A new study suggests that today’s leading AI models, despite their remarkable language capabilities, may lack the ability to perform the complex reasoning that underpins genuine human intelligence. Researchers point to failures in problem-solving architecture that call into question how close we really are to machines that think like us.