Do Models Doubt? (www.psychologytoday.com)

Recent research has sparked a philosophical discourse on the nature of thought and doubt in AI, inspired by René Descartes' method of radical doubt. While large language models (LLMs) like Anthropic's Claude can illustrate complex internal computational processes—referred to as the "J-space" where concepts form before they are articulated—it becomes crucial to differentiate between computation and human cognition. The author highlights that LLMs excel at predicting answers based on learned probabilities but lack the essence of human thought, which is shaped by uncertainty, struggle, and the learning journey itself. This distinction is significant for the AI/ML community as it drives home the idea that the true strength of human cognition lies not just in arriving at correct answers but in the transformative process of grappling with questions. The fragile nature of AI responses suggests that while remarkable, these models do not experience the same intellectual struggles that characterize human learning. As the discourse evolves, the author proposes a reevaluation of Descartes’ ideas, suggesting we could say, "I doubt, therefore I think, therefore I am," emphasizing the importance of the cognitive journey over the endpoint in understanding both AI capabilities and human intelligence.