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...
Often folks hit us up because they are thinking of starting an incubator and want advice.
Typically their motivation is either that (a) they have a list of specific things they want built that no one is building, or (b) they think an ecosystem needs more new projects generally to absorb more talent and deploy more funding effectively.
Here are six questions we often ask prospective teams, to help them figure out what to do. If you're incubator-curious...
The Straw and the Camel's back
I recently had a colleague complain that oat milk was a 'luxury' that the work coffee machines didn't need. And this tiny little comment kind of broke me. I feel like I am so careful not to judge or lecture everyone around me for their insanely massive moral failings around animal welfare, or donating - yet apparently people can't even just let me have my suffering-free milk in peace.
Which prompted me to re-evaluate something I hadn't really thought about in a long time - being EA (or EA-adjacent or however people wanna identify) is just really hard sometimes. I used to be more actively advocatey about things, but it can be exhausting, and at some point I just kinda stopped. But now I feel very motivated to figure out how to start being a lil more vocal again, because it turns out that pretending like I don't have strong opinions on these things is also exhausting!
Which is all to get to the point of: there are a lot of posts on here about EA being hard, and how to talk about EA, and reading those posts helps get a feeling of support but knowing this doesn't magically make it all easier. I am just really grateful for this awesome community, and want to just normalise a bit more to share when it gets hard because thats ok. We are doing a hard thing.
(Note: while this one colleague clearly pushed my buttons, further reflection got me very happy that clearly a bunch of other people had been advocating to get the oat milk at LUMC and I'm very happy they exist and that they succeeded)