Bob Murphy Show Ep. 470 Are There Theoretical Limits to AI?
13 snips
Dec 5, 2025 Adam Hayman, a podcaster known for the Hayman Nature show, joins forces with AI commentators Jobst Landgrebe and Barry Smith to dive into the limitations of artificial intelligence. They explore the Clever Hans analogy, arguing LLMs merely mimic patterns without true understanding. The trio debates the feasibility of simulating evolution and how Gödel's Theorem poses limits on computational minds. They touch on LLM hallucinations and the challenges of creating embodied AI, while pondering the ethical implications of digital simulations and information in life.
AI Snips
Chapters
Books
Transcript
Episode notes
Pattern Matching, Not Understanding
- Jobst Landgrebe and Barry Smith argue LLMs mimic patterns, not genuine intelligence.
- They compare ChatGPT to Clever Hans: impressive outputs without true arithmetical understanding.
Novelty Is Central To Intelligence
- Intelligence requires meaningful responses to wholly novel situations without prior training.
- Landgrebe claims LLMs never truly do novelty because they rely on extracted patterns and repetition.
Clever Hans Read The Crowd
- Bob Murphy recounts the Clever Hans story and clarifies the horse read crowd cues, not do math.
- He stresses the horse also used subtle crowd signals, making the illusion more convincing.
