in predicting the likelihood of someone contracting diabetes, a person’s weight and body mass index are strong features. AI algorithms do indeed factor in these strong features, but they also look at thousands of other weak features: peripheral data points that might appear unrelated to the outcome but contain some predictive power when combined across tens of millions of examples. These subtle correlations are often impossible for any human to explain in terms of cause and effect: why do borrowers who take out loans on a Wednesday repay those loans faster? But algorithms that can combine
...more
