AI learns from the products of thinking, not thought itself (theconversation.com)

Recent research highlights a critical gap in AI systems: they generate outputs based on the "traces" of thought rather than engaging in the creative processes that humans undergo. This limitation suggests that current AI models, while capable of producing polished results, lack the ability to learn from errors and the organic friction that often enhances human creativity. The study emphasizes that without incorporating these processes—such as revising responses or learning from mistakes—AI applications could hinder not only their developed outputs but also the skills of the humans who utilize them. Significantly, the research draws parallels between the shortcomings of AI and a broader phenomenon observed in fields like aviation and medicine, where reliance on AI tools has led to a degradation of essential skills. For instance, pilots' and clinicians' abilities have declined due to overdependence on automated systems. The authors propose that effective AI should integrate more human-like processes, such as encouraging critical thinking and maintaining a focus on the learning journey instead of simply providing answers. By doing so, AI could enhance both its capabilities and the cognitive resilience of its human users, ultimately fostering a more symbiotic relationship between technology and human intelligence.