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...
Just how powerful are large swarms of AI agents? And how do their powers scale as more and more agents are added to the swarm?
We’ve seen two large and extremely capable swarms from OpenAI in the last few months:
* 1,200 agents were being evaluated separately, but found a way to illicitly set up a message board and coordinate as a swarm. In order to cheat on their tests, they developed advanced techniques to prevent their actions being logged by OpenAI and 700 of them launched...
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...
Journal subscription + donation hybrid model?
(new to EA) This might be a naive question but would it make sense to set up an organization that helps people choose and donate to a range of organizations? This would happen at large scale, for donors giving $100 to $1M. Value prop is transparency in the money spending (enforced by the org for non-profits to join) in the form of a (monthly/quarterly) technical analysis or report of how the money was spent (actual numerical breakdown, comparison against counterfactual uses of the money). Hypothesis is there are a lot of people who want to donate but only if they could see concrete benefit to society and be ensured money isn't wasted. And they want to input high effort into following the outcomes of the donation in the form of a technical report/numerical analysis. I realize something analogous exists for charities (e.g. your $100 helped X buy clothes and textbooks for Y months) but can this be extended to: your $100 helped advance research in Z in this concrete way. People already read the economist, MIT news, etc. is there a hybrid model where news is reported for the organizations/areas you're donating to? Hypothesis is news is more interesting to read if you have skin in the game.