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...
I am an undergraduate majoring in applied math and I am trying to spin towards alignment research. I am finishing up a course in the math of machine learning offered by UIUC. Towards helping others who are in a similar position, I've been working on an article about who might be well-suited for the course, the concepts that you learn about, and ultimately how helpful the course is in becoming a better alignment researcher in comparison to other avenues. At this time I am still writing the post, but I am not very certain about the verity of even the core arguments that I make. So I hope that I can get feedback from others. If the article has misconceptions, the feedback would help me rectify them before I publish this for public viewing. (A potential counterpoint is that the voting mechanism would result in the prominence of the post organically being reduced if many people find fault with it.)
Send me a message or comment if you're interested. I appreciate anyone who'd be willing to provide feedback on this.