Comments
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...
A strong skepticism of flashy marketing.
Really, I think this one is massively under-emphasised because of the necessity of marketing tactics to many of EA's outreach efforts.
But I think that learning to decouple actual measurable impact from a slick, well-presented, eye-catching markety whatever is a huge part of becoming the kind of EA that can be trusted with power. And that happens by developing an almost aesthetic revulsion to anything that looks too flashy: a personal warning system to proceed with caution and not be taken in by something attempting to hijack your importance-measuring-sense.