Embed Notice

Observation 1: machine learning algorithms tend to get stuck in so-called "local maxima" where any intermediate steps towards the optimal solution look less promising than where they currently are. The solution is to throw in some complete randomness so that they can jump to a different part of the solution space.

Observation 2: probably about half of all human inventions throughout history have been made by someone doing something completely boneheaded, then realising the outcome is interesting and promising in its own right.

Conclusion: being an utter ditz is a human evolutionary strategy. Stand proud!