nostr:npub1qq7hl5sl6z0l0ahk8f6a4uv5mkvlama7dyvucdmtwdva2zg24xnqyafedr I think it a little reductive but it's not a _bad_ analogy; it is both adversarial and generative, and through anonymity strips individual contribution of anything distinguishing. But there's something more human about it -- instead of top-down hyperparameter adjustment to force an output for a given input, its output is always voluntarily provided and lacks any form of attention like the generative models we all love right now.

I like to think of it as an upgrade to the oral tradition that has served humans for 30,000 years. A node can only benefit from the network by putting in the effort to align with how the network presents and receives information, which is not the case for neural networks, and at any given point a node can spin off an entirely new network should that node not like the output of the imageboard.

We've always argued, we've always shitposted, we've always admired feet and we've always spammed /b/ with porn to slide the meaningful threads. It's so human that treating neural networks as on the same level is just a little insulting.