新智元报道
点开菲尔兹奖得主陶哲轩的博客的评论区，你会感到一种前所未有的震撼——当今世界最顶尖的一批数学家、物理学家、AI圈大佬们齐聚于此，各抒己见，吵得好不激烈。
字里行间，探讨的竟是「数学界的生死存亡」与「人类学者的终极价值」。
而引爆这场世纪大辩论的导火索，是接连几天内发生的极具戏剧性的巨震。
就在这场大吵爆发的前一天，一段关于陶哲轩发声要求「放缓AI发展」的视频，刚刚在全网疯狂刷屏。
在视频中，这位向来拥抱新技术的顶级学者面色凝重，发出了史无前例的严重警告：
「我们必须放缓AI。现在的速度简直疯了，根本没有理由这么快——毫无理由可言！」
他接着说道：「令人惊讶的是，我们在完全不知道后果会怎样的情况下，竟然如此心甘情愿地改变一切。这是一种极端的非线性动态。」
更让人感到一种魔幻现实主义的，是时间线上的冲突——就在刚刚过去的上周末，陶哲轩本人才刚刚提议并大力推进了「SAIR：开源数学模型，以及如何做出贡献」的项目。
上周还在亲自为AI开源数学模型招兵买马，规划蓝图；昨天就突然拉响警报，疾呼「速度疯狂、必须放缓」；今天，他的博客就成了数学界直面AI的决战之地。
短短几天内，陶哲轩态度的激烈碰撞，彻底撕开了当前学术界面临的恐怖现实：延续近千年的数学研究体系和博士培养机制，正面临被AI彻底摧毁的绝境。
为什么陶哲轩会感到如此警惕，甚至认为这是一种「极端的非线性动态」？
因为AI攻占数学界的速度，已经超越了人类心智的承受极限。
在过去，数学研究是人类智慧王冠上的明珠。
一个数学博士，往往需要花费数年时间，在浩如烟海的文献中寻找灵感，为了证明一个小小的引理而绞尽脑汁。如果能做出哪怕一点点「原创性贡献」，就足以顺利毕业，并在学术界谋得一席之地。
但现在，时代变了。
随着AI大模型的极速进化，尤其是自动定理证明器（如Lean）与大语言模型的深度结合，AI已经拥有了令人胆寒的数学发现能力。
极具讽刺意味的是，陶哲轩曾大力推广的协作解题与Lean形式化验证工具，原本是为了方便人类学者交流，如今竟成了机器大军「围剿」人类的顶级武器。成千上万的AI智能体正不知疲倦地在简陋的留言板上协作，以光速探索着数学潜在空间的每一个阴暗角落。
面对不知疲倦、算力无边的机器，要求一个普通的人类数学博士在「证明定理」上去做出原创性贡献，已经变得极其困难。
有悲观者绝望感叹：传统的纯人类数学研究方式已彻底宣告终结。在短暂的过渡期后，一个由机器主导的新时代将悄然降临。而未来的人类学者，或许将可悲地沦为只能去「理解机器发现」的平庸工程师。
25万数学博士面临绝境？数学，这个延续了千年的古老学科，难道真的只剩下让AI疯狂刷榜、「做题」这一层意义了吗？
在这个生死存亡的十字路口，今天冲上热搜的陶哲轩博客文章，犹如漫漫长夜中的一座灯塔。这篇由著名数学科普大V、3Blue1Brown创始人 Grant Sanderson 撰写的客座文章，题目极具「宣战」意味：
《如果数学不仅仅是证明，我们需要更好地颂扬它的其余部分》
这篇文章，为面临AI冲击的数学界，指出了一条全新的自救之路。
Grant Sanderson 在文章开篇就抛出了一个振聋发聩的观点：
「当前数学界回荡着一种共识：解决问题和生成证明，一直以来只是数学家真正目标的『代理指标』。而数学家的真正目标，是增进人类的理解。当AI在没有任何『理解』的情况下也能生成证明时，『证明』作为衡量指标的价值就被彻底削弱了。」
这句话，狠狠扯下了学术界长期以来的遮羞布。
一直以来，学术界重「证明」，轻「阐述」。
你证明了一个千禧年大奖难题，你就是神，就能拿菲尔兹奖；但如果你只是把一个复杂的定理讲得深入浅出，哪怕惠及百万学子，在传统的学术评价体系里，你也只会被打上「搞科普的」标签，算不上硬核学术贡献。
Sanderson 一针见血地指出：AI现在就是一个无情的「证明生成机」。
如果外界认为能写证明的机器让数学家变得多余，那么，数学界就必须站出来，用全新的奖励机制向世界大声宣告——人类数学家的核心价值，是对知识内在联系与清晰性的追求，绝不仅仅是一台证明机器！
