Comment by tweakimp

You could use it as negative reinforcement to tell the next big AI model what it should not do