Researchers warn that relying on ready-made answers from AI can undermine intellectual effort and risks losing basic skills, especially when lacking supporting tools.
Wired magazine published an article titled "Using AI for Just 10 Minutes Might Make You Lazy and Dumb, Study Shows," reporting on research from leading universities including Carnegie Mellon, MIT, Oxford, and UCLA. The study highlights that even a brief 10-minute reliance on artificial intelligence capabilities can significantly impair human analytical thinking and problem-solving processes.
Experiments involving three test formats with several hundred participants showed that users who received AI assistance for solving problems—such as basic fraction calculations or reading comprehension—tended to give up more easily or performed less accurately when the AI help was later removed, compared to those who never used AI from the start.
Michiel Bakker, assistant professor at MIT and one of the researchers, explained that the key issue is not banning AI use in education or work, but recognizing that AI boosts short-term productivity at the cost of fundamental problem-solving skills. The willingness to face and endure difficulties in problem-solving is crucial for developing skills and the capacity to learn new things.
Moreover, the research proposes that AI developers shift from providing direct answers to creating systems that support learning or act as coaches encouraging users to find answers themselves. This approach aims to prevent long-term intellectual decline, aligning with moves by companies like OpenAI to reduce model behaviors that simply conform to users in newer GPT versions.
However, risks remain with the use of autonomous or Agentic AI systems, which operate independently and can produce unexpected errors. Without skills in verification and critical thinking, users relying heavily on AI might be unable to identify or fix AI-generated mistakes. For example, using AI to write system commands without truly understanding them could cause serious system damage.
/source:Wired