Past Performance is Not Indicative of Future Results

This week on my podcast, I read my November 2020 Locus column, Past Performance is Not Indicative of Future Results, a critical piece on machine learning and artificial intelligence that takes aim at the fallacy that improvements to statistical inference will someday produce a conscious, cognitive software construct. It’s a followup to my July 2020 column, Full Employment, which critiques the fear/aspiration of automation-driven unemployment.

The problems of theory-free statistical inference ...