"Google told partners that it plans to use both human moderators -- the company recently announced it will have 10,000 employees focused on the task -- as well as artificial intelligence software to flag videos deemed inappropriate for ads."

They will record decisions people make as training examples and will use it to train ML models.