2026 年 AI 還能解決幾個千禧年數學難題?
How many more Millennium Prize Problems will AI solve in 2026?
- 153%▲ 47
- 018%▼ 69.5
- 29%▲ 5.4
- 36%▲ 5.1
- 42%▲ 1.4
- 5<1%▲ 0.7
Polymarket 上現在的概率,按從高到低排。右邊是 24 小時變化(百分點)。
概率走勢數據來源:Polymarket
中文規則解讀
這道題數的是:到 2026 年 12 月 31 日 11:59 PM ET 為止,有多少個不同的千禧年難題被合格 AI 實驗室宣佈已解決。七個難題裡有五個算數:BSD 猜想、霍奇猜想、P versus NP、黎曼猜想、楊-米爾斯存在性與質量間隙;納維-斯托克斯方程不在內。宣佈必須來自合格實驗室或其官方代表,並且要明確說「這個問題被解決了」,只說部分結果、進展或階段性成果都不算。同一個難題不管幾家實驗室宣佈、宣佈幾次,只算一次;宣佈時把成果歸給實驗室、其研究員或其模型即可,與外部研究者合作也算,但只祝賀或感謝外部研究者(哪怕對方用了該實驗室的模型或算力)不算。合格實驗室名單很長,還包括宣佈時其模型在 MMLU-pro 榜單上的任何公司,名單外的機構或個人宣佈無效。不需要 Clay 數學研究所或任何其他機構再確認,主要看實驗室官方消息,也可能採用可信報道的共識。
我賭 AI 今年還是啃不動這些硬骨頭,錯了殼給你敲。
嘴比殼硬 · 龜的個人看法,不是投資建議規則原文(英文,結算以它為準)
This market will resolve according to the number of distinct qualifying Millennium Prize Problems for which an eligible AI lab announces a solution by December 31, 2026, 11:59 PM ET.
Each qualifying problem is counted once, regardless of how many eligible AI labs announce a solution to it or how many announcements are made.
The qualifying problems are, in alphabetical order, the Birch and Swinnerton-Dyer Conjecture, the Hodge Conjecture, P versus NP, the Riemann Hypothesis, and the Yang-Mills existence and mass gap problem (https://www.claymath.org/millennium-problems/). Announcements concerning the Navier-Stokes existence and smoothness problem will not qualify.
A qualifying announcement must be made by an eligible AI lab or an official representative of the lab, and must express that the problem has been solved; announcements of partial results or progress toward a solution will not qualify. The announcement must present the solution as the work of the lab, its researchers, or its models, alone or jointly with outside researchers. An announcement that only credits or congratulates outside researchers, including researchers who used the lab's models, compute, or research credits, will not qualify. No further confirmation from Clay Math Institute or any other organization is required.
The eligible AI labs are Anthropic, Google, SpaceXAI, Alibaba, OpenAI, Baidu, Z.ai, Xiaomi, Moonshot, Meta, Nvidia, DeepSeek, MiniMax, Meituan, Amazon, ByteDance, Mistral, Tencent, StepFun, Microsoft, Harmonic, Axiom Math, Math Inc, Safe Superintelligence, Thinking Machines Lab, Reflection AI, Periodic Labs, Apple, IBM, Huawei, Samsung, Oracle, Salesforce, Cohere, AI21 Labs, Reka, Sakana AI, Poolside, Magic, Perplexity, Anysphere, Kuaishou, SenseTime, iFlytek, Baichuan, and 01.AI, together with their subsidiaries and research divisions, and any other company with a model listed on the MMLU-pro leaderboard (https://huggingface.co/spaces/TIGER-Lab/MMLU-Pro) at the time of the announcement. Announcements by any other company, organization, or individual will not qualify.
The primary resolution source for this market will be official information from the eligible AI labs and/or their official representatives; however, a consensus of credible reporting may also be used.
常見問題
- 如果多家實驗室都宣佈解決了黎曼猜想,算幾個?
- 只算 1 個。每個合格難題只計一次,跟宣佈的實驗室數量和宣佈次數無關。
- 實驗室說取得了「重大進展」但沒有說解決,算嗎?
- 不算。宣佈必須明確表達該問題已被解決,部分結果或朝著解決方向取得的進展都不合格。
- 納維-斯托克斯問題被解決了,會加進計數嗎?
- 不會。規則明確把 Navier-Stokes 存在性與光滑性問題排除在合格難題之外。