智源研究院學術副院長，清華大學教授唐傑發佈悟道2.0成果丨智源大會

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2021 年 6 月 1 日，由"},{"type":"link","attrs":{"href":"https:\/\/www.baai.ac.cn\/","title":"","type":null},"content":[{"type":"text","text":"北京智源人工智能研究院"}]},{"type":"text","text":"主辦的"},{"type":"link","attrs":{"href":"https:\/\/2021.baai.ac.cn\/","title":"xxx","type":null},"content":[{"type":"text","text":"2021 北京智源大會"}]},{"type":"text","text":"在中關村國家自主創新示範區會議中心開幕，會期三天。智源大會是北京創建全球人工智能學術和創新最優生態的標誌性學術活動，本文爲 InfoQ 記者帶來的現場報道。"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"以下內容根據唐傑的演講速記進行整理，未經本人確認。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“悟道”大規模智能模型取得多項國際領先的AI技術突破和多個世界第一，同時也做出了多項行業貢獻。建設並開放了全球最大的中文語料數據庫WuDaoCorpora2.0，提出了自然語言評測新標準“智源指數”，包含6種主要語言、30餘項主流任務與相關數據集，形成了大模型“權威考卷”。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/5a\/09\/5a39474d96b6d3ec00eeeeb44ed30409.jpeg","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"生命模型，有賴於腦科學特別是神經科學的進步，首先要對大腦進行高精度解析，再在大算力的支持下進行高精度仿真模型。針對生命模型，智源研究院成立了生命模擬研究中心，負責開發高精度生命模擬仿真平臺“天演”，構建和探索感覺、知覺、記憶等智能模型，爭取成爲新一代人工智能發展的不竭源泉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"生命模型本身就是腦科學研究的中心任務，是探索人類自身、解密智能成因的科學探索需要。“天演”生命模型系統已經開展斑馬魚全腦、靈長類視覺、人類心臟、人類認知等模型構建，推動生命科學、醫療健康和人工智能等前沿科研探索和應用發展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最後一條路線的挑戰最大，核心是構建地球乃至宇宙的物理模型，培育新一代自主人工智能，應對各類自然挑戰。智源已經啓動了“天道”物理模型建設步伐，從基本粒子模擬開始，已建立了“分子模擬研究中心”，突破了分子模擬“快”與“準”難以兼顧的瓶頸，取得重要突破，將在未來五年內打造全球最具影響力的材料數據庫，推動人工智能驅動的新一代分子動力學模擬邁上新臺階。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"上述三條路徑的核心任務分別是構建信息模型、物理模型和生命模型，共同需求是算力。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲建設超大規模智算平臺，北京市專門安排了專項資金，建設體系架構先進、高速互聯互通、可擴展高效並行的軟硬件系統。在強大的智算體系支持下，三大模型性能快速提升，共同支撐新一代人工智能模型的建立，進而賦智各行各業，賦智萬物。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"智源研究院經過兩年多探索，已經確定將打造智能模型作爲源頭創新的核心任務，在機構設置、資源保障和機制體制上進行了系統部署，全力構建信息模型、生命模型和物理模型“智能三劍客”。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"簡單總結來說，首先是“悟道”信息模型：數據是智能的素材，通過“大數據+大算力+強算法”，訓練超大規模信息模型，爲各行各業智能化提供切實可用的共性基礎智能模型，是當前及未來一段時間人工智能全面賦智經濟社會發展的中心任務。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"然後是“天演”生命模型：生命是智能的第一載體，模擬仿真億萬年進化而成的生物神經系統和身體，構建生命智能模型，挖掘生物智能機制機理，是發展新一代人工智能的不竭源泉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最後是“天道”物理模型：自然是生命的搖籃，地球是人類的搖籃，構建地球環境物理模型，是實現類人智能和人機共處共融的必要條件。建立宇宙物理模型，發展強人工智能，推動智能走出地球、穿越時空。本次智源大會，各位將親眼目睹這三大模型的最新進展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二方面的工作是建設學術生態。智源從成立之初，就開始遴選智源學者，建設智源社區，積極打造全球人工智能學術和技術創新生態。“智源學者計劃”主要支持頂尖科學家開展人工智能重大方向的基礎前沿研究，支持青年科學家開展開放性、探索性研究。包含智源首席科學家、智源研究員、智源青年科學家等六個層次。已經圍繞六大研究方向，遴選上百位智源學者，形成了AI科研的“北京力量”。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面介紹一下六大重點研究方向的目標及工作進展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“人工智能的數理基礎”方向旨在建立以數學和統計理論爲第一原理的新一代人工智能方法論。已取得了一系列成果。深度勢能團隊首次採用智能超算與物理模型結合，將原來可能需要60年才能完成的分子動力學模擬任務縮短到了1天，成果獲得國際高性能計算應用領域最高獎——戈登貝爾獎。應用層面，團隊提出的MOML算法已應用在冬奧賽場溫度和風速預測，預測準確率大幅提高。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“人工智能認知神經基礎”方向旨在將認知科學、神經科學和計算科學交叉融合，爲人工智能的架構和功能探索新的模型和發展思路。目前正在建立全球規模最大的高精度生物智能開源開放平臺及模擬系統。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“機器學習”方向旨在解決現有機器學習算法存在的可解釋性缺失、大數據依賴、模型場景化等問題，開展“用戶友好的AI”研究。已經取得了階段性進展。黃高團隊提出的高效空間自適應推理，可實現高效的動態特徵提取。