x
Uncertainty Without Qualification: The Standard Output Problem in Autoregressive Language Models — LessWrong