Easy -- their methods aren't sufficient to begin with. No amount of training data would be enough. But perhaps they can develop new methods with what they've learned.