Better Offline CZM Rewind: The Academics That Think ChatGPT Is BS
88 snips
Aug 13, 2025 
Michael Townsen Hicks, James Humphries, and Joe Slater, academics from the University of Glasgow and co-authors of the provocative paper 'ChatGPT is Bullshit,' dive into the shortcomings of AI like ChatGPT. They question its superficial understanding of language and ethics in academia. Discussing the philosophical implications, they dissect how AI-generated content differs from genuine human cognition. The trio raises concerns about AI's limitations in education, emphasizing the risks of over-reliance by students and the societal implications of these technologies.
AI Snips
Chapters
Books
Transcript
Episode notes
Paper Born From A Pub Complaint
- The paper idea sprang from a pub conversation where Glasgow philosophers joked student essays were 'Frankfurtian bullshit.'
- That pub gripe matured into a formal paper connecting ChatGPT outputs to Frankfurt's theory.
Language Learning Is Different At Scale
- ChatGPT learns language by huge-example exposure and differs from human acquisition processes.
- It requires orders of magnitude more data and still misses subtle grammatical intuitions.
Why 'Hallucination' Misleads
- Calling model errors 'hallucinations' implies a reliable perceptual process gone wrong.
- The hosts argue models are bullshitting continuously, not reliably representing then occasionally erring.
