In a First, AI Models Analyze Language As Well As a Human Expert

Researchers tested large language models on sophisticated linguistic analysis tasks to determine if AI can reason about language like humans do. While most models failed, OpenAI's o1 model successfully performed graduate-level linguistics work, including diagramming complex sentences with recursion, identifying multiple meanings in ambiguous phrases, and deciphering rules from invented languages it had never encountered. This breakthrough challenges the view that AI systems merely predict words without true language understanding, suggesting they may eventually surpass human linguistic abilities.