PSC VB code show any sign of having most of the other starting conditions I listed, nor any sign of doing what I specified was needed as a result:
| Quote |
Can Gary program a system that will itself program a controller for a robot agent in 2D that would yield a *provably* optimal strategy in finding high resource concentrations?
Starting conditions:
* Resource is distributed as a gradient with a single peak randomly placed in a rectangular discrete bounded 2D grid with prime or relatively prime sides. [Gary sets up discrete, non-gradient resources.]
* The robot agent can sense the difference in concentration of a resource gradient between the current position and one step ahead in the direction of current movement. (Knowledge of the resource is only local.) [Gary's program permits identification of any resource from any position in the space, thus Gary's sensory system implementation is global.]
* The robot agent is restricted in its means of changing heading. It can only initiate a random tumble, as with ciliary or flagellar movement, resulting in a new random heading. [Gary is quite adamant about non-random settings in his PSC VB code.]
* The robot agent is restricted in its movement ability. It can only initiate a single-step movement ahead in the current direction.
* The robot agent controller has a Turing-complete computer language and hardware including three registers and three stacks of 10 numbers. [Gary has nothing of the sort implemented in his code.]
* The initial control program in the simulation includes no heading-change commands and no movement commands. [Gary's code has these things in the controller at the outset.]
End results:
The point of the paper is about the production of effective methods in a particular class.
| Quote |
In testing the capability of evolutionary computation to produce effective methods utilizing movement strategies to intelligently exploit spatially-distributed resources, our results show that such strategies do emerge and that in about 12% of shorter runs and in 80% of longer runs the ?nal movement strategy used by the majority of the population at the end of the run is in the class of optimal response for our environment, that of gradient ascent.
|
If Gary's approach does not generate an effective method, it isn't a replication of the paper. [Gary's PSC VB code does not do this.] If Gary can't prove the optimality of the generated effective method, anything else he does doesn't matter. [Gary does not do this.]
|
Nor does Gary address the various unresolved issues from his laughably inept critique of metadata from the 2009 paper:
| Quote |
| Quote (GaryGaulin @ July 04 2013,15:32) |
Saying "one can research intelligent systems in an evolutionary computation framework" does not help your case either, that's what I have been saying! And you are in fact influencing the molecular behavior of imaginary cells that in turn influences the behavior of your imaginary virtual critter. Apparently when you say that it's science it's science, but when I do (along with even more detail than you can provide) it's religion.
In this paper you are one of 4 authors who only experimented with GA software called the "Avida digital evolution research platform" which was developed by someone else, not by you. I also experimented with Avida, and would be ashamed of myself for thinking that my playing with it was worth writing a paper for. But seeing a sciency looking abstract in a science journal does look highly scientific and very impressive to someone who does not know what it actually is, like Texas Teach and other gullible science teachers.
In looking at the number of citations (by which the success of a science paper is judged) there were none listed. I expect that is because the scientific community found the "research" as much of a yawn as I do.
And the metrics were quite revealing!
Metrics: 17 Total downloads since Feb. 2011
With only 17 downloads someone like myself has to wonder why you would even bother to publish science papers at all, except of course to advance a career that requires smoke and mirrors to make the big-bucks, and control others who don't know your trick. |
Gary does not even attempt to critique the content of the paper. Gary instead attempts, and fails, to critique meta-data about the paper.
For instance, here's the author list:
Author(s)
Elsberry, W.R.
Michigan State Univ., East Lansing, MI
Grabowski, L.M. ; Ofria, C. ; Pennock, R.T.
"Ofria, C." is Charles Ofria.
Wikipedia:
| Quote |
Avida is under active development by Charles Ofria's Digital Evolution Lab at Michigan State University and was originally designed by Ofria, Chris Adami and C. Titus Brown at Caltech in 1993.
|
If there is any one person who could be said to have written Avida, Prof. Ofria would have the best claim to that. Nor is the paper about use of unaltered Avida, as any competent reader looking at the Methods section would have learned. Most of the coding for the additional instructions and integration of them into Avida was mine.
The IEEE site apparently doesn't do meta-data well. Google Scholar knows of three citations of the paper; not stunning, but not non-existent, either. And the number of downloads at the IEEE site is a count of people who paid either $31 or $13 for the privilege. I don't know how many people downloaded the PDF for free from Charles Ofria's website instead. At a rough valuation, then, the IEEE has realized between $221 and $527 from sales of my paper. Gary said that he had seen the paper before; I wonder whether he paid the IEEE or downloaded it for free from another source. If he didn't buy my paper from the IEEE, then Gary knew that it was not only available from IEEE and therefore whatever download count they had was meaningless as a measure of community interest, and would have made his claim with intentional malice. If Gary did buy the paper from the IEEE, then he knew that the download figure had a real cost associated with it and was not a simple measure of unhindered community interest as his statement implies; again, it is difficult to see how one could posit Gary making that statement without actual malice. Gary has previously claimed to have superior habits in paying attention to detail. Here's a detail from the IEEE "Metrics" tab that gives the download number Gary uses and quotes above; I'll provide it again here:
| Quote |
17
Total downloads since Feb. 2011
|
However, the paper was published in 2009. The IEEE site doesn't have download data for the period of time closest to publication, which is when most interest in papers is expressed. If Gary didn't notice the mismatch between the publication date and the download statistics date, it argues that Gary has sub-standard attention to detail, contrary to his previous claims. If Gary, on the other hand, did notice the mismatch in dates and chose to make the argument seen above, he was deliberately misleading readers as to the truth of the situation. Again, Gary's handling of meta-data shows near-total incompetence or actual malice.
Gary:
| Quote |
Saying "one can research intelligent systems in an evolutionary computation framework" does not help your case either, that's what I have been saying! And you are in fact influencing the molecular behavior of imaginary cells that in turn influences the behavior of your imaginary virtual critter. Apparently when you say that it's science it's science, but when I do (along with even more detail than you can provide) it's religion.
|
Gary has previously stipulated that his PSC VB code contains no evolutionary computation component. Then there is this from Gary:
| Quote |
As a result the Theory Of Intelligent Design is an 'origin of life'¯ theory that requires terminology found primarily in robotics and Artificial Intelligence and never once mentions or borrows from Evolutionary Theory.
|
So the above is once again a blatant falsehood by Gary; Gary has explicitly stated exactly the opposite of what he claims now. Does Gary think that his past words can't be consulted?
The stuff about "molecular behavior" as a component of our paper is a bizarre invention on Gary's part. It seems unlikely that Gary has read the paper; at the least, his strange statements about it indicate that he did not comprehend it even if his eyes were exposed to reflected light from its pages.
|
--------------
"You can't teach an old dogma new tricks." - Dorothy Parker
|
