The trouble with these black box models isn't that we're afraid of uncertainty, it's that we're worried they're using a bad proxy. Throwing more data at the alg

The trouble with these black box models isn't that we're afraid of uncertainty, it's that we're worried they're using a bad proxy. Throwing more data at the algorithm won't solve an omitted variable problem or an endogeneity problem.