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, ...
I think this article in Vox covers some of the main points.
"...improving the American education system, while important, is neither a neglected cause nor a tractable one. It is a system on which hundreds of billions of dollars are spent annually by diffuse governments whose policies are difficult and expensive to change, where matters of importance are intensely contested, and where interest groups tend to fight each other to a standstill.
And it’s a system where, even after investing millions if not billions in research, we still don’t have a lot of confidence as to which interventions are helpful and which are not. The views of key actors, notably the Gates Foundation, have tended to shift rapidly on those substantive questions.
If every issue in the world were as crowded and hard to make progress on as education in the US, then I’d understand why foundations like Gates and Broad keep chugging. But that’s not the case..."