朱軍團隊提出第三代安全可控人工智能，研發出全球首個針對算法模型本身進行安全檢測的技術平臺RealSafe，提供從安全測評到防禦加固的整體解決方案。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“智能體系架構與芯片”方向旨在從整機、編程、芯片、方法四個方面探索智能計算系統的發展。提出了“類腦計算完備性”及計算系統層次結構，相關工作發表於Nature。探索出分形馮·諾依曼體系、支持CNN+RNN的混合神經網絡處理器的新型架構。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“智能信息檢索與挖掘”方向旨在建立智能信息檢索與挖掘領域的“北京學派”，已經研發並上線了“交互式智能政務助手”，支持了60多個北京政務應用。針對信息檢索的可解釋性和公平性，提出雙穩健排序等模型，連續兩年在ACM SIGIR會議上貢獻了約10%的學術論文。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“自然語言處理”方向旨在提出“大數據+富知識+跨模態”共同驅動的新範式。在中文預訓練模型、知識圖譜、文本生成等領域，取得了一系列具有國際影響力的學術成果。研發出最大的中文預訓練語言模型“悟道·文源”，研發的神經機器翻譯工具包，成爲國際三大神經機器開源軟件之一。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"智源研究院和國際AI技術生態深入融合，建設智源社區，今年將匯聚10萬以上海內外人工智能科研人員。日常，智源社區舉辦豐富的學術交流活動，包括學術報告會、主題沙龍等，以及一年一界的智源大會，打造聚合全球一流AI學者前沿思想、連接世界AI學術與產業資源的平臺，構建全球人工智能學術和技術創新生態。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“青源會”是智源研究院發起的青年科學家學術組織，爲海內外的AI青年科學家和技術人員建立的寬鬆、活躍的學術交流平臺，促進學科交叉，支持青年科學家提出具有引領作用的原創思想，開創新的科學前沿。在6月2日上午舉辦的“青源學術年會”上，將舉辦“青源會”成立儀式，詳細介紹“青源會”發展目標及未來規劃。屆時歡迎各位參加。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當前，人工智能已成爲經濟發展的新引擎。智源研究院加快推進AI原創成果落地，加快人工智能深度應用，孵化AI創新企業，推動人工智能產業發展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"健康醫療也是人工智能最熱門的研究和應用領域之一，北京在發展智能醫療方面，有着得天獨厚的優勢。爲此，我們成立了“健康醫療研究中心”，構建醫療大數據和高質量小數據等基礎數據，加快智能醫療核心技術突破，加快人工智能在健康醫療領域的深度應用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，我們與清華智能產業研究院開展深度合作，推動以“個人”爲核心的、AI驅動的主動健康管理。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在推動AI原創成果轉化方面，去年建立了多個創新中心，加速AI原創成果熟化及工程化。智源爲創新中心提供場地與研發資金支持。目前，已建設智能信息處理、認知知識圖譜、安全人工智能等8個創新中心。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一年來，智源創新中心建設取得重大進展，例如："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"視網膜芯片技術創新中心，已研製出超高速脈衝視覺芯片，用於高速場景、弱小目標、全時檢測等機器視覺應用。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"安全人工智能創新中心，研發出RealSafe人工智能安全平臺、DeepReal深度僞造內容檢測平臺等，並在金融、能源等行業進行應用。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"智能政務信息處理創新中心研發的政務問答系統，已在60多個北京政務場景中得到應用。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"疾病腦電智能技術創新中心，研發的臨牀腦電圖智能評估系統，準確率已達到落地應用水平，產品覆蓋全國11個省市，累計服務超過上萬名患者。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"今年，我們按照國家和北京市促進科技成果轉化的最新部署，積極整合技術、場景、人才、資金等資源，推出支持AI創業的“源創計劃”，爲AI創業團隊開放大模型、數據集等生態資源，爲來自學術界的AI科學家創業團隊對接應用場景，爲來自產業界的創業團隊對接業界領先的AI技術，加快形成可落地應用的AI產品。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外，還將通過創業基金等方式，加快孵化一批具有國際領先技術能力的AI創新企業，爲北京AI產業的持續發展儲備力量。“源創計劃”將在6月3日上午召開的“AI創業論壇”正式發佈，敬請期待。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當前，AI倫理和治理成全球共識，國際社會正探索建立廣泛認可的AI倫理原則，推進敏捷靈活的AI治理。智源研究院在2019年成立人工智能倫理與可持續發展研究中心，率先在國內開展AI倫理安全研究，並積極參與國際AI治理。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AI倫理規範方面，2019年5月，發佈了我國首個AI發展與治理準則——《人工智能北京共識》。去年9月，又發佈了我國首個《面向兒童的人工智能北京共識》。另外，我們建設了我國首個“人工智能治理公共服務平臺”，針對人工智能技術研發及應用中存在的潛在倫理問題提供檢測服務，幫助AI研發機構研發更符合AI倫理規範的技術和產品。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"聯合國在2015年正式通過了可持續發展目標，共17大項，涉及社會、經濟和環境三方面的發展問題。爲此我們發起成立了“面向可持續發展的人工智能國際協作網絡與智庫”，目前已匯聚了來自15個國家的專家學者。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同時，我們也聯合百度、小米、曠視、滴滴等企業發起了“面向可持續發展的公益計劃”，目前已圍繞城市治理、自動駕駛、勞動者就業影響、青少年心理影響等方面發佈了首批4個研究題目，並已啓動研究工作。後續，完成後的研究成果將會面向全球公開，服務於人類命運共同體的構建與實現。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上是智源研究院一年來的工作進展，我的介紹就到這裏，謝謝大家。"}]}]}