为此，Sanderson 正式提出了一个可能改变未来数学界格局的新概念：「动机阐释」。
什么是「动机阐释」？它和冷冰冰的「证明」有什么区别？
在证明中，定义往往放在开头。 传统数学家喜欢直接甩出一个全新的、晦涩的结构，然后开始冷酷地推导它的性质。
在动机阐释中，定义坐在中间。 只有当我们要解决的问题已经被清晰地建立，新的结构才被允许引入。你要告诉读者，我们为什么要发明这个概念？
在证明中，每句话都必须是100%正确的， 每一个主张都必须是前文的必然推论。
在动机阐释中，你可以从一个「虽然简单但却是错误」的想法开始。 这叫作「探索式虚构」。你可以带领读者经历「提出错误想法 -> 发现哪里崩溃 -> 修复问题 -> 发现新问题」的过程。因为，这才是人类大脑认知世界的真实过程！
证明的目的是告诉你：「这个定理是对的。」
而动机阐释的目的是告诉你：「这个定理为什么是对的？它是怎么被想出来的？为什么一开始我们要提出这个问题？它在更宏大的数学宇宙中处于什么位置？」
网友 Mark Miner 在评论区的一段话极其生动地点明了这一点：「100%赞同。如果我六年级的侄女听不懂这个证明，那它就还没有被解释得足够好，不能称之为证明。」
他举了黎曼猜想中Zeta函数的例子：「看看那些可视化图形，你就能凭直觉感受到为什么实部必须是1/2，函数才会在范围内归零。如果小于1/2，线圈变红膨胀；大于1/2，线圈变蓝收缩。你可以像小孩子玩蒙特梭利玩具一样自己调整参数。这才是真正的理解！」
为了让「动机阐释」不再是毫无学术回报的「用爱发电」，Sanderson 在陶哲轩的博客上提出了一个极具颠覆性的倡议：设立「开放式阐释问题」。
这个想法其实早有渊源。
早在2007年，数学家 Timothy Chow 就曾提出过类似的概念：解决一个「开放式阐释问题」，意味着以一种完全透明、清晰的方式解释一个数学主题。每一步都应该有动机；最理想的状态是，让学生觉得「我上我也行」，他们自己也能推导出这个结果。
但当年，这个提议响应寥寥。因为那时候，人类还在为证明定理焦头烂额。
而今天，情况变了。Sanderson 指出：「目前，AI生成的每一个证明，天生就是一个『未解决的阐释问题』。」
想象一下未来几年，AI疯狂吐出成千上万个冷冰冰、毫无逻辑美感、甚至长达几千页的代码级证明。
这就像机器扔给了人类一座座矿山，如果人类不去提炼、不去解释、不去建立概念之间的桥梁，这些「真理」对人类的心智而言将毫无意义。
已故数学大师、菲尔兹奖得主 Bill Thurston 在1993年那篇著名的《论数学的证明与进展》中就写道：
计算机的飞速发展突显了这一点……当Appel和Haken利用海量计算完成四色定理的证明时，引发了巨大争议。
我认为这并不代表人们怀疑定理的真实性，而是反映了人类一种持久的渴望：除了知道定理是真的，我们还渴望通过人类的心智去真正理解它。
为了推动这一转变，Sanderson 提出了几条极具实操性的呼吁：
1.重塑博士答辩： 导师不要只让博士生去解题，可以让他们用AI解题，但最终的考核指标是：你必须举办一场演讲，把这个晦涩的难题或AI证明，给全系的师生讲得明明白白。
2.现代版的希尔伯特问题： 呼吁陶哲轩这样的数学界领袖，发布一份「未解决的阐释问题」清单，告诉全人类，哪些重要的定理我们至今仍是「知其然而不知其所以然」。
3.改革终身教职评定： 将编写优秀的教科书、制作具有高度启发性的学术内容，置于和发表顶级期刊同等甚至更高的地位。
最近，研究者 Liam Price 利用 GPT-5.4 Pro 解决了一直难以攻克的 Erdős 第1196号问题。
然而，最初的AI证明虽然存在，却像天书一样，人类的理解并没有因此推进。
随后，多位人类数学家（包括陶哲轩本人在内）接手了这篇「天书」。他们破译了AI的方法，提取出了核心思想，并将其整理成了人类可读的论文。更重要的是，人类学者阐明了这个核心思想不仅解决了原始问题，还澄清了周围的许多相关猜想。
