Jack Roycroft-Sherry The Secret to Human Thought AI Will Never Understand
Mar 20, 2026
Elliot Murphy, a neuroscientist who studies language, cognition and the brain, offers a brisk tour of what makes human language unique. He explores innate language capacity, structural ambiguity and sentences humans grasp that AI misses. He examines language’s evolution, its role in deception and whether something beyond our current language could exist.
AI Snips
Chapters
Transcript
Episode notes
Language Evolved For Deception As Well As Communication
- Language evolved as a powerful tool for deception and manipulation as well as communication, so clarity isn't its primary design goal.
- Murphy cites Dawkins and argues ambiguity and misdirection are intrinsic advantages of language.
Don't Mistake Prediction Effects For Language Mechanisms
- Interpreting brain signals as evidence that language is next-token prediction is circular if you define language that way.
- Murphy warns neuroscience must target formal algebraic properties, not just prediction effects captured by LLM-like models.
Specify Formal Commitments In Language Neuroscience
- When studying language neuroscience, specify formal commitments and the exact grammatical computations under study.
- Murphy recommends against vague language localizer tasks and for targeted experiments teasing apart components like merge and semantics.