Genetic Programming

Genetic programming is a technique to create algorithms that can program themselves by simulating biological breeding and Darwinian evolution. Instead of programming a model that can solve a particular problem, genetic programming only provides a general objective and lets the model figure out the details itself. The basic approach is to let the machine automatically test various simple evolutionary algorithms and then “breed” the most successful programs in new generations.

While applying the same natural selection, crossover, mutations and other reproduction approaches as evolutionary and genetic algorithms, gene programming takes the process a step further by automatically creating new models and letting the system select its own goals.

The entire process is still an area of active research. One of the biggest obstacles to widespread adoption of this genetic machine learning approach is quantifying the fitness function, i.e to what degree each new program is contributing to reaching the desired goal.

Problems

See the wiki page o Artificial Intelligence fo a discussion on the problems found in selecting a goal that will avoid inappropriate results.