

The same image is shown to a lot of people. If a majority of people click on the same things, that is assumed to be the correct answer. And it is added to the training database. Occasionally you’ll get one that hasn’t been shown to enough people yet to know for sure. For those, they’ll usually accept any answer, even wildly incorrect ones. The thing is, you as a user never know which ones they already know and which they don’t.
It’s fancy text completion - it does not have judgement.
The way he talks about it shows he still doesn’t understand that. It doesn’t matter that you tell it simmering in ALL CAPS because that is no different from any other text.