We report the results of a survey of truck drivers: Those who are most concerned about automation are, counterintuitively, also most likely to say they intend to reinvest in that occupation. This finding does not hold among drivers who are most familiar with the low costs of community college. This is consistent with a simple model we develop in which noise in the perceived cost of retraining can lead to inefficient outcomes, which suggests that providing information on retraining can have large welfare consequences.