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...
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...
This is part one of a two-part series. Part two is available here.
“Incubator” is a very vague term[1]...
I enjoyed Cloud Atlas (fiction, both book and movie). It's a movie where relatively normal people did hard things because they were the right thing to do, not because movie morality entailed a just world hypothesis where good things happen to good people, or because the movie's characters were personally and emotionally invested in specific outcomes.
I think this is relatively rare in fiction, which is why I like the movie. I've told several EA friends about the movie and they liked it too.
I also watch movies about ordinary people resisting during World War II (eg, John Rabe for nonfiction and Jojo Rabbit for fiction), especially if it's from the perspective of the Axis.
+1, a lot of David Mitchell's metafiction-y books have EA vibes (huge scope, interwoven characters doing the right thing for unknown future others)