王垠：不懂编程的人想用 AI 成功做项目，几乎是妄想！

Core Viewpoint - The article emphasizes that while AI tools can assist in programming, they cannot replace the need for deep understanding and expertise in computer science to produce high-quality code [3][5][6]. Group 1: AI Programming Experience - The author shares experiences with various AI programming tools, highlighting that non-programmers are unlikely to successfully complete projects using AI alone [3]. - Despite generating over 60,000 lines of code with Cursor, the author only accepted around 5,000 lines due to frequent errors and misunderstandings by the AI [3]. - The complexity of AI-generated code often leads to confusion and the need for extensive corrections, indicating that AI struggles with even simple functions [3][5]. Group 2: Limitations of AI - The article argues that AI cannot produce "high-level" code due to the mediocre quality of the training data it relies on, which primarily consists of average programmers' work [5]. - The distinction between computer scientists and programmers is emphasized, with the assertion that AI may replace ordinary programmers but cannot substitute for computer scientists [5]. Group 3: AI as a Tool - AI is viewed as a valuable tool that can enhance productivity, but it is limited to amplifying existing skills rather than compensating for a lack of knowledge [6]. - The author acknowledges the benefits of using AI for problem-solving but stresses that without foundational skills, AI's effectiveness is nullified [6].