在这个过程中，AI只是「挖矿机」，而人类数学家才是真正的「珠宝设计师」。 那篇理清脉络、升华思想的人类论文，远比AI输出的原始代码值得庆祝。
陶哲轩的这篇博客一出，X上和评论区立刻化身为大型「学术辩论场」。
最先赶来「嘲讽」加「欢迎」的，是隔壁物理学界的同行。
著名物理学家 Lenka Zdeborova 发了一条火药味与幽默感兼具的推文：
「25年来，我们搞理论物理的总是被告知『你们那不叫数学』，因为我们优先考虑的是『理解』而不是『严格的证明』。现在看到数学界因为AI而产生这样的辩论，我感到非常暖心。突然之间，许多数学家开始解释：数学是关于理解的，而不仅仅是证明。欢迎来到我们的世界！」
有网友附和道：「物理学家和数学家的核心哲学分歧就在于：物理学家认为数学背后有一个现实，数学只是在试图逼近现实；而数学家认为，数学本身就是现实。现在AI把数学高高在上的『神性』打碎了，数学家终于要回归人类的直觉了。」
当然，数学界内部并非毫无保留地接受这种转向。许多学者表达了深深的担忧：
反对声音一：这是对不善言辞的天才的埋没！
有人留言提到已故的数学天才Jean Bourgain。
布尔甘以其粗犷、直接、甚至极其难读的论文风格著称，但他解决问题的能力是核弹级的。
「如果你把评价标准转向『阐述』，那布尔甘这样不善言辞但能证明惊世定理的天才是不是就被埋没了？难道要让未来的布尔甘流失到工业界吗？」
对此，陶哲轩本人亲自下场回应。
他温和却有力地反驳：「我认为布尔甘牺牲阐述而专注于解题，是那个时代奖励机制的『后果』，而不是这套机制合理的『证明』。我私下与他接触过，他完全具备极佳的阐述能力。如果当年的学术环境能给『阐述』更多的权重，他调整写作风格后，对数学界乃至更广泛群体的贡献，绝对比现在还要巨大！」
反对声音二：AI迟早也会掌握「完美阐述」！
还有学者（如Tsimerman）指出，不要太天真了。
AI现在的阐述能力确实一般，但未来大模型必定能生成比人类更完美的教科书级别的「动机阐释」。
如果连「理解和解释」都被AI包揽了，人类数学家还能干什么？
一位匿名网友一针见血地回应了这种「AI万能论」：
「我对某些人对AI的盲目崇拜感到惊讶。什么是『最好的阐述』？这高度取决于试图理解数学的那个特定人类。没有一个静态的『阐述有效性函数』。我们之所以要关注阐述，是为了促进『人类对数学的理解』，而不是『机器对数学的理解』。」
没错，哪怕AI把菜做得再好吃，人类也需要自己咀嚼和消化，才能获得营养。数学不是为了把冰冷的真理堆满服务器的硬盘，而是为了让人类的心智得到升华。
一位提议将「有限单群分类」作为首个「千禧年阐释问题」的学者留言说得令人动容：
「这项证明涉及数千页论文，甚至很多专业数学家都无法完全理解。随着老一代专家逐渐退休，这门手艺有成为『失传手艺』的危险。如果我们在面对未来AI生成的数千页证明时不想束手无策，我们就必须从现在开始，进行大规模的跨界合作，为这些宏大的数学建筑提供『动机阐释』。这是一次人类保卫自身理性的演习。」
回到文章开头，上周还在开发开源数学模型的陶哲轩，昨天为什么要在视频中发出「疯了，必须放缓AI」的严重警告？
因为极端的非线性动态，意味着失控。
陶神的绝望和呼吁，并非因为人类在算力上输给了机器，而是因为我们在文化、体制和心理上，根本没有做好迎接这场大变革的准备。
如果学术界依然死守着「唯证明论」、「唯论文论」的僵化体制，那么在这场AI的海啸中，成千上万的青年学者确实将面临绝境，人类的数学信仰也将受到毁灭性的打击。我们在不知道后果的情况下，正在亲手摧毁延续千年的学术生态。
但是，如果我们按下暂停键，好好审视一下 Grant Sanderson 提出的这场「文化转向」呢？
正如他在文章末尾所言：
「对于现在的年轻学生来说，这其实是进入数学领域最令人兴奋的时代。因为没有比进入一个『具有极强可塑性』、并且你『有机会积极塑造其未来面貌』的领域更令人激动的了。」
当繁重、枯燥的逻辑推演被AI接管，人类数学家终于可以从流水线上的「证明工人」，进化为真正的「数学建筑师」和「思想哲学家」。
