Lex is right, Sam was wrong. Our research shows that people don’t really check the work of AIs, once those AIs cross a certain threshold of quality. We found that consultants, for example, “fell asleep at the wheel” and did not look for AI errors. Training didn’t help, either.