Emergence. My favorite word for a while. Explained everything that I cared about. Then love turned to hatred. I would stop reading when it was mentioned. Because I realized, it had no explanatory power.
The sun rose again, and the cold in my heart thawed. I’m beginning to warm up to emergence again, because I think I’ve managed to formulate an example for myself.
Consider OpenAI and Anthropic. Classic duopoly, but early stage.
I can think of 5 other timelines AI could’ve gone so far. This is not about the future, I’m thinking of all that could’ve happened so far already. And the timeline we’re in is the best of them.
Why?
Because of the duopoly. If it’d been a monopoly — an outcome that, prima facie, seemed more likely, given the winner-take-all dynamics that frontier models can produce through recursive self-improvement (though we might not yet have reached that stage yet!) — then things would’ve been worse, for all the parties, human or AI, involved (except the monopoly, of course)
But it’s not a monopoly currently, it’s a duopoly, and that’s saving our tofu.
The thing is — this is emergence! We could’ve had the exact same set of people and AI models and resources and GPUs and money and CEOs and whatnot, yet the outcome would’ve been very different. The material separation of the exact same set of ‘building blocks’ into two different ‘organizations’ doesn’t change any of the building blocks but changes the outcome.
This, to me, perhaps incorrectly, but still usefully, is Emergence: when a whole made of interacting parts shows properties or behaviours that are not present in the parts taken separately.