Academics only.
For decades, a key question about the challenges of high-dimensional computation resisted proof. New research resolves the KLS conjecture, removing a major barrier to understanding the efficiency of randomized algorithms used throughout science and computing.
msft.it/6011aar01
我觉得 AI 能解出 31 年的数学难题，真正的功臣是那个负责砍掉错误方向的人。
微软研究院刚宣布证明了 1995 年提出的 KLS 猜想，高维几何里的核心难题。过程是这样的：
研究员让多个 AI agent 并行探索，一共跑了 100 多条路线。每一轮人看中间结果，砍掉没希望的，把最好的那条交给下一轮。
最扎心的是另一组对照：差不多的 token 预算，让几家顶尖模型不加引导地反复去解，结果都没成功 🤡
同一天另一组专家也发了独立证明，后来还有 Lean 4 形式化验证。
所以 AI 不缺算力，缺的是一个知道往哪砍的人。你用 AI 干活，是在剪枝还是在许愿？
For decades, a key question about the challenges of high-dimensional computation resisted proof. New research resolves the KLS conjecture, removing a major barrier to understanding the efficiency of randomized algorithms used throughout science and computing.
msft.it/6011aar01
This is EXACTLY why this is the beginning, not the end, of math.
Those who will make 100x the typical human impact in the future will be whose who continue to use 100% of their own cognitive effort and creativity, and use AI to amplify their own efforts.
For decades, a key question about the challenges of high-dimensional computation resisted proof. New research resolves the KLS conjecture, removing a major barrier to understanding the efficiency of randomized algorithms used throughout science and computing.
msft.it/6011aar01
Resolving KLS matters less as a trophy proof and more as a runtime boundary: when randomized algorithms in high-d science actually scale—and when they quietly don't.
For decades, a key question about the challenges of high-dimensional computation resisted proof. New research resolves the KLS conjecture, removing a major barrier to understanding the efficiency of randomized algorithms used throughout science and computing.
msft.it/6011aar01