SE: Anatomy of a Controversy: Deconstructing Apple's "Illusion of Thinking"

Ai looks at a controversy in the AI community ignited by Apple's "The Illusion of Thinking" paper, which claimed advanced AI models exhibit a fragile "reasoning" ability that collapses under complexity. This paper argued that the perceived intelligence of Large Reasoning Models (LRMs) was an "illusion," based on experiments with classic puzzles like the Tower of Hanoi. However, a swift rebuttal, "The Illusion of the Illusion of Thinking," countered that Apple's findings were not due to fundamental AI limitations but rather flawed experimental design, pointing to issues such as token output limits, unsolvable problems, and overly rigid evaluation methods. The debate highlights the philosophical differences in defining and testing AI "reasoning," with the rebuttal suggesting that models might understand algorithms abstractly even if they struggle with exhaustive, human-like procedural execution. This discussion has led to polarized reactions within the AI community, influencing how researchers now approach benchmarking and evaluation of AI capabilities.