The Myth of Artificial Intelligence: Why Computers Can’t Think the Way We Do

by Erik J. Larson

tuning a system to learn what’s intended by imparting to it a desired bias generally means causing it to become narrow, in the sense that it won’t then generalize to other domains.