Freeport-McMoRan stock tumbles after Trump imposes copper tariffs
Investing.com-- Apple Inc over the weekend released a research paper claiming that artificial intelligence models geared towards reasoning had limited capabilities and failed to generate accurate results beyond a certain level of complexity.
In a paper titled “The Illusion of Thinking: Understanding the Strength and Limitations of Reasoning Models via the Lens of Problem Complexity,” Apple (NASDAQ:AAPL) researchers claimed that larger reasoning models (LRMs) had clear gaps in the quality of their reasoning and failed to develop general problem-solving capabilities.
Researchers tested LRMs such as OpenAI’s O1/o3, DeepSeek-R1, Claude 3.7 Sonnet Thinking and Gemini Thinking through increasingly complex problems which also deviated from standard AI testing benchmarks.
Apple researchers used “controllable puzzle environments” to test the models, and found the performance of the LRMs deteriorating, eventually reaching zero in the face of increasing complexity.
“We show that state-of-the-art LRMs (e.g., o3-mini, DeepSeek-R1, Claude-3.7-Sonnet-Thinking) still fail to develop generalizable problem-solving capabilities, with accuracy ultimately collapsing to zero beyond certain complexities across different environments,” Apple researchers wrote in the paper.
Researchers said testing showed LRMs had “fundamental inefficiencies” and clear limits to their scaling abilities. Researchers also questioned the current evaluation methods for LRMs based on established mathematical benchmarks, and said they had designed a more controlled experimental method by using algorithmic puzzle environments.
Apple researchers questioned claims of LRMs being a significant step towards general AI– a theoretical form of AI that can emulate the broad range of cognitive abilities and problem solving skills shown by humans.
General AI has been long touted as an eventual goal by major developers, although it still remains highly theoretical in nature. Current AI models, specifically large language models, use pattern recognition to predict the next word in a sequence to generate new text, which still leaves them prone to a high margin for errors and limit their reasoning capabilities.
Apple’s paper comes just days before the company’s Worldwide Developers Conference on June 9, with expectations low after the company’s AI efforts largely lagged their rivals.
Apple has been struggling to roll out features promised under its AI offering– Apple Intelligence– despite partnering with OpenAI to enable AI features in its flagship devices.