A dog, a cat, and a language model
LLMs in some ways are grift extension of Internet 1.0. Without any interactor or internetwork settlements all risk is on the receiver. Think of email. Think of social media. In an economic based settlement system there are incentives and disincentives that reduce the risk from individual actors. In general, well built networks reduce risk and overtime are generative and sustainable and approximate those found in natural ecosystems. Networks that are risk imbalanced (aka Internet 1.0 and LLMs) result in pretty bad outcomes. We need to bake in economic settlement systems within and between LLMs before things take a turn for the worse. Unfortunately, most of humanities networks are based on similar principles and we have to address those as well.