As artificial intelligence rapidly evolves, a commentary from Nature Human Behaviour challenges a prevailing notion: that language drives human thought. Authored by Evelina Fedorenko (MIT), Steven T. Piantadosi (UC Berkeley), and Edward A. F. Gibson (MIT), the article asserts that scaling up language models will not yield human-like reasoning. Their research, blending cognitive science and neuroscience, demonstrates that language, while essential for communication, is not the foundation of intelligence. AI systems like ChatGPT and Gemini, despite their fluency, simply predict text patterns and do not engage in genuine thought processes. Furthermore, studies reveal that even individuals with impaired language skills maintain reasoning abilities, affirming the independence of language from cognition. The commentary highlights that true intelligence encompasses broader cognitive functions, urging a reevaluation of AI development. As AI evolves, understanding that language proficiency alone won’t lead to human-level intelligence is crucial for future advancements.
