提示与他者心智的问题：我们是否真正了解模型的“理解”？

Interactions with AI models often yield two contrasting types of responses. On one hand, the model may provide answers that perfectly align with the user's intentions, as if it can peer into their mind. On the other hand, similar prompts can result in baffling or incorrect outputs. This phenomenon raises questions about the 'problem of other minds,' particularly in the context of AI. Philosophically, this issue explores how we can ascertain that others have inner experiences akin to our own. While we can observe behaviors and hear words, direct access to consciousness remains elusive. Compared to humans, AI's understanding is even more complex, as models lack consciousness to access, relying instead on patterns and weights. This creates greater uncertainty regarding the model's comprehension capabilities.