Talk:Genetic algorithm
This article has major flaws. It does not distinguish between genetic algorithms, which model only mutation, recombination and selection, and are used to search for answers to specific problems; and evolutionary simulations, which often model the reproductive process as well, and often have survival as their only criterion for success. The criticisms listed often apply to only one of these, yet the accompanying discussion covers both. The article also confuses aspects of the simulated environment with aspects of the 'organisms' that inhabit it.
"Genetic algorithms are a search method used in computer programming in the search for solutions to combinatorial optimization problems (http://en.wikipedia.org/wiki/Combinatorial_optimization#Informal_definition) using methods inspired by biological evolutionary. Because of they were inspired by Evolution theory, some evolutionists claim them as evidence that microbe to man Evolution is possible."
This is a clear example: the first sentence describes genetic algorithms, while the second covers all evolution based systems. This is misleading.
"Genetic algorithms often start with a random "gene" sets."
False. They almost never start with random initial state. The few that do are mostly evolutionary simulations written to investigate a specific aspect of selection.
"Genetic algorithms have no fatal steps."
False. Mutations in evolutionary simulators and genetic algorithms can be as fatal as mutations in living creatures.
"Genetic algorithms have the instructions for critical functions such as reproduction beyond the influence of the mutations, as such no mutation will disrupt those functions. In the real world critical functions can often be destroyed by mutations."
False for evolutionary simulators.
"Genetic algorithms never produce new capabilities beyond what is preprogrammed into them."
False. As an example, an experiment intended to generate a specific type of signal produced a radio receiver instead.
"Genetic algorithms start with fully functional processes designed into it. "
False. The simulated environment may include some processes, but the 'organisms' within it need not.
"Microbe to man Evolution needs to develop those processes from scratch, but they are need for life."
False. Microbes already have those processes. The author has equivocated between evolution and abiogenesis.
"Genetic algorithms are designed by intelligent programmers with a specific problem in mind and fully functional from the start."
Again the author is confusing the environment and the 'organisms' within it. This criticism is like complaining that evolution didn't happen because the earth already existed. It is nonsense.
"The "organisms" would have to be a fully functional program, with a detailed programing language that tells it how to do every thing."
Many are. If the author is unaware of systems such as tierra, they are not qualified to write this article.
"The "organisms" would have to develop the programing language from scratch with no input from a programmer."
False. This is akin to claiming that self-replicating molecules must develop the laws of chemistry.
"The "organisms" would have to develop a system to read and write the programing instructions also from scratch with no input from an intelligent agent."
This is the same point as above, and fails for the same reason.
"The "organisms" would have to develop the entire operating system from scratch with no input from a programmer."
Utterly ridiculous. Unless the author thinks the earth was built by early life forms.
"The "organisms" would have to develop and build the computer memory and processor from scratch with no input from an intelligent agent."
Even more ridiculous. Apparently the author does think the earth was built by the first microbes.
"Genetic algorithms have a narrow definition of fitness. The "fitness" of the "organism" is measured based on how well is fits a specific problem. In the real world organisms ether live or die. If they live long enough, they usually reproduce."
This is meaningless. In the real world all organisms die, and evolutionary algorithms are the same. In the real world organisms either survive to reproduce or they don't; ditto for evolutionary algorithms. The difference is that in evolutionary algorithms survival/reproduction may be based on something other than finding food and a mate while avoiding predators.
"According to Wikipedia this is a type of problem that genetic algorithms (GAs) can not effectively solve."
Argument from authority.
"One example touted by Evolutionists is an "artificial life" program called Avida (http://dllab.caltech.edu/avida/). Despite the claims about this program, it dose not come any wear near showing the possibility of microbe to man Evolution."
Nor can it make maple waffles.
"One flaw is that each bit of the "genome" (http://dllab.caltech.edu/avida/v2.0/docs/cpu_tour.html) makes up a complete command, and one that is actually encoded out side the genome, this dose not fit the genomes real organisms."
The chemical properties of nucleotides, amino-acids and proteins are also 'encoded' outside the genome. The author's objection is groundless.
"Unlike most genetic algorithm programs Avida (http://dllab.caltech.edu/avida/) does include two reproduction commands an part of its "genome" (http://dllab.caltech.edu/avida/v2.0/docs/cpu_tour.html) but they only tell the "organism" when to reproduce and by what mode ( sexual and asexual ) to use. In both cases the actual instructions are out side the "genome" (http://dllab.caltech.edu/avida/v2.0/docs/cpu_tour.html) and are thus unaffected by mutation. This does allow for a mutation that renders an "organism" sterile, but no mutation changes the preprogrammed instructions inside each command."
Other simulations (such as tierra) do however include the possibility of mutations which affect reproduction. The author is choosing to criticise simulations based on what the simulation builders chose not to include, while ignoring the results from other simulations where those features were included. This is unethical.
"These "artificial organisms" do not developed new abilities, that are not designed into the program, but simply rearranges existing abilities."
False. cf the radio example.
"Avida (http://dllab.caltech.edu/avida/) starts with a created kind of "organism" and only produces varieties of that organism,"
True, but
"in perfect agreement with Creation science."
False. Creation science requires many created kinds, not just one. Avida more resembles evolution from a common ancestor.
Some of the above criticisms are minor, and easily correctable. Some require a major overhaul. But the fact that the author believes that evolution can only be simulated by having the simulated organisms build the computer in which they reside is sufficient reason to insist the rewrite be done by someone else entirely.
Roy 12:00, 8 Sep 2005 (GMT)
Delete
I think the text below must be deleted:
In a given environment an organism has two answers by which its fitness for evolution is judged: live or die (right or wrong). Thus, real world organisms have no way to converge on a solution.
Because this is more complex: the fitness for evolution is judged by the number and quality of the offspring left, also.
Luiz Alexandre Silva 13:28, 3 May 2012 (PDT)
