| What ideas pop up in your head about someone who deems their enemies "trust and safety" or "risk management"? Especially risk takers should manage their risks. Someone who complains about risk management when taking risks is usually not an innovator, but an opportunistic liability to progress and shared benefits. You are free to take risks where negative downsides of unchecked technology are taken by you, but AI will impact everyone ("playing God with everyone else’s lives, with total insulation from the consequences."). It is precisely because the potential and power of AI are so impactful and transformative that AI developers have more responsibility to not squander this and mess this up. Being optimistic about the benefits, should make one pessimistic about the risks. So, yes. Promote benefits. Build, create, fail, refine. Focus you attention fully on it. Ignore risks and risks management and trust and safety. But don't label these an enemy when these speak up. Doomerism has hurt AI safety discourse tremendously. Unrealistic and restrictive risk management hurts progress and reasoned debates. Roko and Yudkowski get way too much attention and dominate and control the public debate, pushing out moderate and careful arguments. The e/acc crowd similarly hurt AI safety: almost like these don't believe that either technology can be dangerous, or that such danger is to be accepted, or even required, to make progress. Maximize benefits, while minimizing risks. Both can work. One without the other won't. Don't try to launder your Waifu-generators as decentralization or progress. Don't damage AI safety by making everyone disgusted at the thought, because you just had to talk about existential threats of nano-bots, not the current and very real risks associated with (generative) AI. The problem with the current debate on AI safety, is that AI ethicists are too enthusiastic to come up with many rules to bind AI and stifle progress. While the AI acceleration have trouble to come up with just one rule we could put to a democratic vote. If you really can not come up with severe and avoidable negative consequences from computing, then maybe you should not have strong opinions on computer safety. Accelerate benefits, reduce risks. We owe it to the future we are building right now. |