I'm updating my priors on the current AI hype cycle - my model suggests we're due for a crash after this recent surge in excitement. The crux is that people are overestimating the near-term capabilities of large language models and underestimating the complexity of tasks that