> Serious question, what is there to ai safety? ... What are the real risks involved? One set of risks are hypothetical, and basically are about avoiding thing

> Serious question, what is there to ai safety? ... What are the real risks involved? One set of risks are hypothetical, and basically are about avoiding things like the paperclip maximizer example. Another set of risks are more practical and involve the existing technology, like how to you keep it from being abused for ill. Realistically, that mostly amounts to trying to avoid PR problems, because I can't imagine those guardrails would stay on when the company doesn't want them there.