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...
Not all capabilities matter. For example, the capability to burp really loudly is not a morally important one. If we were trying to improve the world, trying to give people the capability to burp loudly is not in the top 1000 list of things I'd suggest prioritizing.
And plausibly, the reason why this capability doesn't matter is because it doesn't promote wellbeing. And more generally, this might be true of any capability. The reason why the capability to get an education or access healthcare are important is precisely because they reliably lead to people living better lives.