Well no. GP is still the right solution:

1. Specific problems like #SymbolicRegression and #HyperHeuristics.
2. Gradual automated improvement via test cases and objective function.
3. Use cases where regurgitation / copyright infringment has to be avoided at all costs.