The other side of the argument though, is the "reduce existential risk" part. There probably isn't anything meaningful I could do, to reduce the risk of some fu

The other side of the argument though, is the "reduce existential risk" part. There probably isn't anything meaningful I could do, to reduce the risk of some future war against aliens. But with AI, the presumption is that the risk is near and that AI research going on right now could meaningfully impact future AGI designs. AI research, say, 80 years ago wouldn't have the same tradeoff. I'm not sure I buy the assumptions, but the argument does require an assumption of near-term influence (albeit it allows that that influence to be very small).