AI flunks logic test: Multiple studies reveal illusion of reasoning

Recent studies show top AI models struggle with complex reasoning tasks like the Tower of Hanoi puzzle and Math Olympiad problems. Performance declined as complexity increased, with models making basic errors and contradicting themselves, even when given explicit algorithms. These findings challenge claims of AI reasoning capabilities, suggesting pattern matching rather than true reasoning. The results emphasize the need for hybrid approaches combining language models with symbolic logic.