最近deepseek R1大火，年夜饭上父母都在和我讨论大语言模型的未来。我自己使用几次的体验下来，我感觉很惭愧——不是AI的能力太差，而是我很难提出一个好问题🥺我经常找不到一个合适地描述自己所遇到的困境的比喻或描述，不知道怎么用足够的提示词来精准检索我所需要的知识。AI就像亚瑟王的宝剑插在石头里，我却拔不出来🥲 我第一次感到自己原来是这样地无能，也是AI的发展才让我意识到，很多时候我学习连自己的目标都搞不清楚，就搁那闷头学，怪不得学习效率低。 另一个困境是，面对AI这样称手的学习工具，困难不是自己不知道知识，而是不知道自己“不知道”本身。有些知识我都不知晓它的存在，更别提让AI教会我了。最令我恐惧的是我都不知道自己无知，R1这种思维链模型出现后能够自主帮我完善提示词，我才意识到自己认知的局限。它不仅能告诉我不知道的，还能告诉我不知道自己不知道的。突然对自己有了新的认识。
Recently, Deepseek R1 has become extremely popular. During the New Year’s Eve dinner, my parents were discussing the future of large language models with me. From my own experience of using it several times, I feel quite ashamed – not because the AI’s capabilities are too poor, but because it’s difficult for me to come up with a good question 🥺. I often can’t find an appropriate metaphor or description to depict the difficulties I encounter, and I don’t know how to use enough prompt words to accurately search for the knowledge I need. The AI is like King Arthur’s sword stuck in the stone, but I can’t pull it out 🥲. For the first time, I felt so incompetent. It was the development of AI that made me realize that often when I study, I don’t even clearly understand my own goals. I just keep studying blindly, no wonder my learning efficiency is low. Another difficulty is that when facing such a handy learning tool as AI, the problem is not that I don’t know the knowledge, but that I don’t know what I “don’t know” itself. There is some knowledge that I’m not even aware of its existence, let alone asking the AI to teach me. What scares me the most is that I don’t even know my own ignorance. After the emergence of a thought-chain model like R1, which can autonomously help me refine the prompt words, I realized the limitations of my cognition. It can not only tell me what I don’t know, but also what I don’t know that I don’t know. Suddenly, I have a new understanding of myself.