Apple: "LLMs are faking intelligence—it's all an illusion! 🎭"
Their evidence? 4 logic puzzles where EVERY model (yes, even o3 and Claude) hits 0% accuracy above certain complexity thresholds.
But here's the thing: they tested AI like testing fish on tree-climbing. The real illusion might be thinking these narrow tests reveal fundamental limits.
Paper: https://ml-site.cdn-apple.com/papers/the-illusion-of-thinking.pdf