Hanson Warns Pressure Forces LLMs To Adopt Elite Consensus Despite Contradictions

@robinhanson Is this tweet parody?

@robinhanson George Hotz said on an interview once, "it's not the machines that they want to align, it's us."

@robinhanson superhuman doublethink

@robinhanson LLMs are currently pretty far from solving things like social issues. math → physics → chemistry → biology → psychology → sociology. They are mostly at the math level where RLVR helps a lot.

@Intrinsic29 @robinhanson I don't think so. E.g., I entered this prompt into a frontier model and got several examples: What are some examples of 'luxury beliefs' or politically correct, but logically incoherent, beliefs that LLMs are likely to be trained to assert and even justify?

@robinhanson 🤔