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...
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...
It's OK to ask "Who can DM me a quick review of [EA-run service]?"
Problem: It's costly for EAs to find out which EA-run services will actually help them.
Proposed partial solution: Normalise asking for private reviews of each other's services.
Thanks to Jennifer Waldmann and Ozzie Gooen for helping me think this through.
Strong upvote from us.
Two natural places to ask are Bountied Rationality and the EA twitter group.
Following my own advice: I will not be offended if I see someone asking "Has anyone used Pineapple Operations who can send me a quick review in DM?" on the Forum or on Slack etc (although I think we're pretty low-cost to use at the moment, so maybe not the best example).