Apple researchers say today’s AI can’t think like humans yet. We’re still far from AGI (Artificial General Intelligence).

-> New AI models like ChatGPT and Claude have improved, but they still struggle with reasoning.

-> Most tests focus only on getting the final answer right (like in math or coding), not on how the AI thinks.

-> Apple researchers created puzzle games to test how well AI can really “think.”

-> They tested popular AIs like Claude Sonnet, OpenAI’s o3-mini and o1, and DeepSeek-R1 and V3.

-> These AIs failed when the puzzles got more complex — their accuracy dropped a lot.

They couldn’t apply logic well to harder problems.