So many people have this all backwards. The concern here isn't that Erbai 'convinced' the other robots, it's that the other robots COULD be 'convinced' at all. Whatever went on here, the failure was in the machines that deviated from their expected behaviours, not the one that succeeded in its task.