我们可以有更多的时间去问「为什么」，去寻找不同领域之间不可思议的联系，去把高深莫测的宇宙密码，翻译成六年级小女孩也能看懂的跳动图符。
正如网友引用的 Scale AI 创始人 Alexandr Wang 的那段话：「正如我告诉我侄子的，数学会让你自由。」
在钻研得足够深的任何事物中，你都会发现数学。数学是连接整个宇宙的网，是物理学定义世界的语言，更是人类在浩瀚无垠的真理之海中，确认自身存在的最美印记。
AI可以递给我们满满一袋子名为「真理」的答案，但面对这无尽的真理之海，唯有深刻的「人类理解」，才能化作我们航行的罗盘。
旧的时代正在崩塌，但一个真正属于「人类心智理解」的数学文艺复兴，或许才刚刚开始。
参考资料：
编辑：Aeneas
不喜欢
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
New Ji-won.
You'll feel an unprecedented shock... The world's top mathematicians, physicists, and the Ai-circles are gathered here.
“The survival of mathematics” and “the ultimate value of anthropologists” are discussed in the words.
And the trigger for this century-long debate was a dramatic shock that took place in successive days.
Just one day before the outbreak, a video about Tao Cheol-hyun's voice calling for "Slow-Ai's Development" had just been razed across the Internet.
In this video, the leading scholar, who has always embraced new technologies, has a strong and unprecedented warning:
"We have to slow down AI. There's no reason to go so fast — there's no reason!
He went on to say: “It is surprising that we have been so willing to change everything without knowing the consequences. This is an extreme non-linear dynamic."
What is more magical about realism is the conflict over time lines — just last weekend, Tao Cheol-hyun himself just proposed and vigorously advanced the project “SAIR: open source mathematical models, and how to contribute.”
Last week, he personally recruited troops for the ABS mathematical model to plan the blueprints; yesterday he suddenly sounded the alarm and shouted "crazy and must slow down"; today, his blog became the final resort for the mathematical community to confront AI.
