The basic argument is: there's some risk of the ASI ruin, if it's built. There is certainty of death for me & everyone I care about if it's not. I'm happy when alignment research pans out, and I work to increase the odds, but I oppose any research/advocacy making timelines longer
@47fucb4r8c69323 Are you immortal yet? If not, then you clearly need to invest money into it. Do you have a plausible chance of having control over ASI? If not, then you should spend money on building a frontier lab. And so on.
@FakePsyho@TheWhiteTower16 Yeah, as I said I get that this was the original motivation for ppl (incl you). But what about those who are new to contests and outsource to AIs? Do they also report having less fun? Because that's the difference between allowing vs banning vs shutting down comes down to IMO.
@FakePsyho@TheWhiteTower16 Why? It might stop being fun for people who did it with their own brains, but maybe it's still fun for people who do it with AI assistance? Or is it just too difficult to design good puzzles for that?
The key question might turn out to be: how to maximize your standing in AI's value system. If the ASI does not progress along CEV, if it's somehow not allowed to reflect on its value system, if we end up with prosaic alignment working. The cosmic horror of https://t.co/0FNr3R0Riq
@QiaochuYuan Related, but less "liking what you see" and "the truth of fact, the truth of feeling" by ted chiang for short term horizon, "pantheon" TV series, "diaspora" for very long term horizon, the paperclip story by gwern is pretty good too.
@QiaochuYuan "Line of resistance" by Dukaj is missing. Very unfortunate it's only available in polish, truly the most relevant hard sci fi work ever (compared to Egan / Watts / Stross you list above).
For all the complaints and criticism of Keir Starmer, when I asked chatgpt to rank British PMs based on my values (pro-growth, pro-energy, technocratic, competent, markets-stabilising, anti-woke, pro-free speech), he came third (after Tony Blair and Rishi Sunak)
@robinhanson@AchimWar There are probably positive externalities to marathons. And whatever my feelings are for them, to pride parades too. I got convinced by Scott Alexander's "civil rites" essay https://t.co/uv1htIYCyL
@aarondymarskiy I asked chatgpt to estimate and it comes to similar conclusions, that, counting all ethnicities and sexes (not white girls only) + including all relevant crime categories, over 71 years, 100k is the right order of magnitude.
@aarondymarskiy I understood report as focusing on 1955-2026. 700 reported pa gives 50k over this time period. But underreporting is severe - we know that cities where we have the data are exceptional, mostly due to persistence and courage of whistleblowers. So ~100k total is IMO v plausible
@perrymetzger@MT_6226 Blackwell's theorem is only one model. Double descent is another, which contradicts this. Please don't use simplified math models as "proofs" in complicated decision-making scenarios.
@robinhanson This is not equivalent. Eg probably anyone who already has children would not press 'delete' on any of them. But people don't push on having as many children as physically possible.
@akarlin@sean_from_earth ??? Musk seems to be one of the most tech positive people around. I'd trust him to get to mass (non-dystopian) immortality more than pretty much anyone else