Headline finding: I audited 17 AI Safety Talent programmes. Zero of 17 have published any comparison group, rejected-applicant follow-up, matched control or randomisation. Not one. Every programme that mentions a counterfactual does it by asking participants to self-report.
Background
At least $70 million...
In celebration of still being alive and fighting, we are giving away 1,000 Amazon e-books of “If Anyone Builds It, Everyone Dies”. Feel free to send a copy to yourself, a loved one, or a friend—we need all hands on deck.
Today marks exactly one year since If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All, by Eliezer Yudkowsky and Nate Soare...
A year ago I posted that YCombinator (YC) companies didn’t seem to be growing faster since the release of ChatGPT in 2022. I reran that experiment and found that 2023+ YC companies are arguably growing a bit faster than pre-2023 companies (including AfterQuery, ...
If we include the nonhuman animals, the world is not clearly much better now. It might be much worse now than before, or much better, depending on your view on wild animal suffering.[1] But unfortunately the article entirely disregarded the nonhuman animals.
So it is entirely possible that "The world is awful. The world is much worse. The world can be much worse."
But to leave things at a slightly more positive note, let's say "The world is awful. The world is probably much worse. The world can be much better.
A noteworthy point is that even if factory farming really does reduce the amount of suffering, humans deserve little praise for doing it as:
1. We clearly did not do it to reduce wild animal suffering - wild animal suffering was likely entirely not in the minds of people who developed factory farming. But it is fair to say that farmed animals' suffering were in the minds of the people who developed factory farming
2. There are extremely likely other ways that can achieve more reduction of wild animal suffering than cutting rainforests (which reduces the number of wild animals), growing crops on these land and then feed the crops to raise animals we abuse to retroduce another source of suffering.
Hi Fai,
I agree on the point about non-human animals, but I guess if we should account for future beings too. 1k years ago, we had not fully realised how much humanity (or post-humanity) could flourish, because it was not clear settling the galaxy etc. was possible, so I think the utility of the future has increased a lot in expectation (as long as you think the future is positive). If we decrease existential risk, we can increase it even further. In other words:
Thank you for your reply!
I think the future is not clearly positive if we also consider non-human animals (and digital people and animals). I think realising colonizing the galaxy could be a bad things: For example, by spreading wild animal suffering and factory farming.
Fair enough! I agree with Saulius that digital minds might be much more important than WAW in the future. I see you wrote about why the expected numbers of farmed animals in the far future might be huge. I have only read the summary of your piece (added it to my reading list now), but I agree that "digital people will, presumably, have very few incentive to raise animals for food, or even other purposes". In addition, I think there are scenarions in which digital beings dominate total utility by many many orders of magnitude (OOMs), whereas I find it hard to imagine wild animal welfare dominating by much more than the current 8 OOMs (roughly my best guess).