Old and new apps, via modern coding agents by Terry Tao
453 points
• 5 days ago
• Article
Link
Terence Tao 长期使用交互式小程序作为数学研究和教学的可视化辅助工具。上世纪 90 年代末，他手工编写了几个 Java applets 来展示诸如 Besicovitch sets 和 honeycombs 等复杂数学对象。但随着网络标准演进、对原始 Java 的支持逐渐消失，这些工具也相应失效。最近，Tao 启动了一个项目，把自己的网站内容迁移到更可持续的存储库，并借助现代 AI 来恢复这些已过时的程序。
迁移过程中，他委托一个 AI 编程代理将原始 Java applets 移植到 JavaScript 。这一做法非常奏效：该代理在几小时内就恢复了所有旧 applets 。更新版本在图形上有所改进，例如增加了着色功能，代理还发现并修复了原始代码中一些未被注意到的错误。 Tao 指出，尽管基于 LLM 的编码代理偶尔会引入错误，但对于此类非关键的可视化辅助工具来说，风险很小。
在此成功的基础上，Tao 将项目扩展到一些过去手工难以实现的新应用。他借助 AI 实现了长期搁置的设想——为 Minkowski space 制作一个 special relativity 的可视化工具；并为他最近关于 Gilbreath's conjecture 的论文开发了新的交互式可视化。这样的"vibe coding" 体验效率极高，通常只需与 AI 交互数小时，便能产出功能完备、质量上乘的工具。
能够轻松生成此类交互式内容，促使 Tao 计划在未来的学术论文中更多地整合类似可视化。由于这些应用主要作为辅助展示，而非正式数学论证的关键部分，AI 生成代码的当前局限并不会构成重大障碍。展望未来，这些工具为数学研究的呈现与探索开辟了新前沿，为读者提供更直观、更具交互性的体验。
Terence Tao has long utilized interactive applets as visual aids for his mathematics research and teaching. In the late 1990s, he manually programmed several Java applets to visualize complex mathematical objects like Besicovitch sets and honeycombs. However, these tools eventually became non-functional as web standards evolved and support for the original Java version vanished. Recently, Tao began a project to migrate his web presence to a more sustainable repository, using modern AI assistance to revive these obsolete tools.
The migration process involved tasking an AI coding agent with porting the original Java applets into JavaScript. This approach proved highly successful, as the agent managed to revive all of the old applets in only a few hours. The updated versions even feature graphical improvements, such as colorization, and the agent successfully identified and fixed bugs in the original code that had previously gone unnoticed. Tao notes that while LLM-based coding agents can occasionally introduce bugs, the risk is minimal for these types of non-critical visual supplements.
Building on this success, Tao expanded his project to develop new applications that were previously too complex to build by hand. He successfully used AI to complete a long-abandoned vision for a special relativity visualization tool designed for Minkowski space. Additionally, he created a new interactive visualization to accompany his recent paper on Gilbreath's conjecture. This "vibe coding" experience was remarkably efficient, requiring only a few hours of interaction with the AI to produce functional, high-quality tools.
The ability to easily generate such interactive content has encouraged Tao to integrate similar visualizations into his future academic papers. Because these apps serve as supplementary aids rather than essential components of a formal mathematical argument, the current limitations of AI-generated code do not present a significant hurdle. Looking ahead, these tools represent a new frontier in how mathematical research can be presented and explored, providing a more intuitive and interactive experience for readers.