In just a few days, the violent collision of the attitude of Tao Chul-chul has completely ripped the horrors of the current academic world: The mathematical research system and doctoral development mechanisms that have continued for nearly a millennium are facing the complete destruction of AI.
Why would Tao Chul-hyun be so vigilant, even considering it an "extreme non-linear dynamic"?
Because AI is taking the pace of mathematics, it is beyond the limits of human mind.
In the past, math research was the pearl of the crown of human intelligence.
A mathematical doctorate, often taking years to find inspiration in a vast body of literature, stretches his mind to prove a little subjectivity. It would be enough to make even a little "original contribution" to graduate and get a place in academia.
But now, times have changed.
With the rapid evolution of AZLLM, and especially the depth of the A.J.C. (e.g., Lean) and the mega-linguistic model, AI already has a chilling mathematical discovery capability.
It is extremely ironic that the collaborative resolution of this issue, which was promoted by Tao Chul-hyun and the Lean formalization certification tool, which was designed to facilitate exchanges among human scholars, has now become the most powerful weapon of the machine army in “siege” humanity. Hundreds of thousands of AZAI agent are working tirelessly on poor message boards to explore every dark corner of potential mathematical space at light speed.
It has become extremely difficult to demand the original contribution of an ordinary human mathematics doctor on the "prove theorem" in the face of an unwieldy and uncompromising machine.
Pessimists lament that traditional purely human mathematical research has been completely completed. After a brief transition period, a new age of machine-led power will be looming. And future humanists, perhaps sadly, will be reduced to ordinary engineers who can only "understand the discovery of machines."
A quarter of a million mathematicians face a desperate situation? Mathematics, an ancient subject that lasts for thousands of years, is it true that there's only one thing left for AI to say, "do the question"?
At this crossroads of life and death, today's hot search of Tao Cheol's blog article is like a lighthouse in a long night. This is a visiting article by Grant Sanderson, founder of the famous Big V in mathematics and 3 Blue1 Brown, entitled "A declaration of war":
If math is not just proof, we need to celebrate the rest of it better.
This article points to a completely new path of self-help for the mathematical community facing the impact of AI.
Grant Sanderson, at the beginning of the article, casts a deafening view:
"There is now a consensus in the mathematics world that solving problems and generating proof has been a proxy indicator for the real goal of mathematicians. And the real goal of mathematicians is to improve human understanding. When AI can produce proof without any "understanding", the value of "prove" as a measure is completely diminished."
This sentence is a long-standing veil of academia.
In the past, the academic community has focused on "prove".
You've proved a millennium prize dilemma, and you're a god, and you can get a Fields prize; but if you're just telling a complex theorem to the ground, even for the benefit of millions of students, in a traditional academic evaluation system, you're only labeled as a "kipping" and not a hard-core academic contribution.
Sanderson points out with blood: AI is now a ruthless "proven generator".
If the outside world believes that proof machines make mathematicians redundant, then the mathematics community must stand up and speak out to the world with a brand-new incentive mechanism — the core value of human mathematicians is the quest for the intrinsic connection and clarity of knowledge, not just a demonstration machine!
To this end, Sanderson has officially proposed a new concept that could change the future shape of the mathematical community: the interpretation of motivation.
What's "motive interpretation"? What's the difference between it and the cold proof?
In certificates, definitions are often placed at the beginning. Traditional mathematicians like to throw out a completely new, obscure structure and then begin to infer its nature in cold blood.
In the interpretation of motive, the definition sits in the middle. Only when the issues we have to address are clearly established will the new structure be allowed to be introduced. You're going to tell readers why we invented this concept?
In proof, every word has to be 100% correct, and every claim has to be the corollary.
In the interpretation of motivation, you can start with the idea of "simply but wrong." This is called "exploration fiction." You can lead readers through the process of "pulling the wrong idea-- finding where to crash-- fix the problem-- finding the new problem." Because that's the true way the human brain understands the world!
The purpose of the proof is to tell you: "This is the truth."
And the purpose of the motivation is to tell you: "Why is this theorem right? How did it come out? Why did we start with this question? Where is it in the larger mathematical universe?"
