Carl T. Bergstrom and Jevin D. West, professors at the University of Washington, have released a free online course exploring how to navigate a world increasingly filled with content from large language models. The course, hosted at thebullshitmachines.com, consists of 18 lessons, each taking five to ten minutes, designed for self-study or flipped classroom use. It is a follow-up to their earlier course and book ‘Calling Bullshit.’

The course does not require a technical background; Bergstrom and West frame it as a humanities course about learning and working in an AI-influenced world. The materials cover both the potential benefits of LLMs and their risks, describing the models as ‘both powerful tools, and mindless—even dangerous—bullshit machines.’ The professors say they aim to let students ‘explore how to resolve this dialectic’ rather than lecturing them.

On Hacker News, the course announcement drew 930 points and 603 comments. Commenters shared anecdotes of colleagues and acquaintances taking LLM output at face value without verification, with one noting a team member had generated a report on a specific subject from an LLM that was ‘inaccurate and unreliable.’ Others discussed the difficulty of distinguishing simulation of reasoning from genuine reasoning in models, and one commenter who said they work at OpenAI warned the current version could leave students ‘worse off,’ though they acknowledged the ‘autocorrect argument is typically (massively) overstated because RL exists.’