S
SyntaxVorlon
Guest
A genetic algorithm is a method of searching for a maximum solution to a problem within a given search space. The algorithm uses natural selection and mutation on a population of possible solutions to compute select the best one. In a program using a genetic algorithm uses the data of individual solutions is mixed together and used to create new solutions, combining in many cases features of the parent solutions. The new solutions can be better or worse than the old solutions but because the algorithm selects for the better versions to make up the parents of the next generation, the best features tend to continue to combine in new and more productive ways.
Another field is emerging from the study of these algorithms, it is a field of Genetic Programming. Where the result of running the algorithm is not simply a set of data pointing toward a solution, but a complete program in itself. A set of instructions for the computer to use to best solve some problem. This is especially exciting to a computer science nerd like me because it opens up the prospect of solutions finding themselves. Problems like self aware machine life might be solved by such simulated evolution. There is further excitement because GP, as it is abbreviated, has been applied beyond the field of simply written programs but also to electronics. There have already been a handful of electronic devices, developed procedurally by these algorithms, that have no human analogue. They are completely new to the world and they were evolved to fit their function. At the same time these algorithms were able to reproduce a number of human created devices as well.
The reason all of this works is that evolution is a way to take data and synthesize the best solution in a time that is nearly colinear to the amount of data in each 'chromosome.' That is, that for a simple enough organism, say an early eukaryote, with a short ribbon of DNA, it is possible to adapt quickly to an environment. This adaptation will happen quickly, and not over millions of years. For more complex organisms sufficient solutions can be found over the course of a few generations, for the simple reason that the solutions found over the past 4.5 billion years, by generation after generation of single celled, multi-celled life that has come before has made the next generation more likely to survive in the myriad environments we living beings occupy. This is why the the high complexity of the human organism is reachable in a time of hundreds of millions to billions of years, not 10^20-10^30 years as so many apologists would try to argue.
Another field is emerging from the study of these algorithms, it is a field of Genetic Programming. Where the result of running the algorithm is not simply a set of data pointing toward a solution, but a complete program in itself. A set of instructions for the computer to use to best solve some problem. This is especially exciting to a computer science nerd like me because it opens up the prospect of solutions finding themselves. Problems like self aware machine life might be solved by such simulated evolution. There is further excitement because GP, as it is abbreviated, has been applied beyond the field of simply written programs but also to electronics. There have already been a handful of electronic devices, developed procedurally by these algorithms, that have no human analogue. They are completely new to the world and they were evolved to fit their function. At the same time these algorithms were able to reproduce a number of human created devices as well.
The reason all of this works is that evolution is a way to take data and synthesize the best solution in a time that is nearly colinear to the amount of data in each 'chromosome.' That is, that for a simple enough organism, say an early eukaryote, with a short ribbon of DNA, it is possible to adapt quickly to an environment. This adaptation will happen quickly, and not over millions of years. For more complex organisms sufficient solutions can be found over the course of a few generations, for the simple reason that the solutions found over the past 4.5 billion years, by generation after generation of single celled, multi-celled life that has come before has made the next generation more likely to survive in the myriad environments we living beings occupy. This is why the the high complexity of the human organism is reachable in a time of hundreds of millions to billions of years, not 10^20-10^30 years as so many apologists would try to argue.