133 comments • Comments Link
• 使用大型语言模型（LLMs）构建计算机科学教学的交互式可视化工具被证明非常高效，使教师能够创建以前因耗时而难以实现的复杂演示。
• 受人尊敬的专家如 Terence Tao 等人采用 AI 编码代理，表明专业工作流程正在发生重大转变：即便是世界级研究者，也在将这些工具用于辅助性任务，尽管它们目前仍有局限。
• "业余项目"（hobby projects）与"严肃工作"（serious work）之间的界限正在演变。即便主要研究仍然独立于 AI 产生，那些能实现快速原型或制作补充材料的工具，仍起到了倍增器的作用。
• 关于对 AI 的"信任"（trust）的讨论最好视为工作流程设计中的一项挑战。用户需要学会识别 AI 擅长的领域（例如生成仪表板或样板代码）以及它的短板，而不是简单地全盘否定这项技术。
• 虽然有人将 LLMs 视为仅仅是"随机鹦鹉"（stochastic parrots），认为它们缺乏真正的自主性，但也有人指出，这种还原论忽视了 AI 在处理抽象复杂任务方面的效用——类似于人类通过高层抽象进行认知的方式。
• 用 AI 生成的 Web 界面来现代化教学内容，尽管存在取代旧有专业解决方案（如 legacy Java applets）的风险，但这类进步最终提升了学生的可及性与互动性。
• 人们对 AI 倡导者的中立性抱有疑虑：有人质疑知名研究者是否存在未公开的动机，或者他们是否仅处于对新兴技术的长期"蜜月期"（honeymoon phase）。
• 仍有人担心所谓的 Gell-Mann amnesia effect，即人们在自己专业领域之外会过度信任 AI 的输出，而忽视其中的细微错误和表面化问题。
• 在 AI 辩论中，数学（Mathematics）占据独特地位。由于纯数学依赖形式逻辑而非感官数据，一些人认为，相较于需要实证验证的领域，AI 在数学领域最终可能展现出更强的能力。
• 非技术领域对软件开发存在巨大的潜在需求。 AI 降低了那些需要工具来自动化任务或可视化概念的专家的准入门槛，有效地将领域专家变成"公民开发者"（citizen developers）。
此次讨论反映出，包括数学与计算机科学在内的各领域专家在将 AI 整合进专业实践时所经历的更广泛转型。尽管关于 LLMs 可靠性的明确共识尚难达成，大家普遍认为这些工具在快速原型设计、可视化以及其他劳动密集型任务中具有显著价值。参与者既承认过度依赖的风险和生成内容中可能存在的细微错误，又强调通过构建以严格人工审核为特征的工作流程可以缓解许多担忧。归根结底，这场讨论凸显了 AI 辅助生产力带来的即时务实收益，与维持准确性和有效性所需的领域专业知识之间的张力。 • Using LLMs to build interactive visualizations for computer science education has proven highly productive, allowing instructors to create complex demonstrations that were previously too time-consuming to develop.
• The adoption of AI coding agents by respected experts like Terence Tao suggests a significant shift in professional workflows, where even world-class researchers utilize these tools for supplementary tasks despite their current limitations.
• Distinctions between "hobby projects" and "serious work" are evolving. Tools that enable rapid prototyping or the creation of supplementary materials function as force multipliers, even if the primary research remains independent of AI generation.
• Debates regarding AI "trust" are better framed as a challenge in workflow design. Users must learn to identify where AI excels—such as generating dashboards or boilerplate code—and where it fails, rather than dismissing the technology entirely.
• While some view LLMs as mere "stochastic parrots" incapable of true autonomy, others argue that this reductive view ignores the utility of AI in abstracting complex tasks, similar to how human cognition functions through higher-level abstraction.
• The modernization of educational content through AI-generated web interfaces risks displacing older, specialized solutions like legacy Java applets, though this progress ultimately improves accessibility and interactivity for students.
• Skepticism exists regarding the neutrality of AI advocates, with some questioning whether prominent researchers might have undisclosed incentives or are simply experiencing a long-term "honeymoon phase" with emerging technologies.
• Concerns persist about the "Gell-Mann amnesia effect," where individuals trust AI outputs in fields outside their own expertise while failing to notice the subtle errors and superficiality in the generated content.
• Mathematics occupies a unique space in the AI debate. Because pure math relies on formal logic rather than sensory data, some believe it is an area where AI will eventually demonstrate superior capabilities compared to fields requiring empirical verification.
• There is significant latent demand for software creation in non-technical fields. AI lowers the barrier to entry for experts who need tools to automate tasks or visualize concepts, effectively turning domain specialists into "citizen developers."
The discussion reflects a broader transition in how experts across various fields, including mathematics and computer science, integrate AI into their professional practices. While a clear consensus remains elusive regarding the reliability of LLMs, there is general agreement that these tools provide significant utility for rapid prototyping, visualization, and tasks that are otherwise labor-intensive. Participants acknowledge the risks of over-reliance and the potential for subtle errors, yet they emphasize that building effective workflows—characterized by rigorous human review—mitigates many of these concerns. Ultimately, the conversation highlights a tension between the immediate, pragmatic benefits of AI-assisted productivity and the critical need for domain expertise to maintain standards of accuracy and validity.