- Genetic algorithms try to find the maximum value of a fitness function over multiple variables
- An individual or behavior is completely determined by its set of genes. In 2D, (x, y) is an individual whose fitness is f(x, y)
- Theory: if f(xi, yi) and f(xj, yj) have large values, maybe f(xi, yj), f(xi + delta, yi), etc. have even better values.
- Not guaranteed to converge, "messy results" because of dormant genes