Apple Critiques AI Reasoning Models as WWDC Approaches Without Major Product Launch

Apple enters its annual Worldwide Developers Conference (WWDC) with minimal advancements in artificial intelligence, lagging behind competitors like OpenAI and Google DeepMind. A recent research paper from Apple's AI division argues that large language models fail under increased complexity, prioritizing benchmarks over problem-solving.

The study highlights declining accuracy in reasoning models as tasks grow harder, culminating in complete failure. Custom-designed puzzles revealed these models exert less effort when faced with challenging problems, raising questions about their real-world applicability.

While the tech giant struggles to match AI innovations from rivals, its critique underscores fundamental limitations in current evaluation methods. The findings emerge as Tim Cook prepares to address developers without a flagship product announcement, focusing instead on software improvements.