谷歌搜索引入多任務統一模型MUM，可更準確理解信息 ｜ Google I/O 2021

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"美國時間 5 月 18 日，"},{"type":"link","attrs":{"href":"https:\/\/events.google.com\/io\/?lng=zh-CN","title":"xxx","type":null},"content":[{"type":"text","text":"Google I\/O 2021開發者大會"}]},{"type":"text","text":"正式開幕。去年，該會議因疫情取消，今年重新恢復並採用全程線上的形式，對所有開發者免費開放。在剛剛結束的主題演講中，谷歌發佈了 TPU V4 人工智能芯片、"},{"type":"link","attrs":{"href":"https:\/\/www.infoq.cn\/article\/76gYqPA2YU0YXCDHFvIE","title":"xxx","type":null},"content":[{"type":"text","text":"自然語言平臺 LaMDA "}]},{"type":"text","text":"以及一系列原有產品的更新升級。本文，我們將詳細介紹谷歌搜索引入的多任務統一模型MUM。"}]}]},{"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":"color","attrs":{"color":"#494949","name":"user"}}],"text":"每天都有很多人使用谷歌來處理需要多步驟的各種任務，而人們在處理類似的複雜任務時平均會發出 8 個查詢。如今，搜索引擎還沒有成熟到可以像專家一樣回答問題。但隨着“多任務統一模型”（Multitask Unified Model，MUM）的出現，谷歌正在幫助解決這類複雜需求。因此，未來只需要較少的搜索就可以完成任務。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"與 "},{"type":"link","attrs":{"href":"https:\/\/blog.google\/products\/search\/search-language-understanding-bert\/","title":null,"type":null},"content":[{"type":"text","text":"BERT"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" 一樣，MUM 同樣基於 "},{"type":"link","attrs":{"href":"https:\/\/ai.googleblog.com\/2017\/08\/transformer-novel-neural-network.html","title":null,"type":null},"content":[{"type":"text","text":"Transformer 架構"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"，但是它的功能要強大 1000 倍。MUM 不僅可以理解語言，而且可以生成語言。MUM 同時用 75 種不同的語言進行了多項任務的訓練，使其比以前的模型更全面地理解信息和世界知識。此外，MUM 是多模態的，因此它能夠理解文本和圖像中的信息，將來，還可以擴展到視頻和音頻等更多模態。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"以徒步富士山的問題爲例：MUM 可以理解你在比較兩座山，因此海拔高度和路徑信息可能是相關的。它還可以理解，就遠足而言，“準備工作”可能包括諸如健身訓練以及尋找合適的裝備。"}]},{"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":"color","attrs":{"color":"#494949","name":"user"}}],"text":"因爲 MUM 能夠基於其對這個世界的深刻理解來展現自己的見解，所以它可以強調，儘管兩座山的海拔高度大致相同，但秋季是富士山的雨季，你可能需要一件防水夾克。MUM 也能爲更深層次的探索提供有用的副主題：比如頂級裝備或最佳訓練練習，並提供一些網絡上有用的文章、視頻和圖片的鏈接。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/d9\/cf\/d9a8059879eb1ba4d4bea9137e35f6cf.gif","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"消除語言障礙"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"在獲取信息時，語言可能是一個重要障礙。通過不同語言的知識遷移，MUM 有可能打破這些界限。它可以從那些不是用你的搜索語言寫成的資料中學習，並且能幫助把這些相關信息發給你。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"假設有一些關於富士山真正有用的信息是用日語寫的；現在，如果你不用日語搜索，你很可能無法找到這些信息。然而，MUM 可以從不同語言的來源中遷移知識，並利用這些洞察力發現與你的首選語言最相關的結果。所以，在將來，當你搜索有關遠足富士山的信息時，你可能會看到這樣的結果：在何處能欣賞到富士山最美的風景、當地的溫泉，以及受歡迎的紀念品商店……這些信息很容易用日語搜索就能找到。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9e\/64\/9ea06b15a094bb4899c9b5d740a47164.gif","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":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"理解不同類型的信息"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"多模態的 MUM 意味着它能夠同時理解來自不同格式的信息，比如網頁、圖片等等。最終，你可能會拍一張登山靴的照片，然後問：“我能用它去爬富士山嗎？”MUM 將會理解這張圖片，並把它和你的問題聯繫在一起，讓你知道你的靴子會很好用。之後，它會給你發一個博客網址，上面有推薦的裝備列表。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/5c\/a7\/5cff98c01a53d22d31e91bcf703e8fa7.gif","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"帶着負責的態度把高級人工智能運用到搜索中"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"無論何時，當我們使用人工智能來使世界上的信息更容易獲取時，我們都要負責任地這樣做。對於谷歌搜索的每一項改進，我們都會進行嚴格的評估，以確保我們能提供更加相關和有用的結果。那些遵循我們《"},{"type":"link","attrs":{"href":"https:\/\/static.googleusercontent.com\/media\/guidelines.raterhub.com\/en\/\/searchqualityevaluatorguidelines.pdf","title":null,"type":null},"content":[{"type":"text","text":"搜索質量評分準則"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"》（"},{"type":"text","marks":[{"type":"italic"},{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Search Quality Rater Guidelines"},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"）的人類評分者，幫助我們瞭解我們的結果如何幫助人們找到信息。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"就像我們已經仔細測試了 BERT 從 2019 年開始推出的許多應用一樣， MUM 也會經歷同樣的過程，將這些模型應用於搜索。具體地說，爲了避免在我們的系統中引入偏見，我們將尋找可能顯示機器學習中偏見的模式。同時，我們也會運用最新的"},{"type":"link","attrs":{"href":"https:\/\/arxiv.org\/abs\/2104.10350","title":null,"type":null},"content":[{"type":"text","text":"研究成果"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"，比如如何減少 MUM 等訓練系統的碳足跡，以確保搜尋工作儘可能高效。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"今後數月甚至數年，我們將把 MUM 驅動的功能和改進帶到我們的產品中。雖然我們仍處在 MUM 探索的初期，但這是一個重要的里程碑，將來谷歌能夠理解人們自然地交流和解釋信息的各種方式。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}},{"type":"strong"}],"text":"作者介紹："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Pandu Nayak，谷歌研究員兼搜索部門副總裁。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}},{"type":"strong"}],"text":"原文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"https:\/\/blog.google\/products\/search\/introducing-mum\/"}]}]}