A commenter, Mark Miner, in an extremely vivid comment section, makes this clear: "100% agree." If my sixth-grade niece does not understand this proof, it has not yet been explained well enough to be a proof.”
He gave the example of the Zeta function in Lehman's guess: "Look at the visualized graphics, you can feel intuitively why the real thing has to be one second, and the function is zero. If less than 1/2, the coil becomes red-inflated; greater than 1/2, the coil becomes blue-seated. You can adjust your parameters like a kid playing Montsolly. That's the real understanding."
In order to stop the "motive interpretation" from being a "power generation with love" with no academic reward, Sanderson has launched a highly subversive initiative in his blog: "Open interpretation."
It's a long-standing idea.
As early as 2007, Mathematician Timothy Chow had proposed a similar concept: solving an "open interpretation problem" which means interpreting a mathematical theme in a completely transparent and clear manner. Each step should have a motive; the ideal is to make students feel "I'm on my side" and they can produce it themselves.
However, that proposal had received little response in that year. Because at that time, humans were still trying to prove that theorem was a mess.
Today, the situation has changed. According to Sanderson, "at present, every certificate produced by AI is born an "unsolved problem of interpretation".
Imagine that in the coming years, AI's madness has revealed thousands of cold, unlogical, thousands of pages of code-level proof.
It's like a machine throws at humans a mine, and if humans don't refine, explain or build bridges between concepts, these "truths" will be meaningless to human minds.
Bill Thurston, the late Mathematics Master and Fields Prize winner, wrote in the famous 1993 edition of The Certificate and Progress of Math:
The rapid development of computers highlights this point... When Appel and Haken used mass calculations to complete the proof of the Four Color Theorem, a huge controversy arose.
I do not think this is a question of the truth of the theorem, but rather of a long-standing desire of humanity: In addition to knowing that theorem is true, we aspire to truly understand it through the human mind.
To drive this change, Sanderson made several very real calls:
1. Reshaping the Ph.D. ' s response: While the mentors should not leave the doctoral students alone to solve the problem, they can use the AI to solve it, the final test indicator is that you have to give a speech in which this obscure problem or AI proves that it is clear and clear to the entire faculty.
2. The modern version of the Hilbert question: Call on Mathematicians such as Tatzer, publish, a list of "unsolved problems of interpretation ", to tell all humanity that we still have the important theorems that we know but do not know why.
3. Reforming the evaluation of tenure: to place the production of good textbooks and highly instructive academic content on an equal or higher footing with the publication of top-level periodicals.
Recently, the researcher Liam Price used GPT-5.4 Pro to solve the problem of Erdős 1196, which has been difficult to overcome.
However, the original AI proved that, while it existed, human understanding, like the book, did not advance.
Subsequently, several human mathematicians (including Tao Cheol-hyun himself) took over the book. They deciphered AI's methods, extracted core ideas and organized them into human-readable papers. More importantly, anthropologists have articulated this core idea not only as a solution to the original problem, but also as a clarification of many of the relevant assumptions around it.
In the process, AI is just a miner, and human mathematicians are the real jewel designers. That human paper, which clears the context and enhances ideas, is much more to celebrate than the original code of AI output.
As soon as the blog of Tao Chul-Hyun was published, X-ray and the comment area immediately became a large "academic debate forum".
The first person to come to "screech" and "welcome" was a colleague from the physics community next door.
The famous physicist Lenka Zdeborova tweets a gunpowder and a sense of humor:
"For 25 years, we have been told that your theoretical physics is not math, because we give priority to understanding, not strict proof." Now I'm very excited to see that the math community is having this debate over AI. Suddenly, many mathematicians began to explain: math is about understanding, not just proof. Welcome to our world!"
"The core philosophical difference between physicists and mathematicians is that physicists believe that there is a reality behind mathematics, that mathematics is just trying to get closer to reality, and mathematicians think that mathematics is itself a reality. Now AI has broken the "godly" of mathematics, and the mathematicians are finally returning to human instincts."
