特斯拉AI日：發佈自研超算Dojo芯片，7nm製造，每秒可執行1024億次計算

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"D1 芯片來了，Dojo 還會遠嗎？"}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"特斯拉發佈 Dojo D1 芯片"}]},{"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":"北美時間 8 月 19 日，特斯拉 AI Day 正式舉行。在發佈會上，馬斯克介紹了特斯拉在人工智能領域的軟件和硬件進展。其中，自研超級計算機 "},{"type":"link","attrs":{"href":"https:\/\/www.cnbc.com\/2021\/08\/19\/tesla-unveils-dojo-d1-chip-at-ai-day.html","title":"xxx","type":null},"content":[{"type":"text","text":"Dojo D1 芯片"}]},{"type":"text","text":"成爲本次發佈會最大的亮點。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/05\/b1\/05e7570228f7db6599c5b5ecedc27db1.png","alt":null,"title":"Ganesh Venkataramanan 展示 D1 訓練芯片","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"會上，Dojo項目負責人Ganesh Venkataramanan上臺就Dojo的主要性能進行了展示。Ganesh Venkataramanan表示，馬斯克想要一臺超快的訓練計算機來訓練 Autopilot。因此，Project Dojo 誕生了。Dojo架構擁有一個大規模計算平面，極高寬帶和低延遲。作爲Dojo架構的重要組成部分，D1芯片採用7納米制造工藝，處理能力爲每秒1024億次。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/3b\/93\/3b9db9817d8f81e9d6225c44ffce6493.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"曾在芯片製造商AMD工作的Venkataramanan表示，將一組這樣的芯片放置在單個“訓練片”上，以提供每秒九千萬億次的計算能力，並將120個芯片放在多個服務器機櫃上，達到每秒超過1千萬億次的計算能力。這些芯片可以幫助訓練模型來識別特拉斯汽車攝像頭中收集到的各種物品。訓練模型需要大量的計算工作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/58\/ab\/58fd17c8f73be111b91573c85914fbab.jpeg","alt":null,"title":"一個特斯拉 Dojo 機器學習計算塊","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"Dojo的規格、性能十分強大。Dojo是一個真正意義上的機器學習架構，它擁有50多萬個訓練節點。每秒九千萬億次的計算以及每秒36tb的帶寬還都只是Dojo的冰山一角。比這更可怕的是，Dojo的全部潛力都在爲一件事情做準備——使自動駕駛汽車成爲可能。Venkataramanan認爲，特斯拉的技術將打造出最快的AI訓練計算機。他還表示：“特斯拉很快就會組裝首批套件。\""}]},{"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，馬斯克還表示可能會在明年推出名爲“特斯拉機器人”的人形機器人原型，這種具有人類外觀的機器人將執行人們最不喜歡做的工作，“消除危險、重複、無聊的任務”，並“對經濟產生深遠的影響”。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/cf\/30\/cf67e9dbfa7dcbc1fa5acb9d4533c330.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"超級計算機 Dojo"}]},{"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":"事實上，大家對於特斯拉本次發佈 Dojo 芯片並不意外。在特斯拉本月初發布的邀請函上，一顆大規模芯片模組的結構圖格外吸睛，彼時就有不少人猜測，這是特斯拉爲超級計算機 Dojo 設計的專用芯片。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/fe\/1d\/fe4c823fd009067743dc162716c9e91d.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"而在特斯拉發佈邀請函前兩日，自動駕駛領域專家 Dennis Hong 就已在其 Twitter 上發佈過這張圖片，並打上了 Tesla AI Day 標籤。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/7f\/a0\/7f98eff476b1379776eff253190d92a0.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"據瞭解，Dennis Hong 是加州大學洛杉磯分校（UCLA）機械和航天工程學院的教授、機器人實驗室 Romela 負責人，主要研究方向爲類人機器人、機器設計以及自動駕駛汽車研究。2007 年，Dennis Hong 及團隊參加了美國 DARPA 無人駕駛汽車挑戰賽，並獲得了第三名的好成績。他們基於一輛福特 SUV 製造了一款自動駕駛汽車，無需干預即可自動到達目的地，如果按照美國汽車工程學會（SAE）的標準，這輛車的自動駕駛能力可以達到 L4 級。2011 年，Dennis Hong 發表了一個以“爲盲人制造汽車”爲主題的 TED 演講，在自動駕駛領域內頗具影響力。"}]},{"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":"外媒 "},{"type":"link","attrs":{"href":"https:\/\/www.teslarati.com\/tesla-dojo-chip-images-dennis-hong\/","title":"xxx","type":null},"content":[{"type":"text","text":"Teslarati 推測"}]},{"type":"text","text":"，Dennis Hong 可能已經在私下裏與特斯拉成立了合作關係，前者將作爲編外專家幫助後者進行 Dojo 超級計算機的調試與完善，不過特斯拉官方和 Dennis Hong 本人都未對此觀點表示承認或否認。"}]},{"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":"事實上，早在 2019 年的 Autonomy Day 上，馬斯克就提到過 "},{"type":"link","attrs":{"href":"https:\/\/www.theverge.com\/2021\/8\/18\/22627101\/tesla-event-ai-day-rumors-elon-musk-what-to-expect","title":"xxx","type":null},"content":[{"type":"text","text":"Dojo"}]},{"type":"text","text":"，並表示 Dojo 能夠利用海量的視頻（級別）數據，做“無人監管”的標註和訓練的超級計算機。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/b7\/5f\/b750aa14e86b0081a08261a0ca50aa5f.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"2020 年 8 月 15 日，馬斯克發推稱“特斯拉正在開發一款名爲 Dojo 的神經網絡訓練計算機，用於處理大量視頻數據”。當時馬斯克還暗示了 Dojo 的計算能力，聲稱它具備 exaFLOP 性能，即每秒能夠執行 1018 萬億次的浮點運算。超越日本富嶽，成爲世界第一。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/resource\/image\/37\/d0\/372751f8a7789f000524990805cf48d0.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"至於爲什麼要自研超算 Dojo，馬斯克也給過答案：“解決自動駕駛的唯一方法是解決現實世界中的 AI 問題，無論是硬件還是軟件，而這也是特斯拉正在做的事情。除非一家公司具有很強的 AI 能力以及超強算力，否則很難解決自動駕駛難題。”"}]},{"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":"如今 Dojo 芯片已來，根據馬斯克的計劃，明年 Dojo 會投入運營。至於 Dojo 未來會爲特斯拉以及自動駕駛帶來哪些改變，一起拭目以待。"}]}]}