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...
note: crosspost from my substack
As a vegan for almost thirty years, I’ve long had second thoughts about how effective veganism is for helping animals. I am not alone in this. Recently others have expressed doubts about veganism (see for instance here,...
TLDR: Everyone’s talking about what the money could do, but few about how to decide where it goes.
This post is part of the new series of articles on cross-cause giving and the new wave of philanthropy. Stay tuned to the EA Forum and our Substack for the latest takes on topics such as giving now vs. later, common pitfalls in cause prioritization, and other crucial considerations from the Cross-Cause Fund (CCF) team...
Some quick observations on the attacks on METR, EA, and other AI safety orgs (from my perspective as a political campaigner and former Communications Director for a union and mayor during COVID):
I want to write something longer on this but don't have time right now. Lots of opportunity for good crisis comms here. Let me know what's wrong about this take and if you'd like to hear more!
3 & 5 make sense.
1 / 4 / 6 feel like they might make sense strategically, but I'm proud to be part of a community that often values epistemic humility and correct argumentation above strategic communication.
That's a big part of why I started posting in the EA Forum and on LessWrong in the first place. I'm learning a lot from it!
Here are some quick thoughts:
1, 4 and 6 aren't asking anyone to say something they don't believe, they're more about who says what and where. If someone outside the AI safety world (or better, someone who's disagreed with METR before) says "the corruption story isn't true," that's not less epistemically correct than an insider saying it, it's just more believable to the general public.
I'd gently push back on the bigger part of your arguement though. If the public conversation ends up being "is METR independent" instead of "should anyone outside the labs be checking their work," there's a real cost to that, especially if we end up losing the larger comms battle.
Curious to hear more about what feels wrong to you or where the line would be crossed.
Says "if you're explaining you're losing". IMO explanations are good. You might be correct that needing to explain is a bad sign for whether you are winning a comms battle. But I would still rather offer people genuine explanations, so we can come to a shared understanding of the truth (exactly like this convo).
My reading of this is that you're claiming who says something is more important than what is said. Again, you may be correct here on how to optimally persuade. But I personally want to value good arguments, no matter who makes them.
Feels like it is straightforwardly asking folks to not engage in debate or disagreements. I don't really get how we come to a shared understanding of the truth if we don't offer clear reasons for why we disagree with particular points.
Again, I very much am not saying you are wrong about the implications of ignoring strategic communication norms. You are likely right. I would nonetheless prefer to be in a community that decided to communicate honestly and earnestly, even with people that it disagrees with.
Is that actually true of X's reply mechanism to a significant degree? My impression was that replies on X are mostly just seen by people who tap on the original post. Replies wouldn't necessarily have the effect of amplifying the original post?
Certainly writing Community Notes would not present much risk I presume??
Thanks for pointing that out.
I meant "if you're explaining, you're losing" for any platform, not X replies specifically.
I agree Community Notes are very likely net positive (they certainly have an impact on me). But where I still see some risk is volume: Replies are a signal and if a lot of people pile on to rebut a post, perhaps that pushes it to more people? (I don't know X's algorithm well enough to say that with much certainty.)
I also don't want to double down too hard on illusory truth. I defined it a bit too loosely in my quick take, and there are follow-up studies showing a good correction usually wins with the people who read it.
For me it's more about Zaller's Receive-Accept-Sample (RAS), which I find helpful for thinking about mass/political comms. Voters have to:
Continuing to respond to a critique that's wrong makes it more likely people 1) hear it and 3) sample it later. Step 2 is where I think it matters most: People who are already predisposed to accept the critique will likely reject the correction, so for them the rebuttal could deliver the attack?
Thanks for continuing to engage.
Thinking aloud here. I don't think you necessarily have to respond by quoting the person. You can just provide a frame or facts which counter or inoculate against whatever critiques are currently widespread. E.g. if someone accuses you of a crime, you can mention that you were someplace else at the time the crime was committed, without directly repeating the criminal accusation. However, your X replies might fill with accusation talk in that case.