Loading...
I think is a very high stakes question. I wrote about it two weeks ago, if you have an AI swarm doing a bad thing, one solution may be a reward hotline for a model that blows the whistle. But only works if the model reasons you’ll honor yours commitment.