The development of universal problem solvers, like the Gödel machine and Markus Hutter's method, come with the overheads of proof search, ensuring that the solutions found are optimal. However, non-universal techniques like recurrent neural networks and local search are much more practical for solving smaller problems.