Of course, this shift is not accepted unreservedly within the mathematics community. Many scholars expressed deep concern:
Opposed to sound one: This is the burial of unspoken geniuses!
A message was sent about the late math genius Jean Bourgain.
Burgan is known for his crude, direct and even extremely difficult thesis style, but his ability to solve problems is nuclear-grade.
"If you turn the evaluation criteria to the `explaining', does a genius like Burgan, who doesn't speak nicely, but can prove that theorem of the world is buried? Should the future of Burgan be lost to industry?"
In this regard, Tao Chul-hyun himself responded.
"I think Burgan's sacrifice and his focus on solving the problem is the `effects' of the incentive mechanisms of that era, not the `proves' of the system. I have had private contact with him, and he is well equipped to speak. If the academic environment of the year had given more weight to the `description', his contribution to mathematics and the wider community would have been greater than it is now."
Opposed voice two: Sooner or later, AI will have the perfect narrative.
Some scholars (e.g. Tsimmerman) have pointed out that not too naive.
AI does have the same narration skills, but the future of LLM is bound to produce a "motive interpretation" at a level more perfect than humans.
What else can human mathematicians do if even "understanding and explain" are covered by AI?
An anonymous netizens responded with blood:
"I am surprised that some people are blindly worshipping AI. What's the best description? It depends highly on the particular human being who tries to understand math. There is no static `description function'. We're concerned about this to promote "human understanding of mathematics" rather than "machine understanding of mathematics".
Yes, even if AI cooks the food better, humans need to chew and digest themselves to obtain nutrition. Mathematics is not intended to fill the server with cold truths, but rather to enhance the human mind.
One of the authors who proposed the term “limited taxonomy” as the first “jubilee interpretation issue” is moving:
“This certificate involves thousands of pages of paper, which even many professional mathematicians cannot fully understand. As the older generation of experts gradually retires, this trade risks becoming a "missing trade". If we do not want to be helpless in the face of the thousands of pages of proof generated by the future AI, we must now engage in large-scale cross-border cooperation to provide a `motive interpretation' of these ambitious mathematical architectures. This is a human exercise to defend its own reason.”
Back at the beginning of the article, and last week, Tao Cheol-chul, who was developing an open source mathematical model, why did yesterday send a serious warning in the video that "Crazy, I have to slow down"?
Because extreme non-linear dynamics mean out of control.
The despair and appeal of the pottery are not due to the loss of human energy to the machine, but because we are, culturally, institutionally and psychologically, not ready for this great change.
If the academic community remains in a rigid system of “provenism” and “discretionaryism”, it is true that in the tsunami in AI thousands of young scholars will face a desperate situation and that human mathematical beliefs will be devastated. Without knowing the consequences, we are hand-in-hand destroying the academic ecology that continues for thousands of years.
But what if we push the pause and look at this "cultural shift" that Grant Sanderson proposed?
As he said at the end of the article:
"This is actually the most exciting time for young students today in mathematics. For there is nothing more exciting than entering an area of "extreme plasticity" and your "opportunity to shape its future."
When heavy, boring logic was taken over by AI, human mathematicians finally evolved from "prove workers" on the waterline to "mathematic architects" and "ideological philosophers".
We can have more time to ask "why," to look for incredible links between different fields, to translate the inexplicable cosmological code into a sixth-grade girl who can read the beats.
As the netizens quoted from the Scale AI founder, Alexandre Wang, said, "Mathematics will set you free, as I told my nephew."
You find math in anything that's deep enough. Math is the web that connects the entire universe, the language of physics that defines the world and, above all, the most beautiful mark of human existence in the vast sea of truth.
AI can give us a bag full of answers called "the truth," but in the face of this endless sea of truth, only a deep "human understanding" can become our compass.
The old age is falling, but a mathematical renaissance that really belongs to "human understanding of the mind and mind" may have just begun.
References:
Editor: Aeneas
I don't like it.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.