A perfect example of my recommended approach to AI: experiment but double check the outputs