Large Language Models and Scientific Discourse: Where's the Intelligence?

A research paper argues that Large Language Models lack true intelligence and understanding compared to humans, as they rely on written discourse rather than tacit knowledge built through social interaction. The authors demonstrate this through examples like the Monty Hall problem, showing that LLM improvements come from changes in training data rather than enhanced reasoning abilities.

• →LLMs cannot access tacit knowledge built through spoken discourse within expert communities, which is crucial for early scientific knowledge formation.
• →ChatGPT's improvement on the Monty Hall problem was due to changes in available written discourse, not enhanced reasoning capabilities.
• →LLMs fail when faced with small prompt variations that make established answers nonsensical due to 'overshadowing' by dominant discourse patterns.
• →The research argues that true intelligence resides in humans rather than in the LLMs themselves.
• →Scientific knowledge creation relies heavily on social discourse and expert consensus that LLMs cannot currently replicate.