Been listening to an Economist podcast about all the wonderful things good people are doing to make AI models safe. My problem with this is that I don’t see anything to prevent bad people from making unsafe AI models. And so much of the development is easy that this is not just possible, but inevitable.