President Vladimir Putin has often proclaimed that Russia must lead the world in artificial intelligence. In reality, the country is stuck on the sidelines as others pull ahead. PREMIUM Russian President Vladimir Putin, in red tie, at an AI forum in Moscow with Sberbank CEO Herman Gref. The state-owned lender leads Russia’s AI efforts. As…
在現代科學研究中，實驗室的規模與員工人數已不再是最重要的資源。Meta(META-US) 執行長祖克柏（Mark Zuckerberg）與醫生妻子 Priscilla Chan 表示，對於他們的慈善機構陳和祖克柏基金會（Chan Zuckerberg Initiative，CZI）來說，最關鍵的資源是運算能力，尤其是 GPU。 Meta祖克柏：研究人員不需更多實驗室或員工、他們只要GPU。(圖：Shutterstock) 根據《Business Insider》報導，Priscilla Chan 近日談到 CZI 旗下支持的生物研究中心生物樞紐（Biohub）時表示：「我們不是在擴張實驗室空間，而是在擴張運算能力。」 祖克柏補充：「研究人員不需要更多員工，也不需要更多實驗空間，他們只需要 GPU。」 Priscilla Chan 指出，GPU 可視為「新型實驗室空間」，但其成本遠高於傳統濕實驗室。 當被問及 CZI 是否會增加生物樞紐的人員或採用多點網絡模型時，祖克柏表示，兩者可能都會進行。 他透露，CZI 已經新增了多個生物樞紐分支，並正在組建一支中央 AI 團隊。 本月初，祖克柏與 Priscilla Chan 宣布，CZI 將重新調整重心，更加專注於生物樞紐的運作。 生物樞紐也宣布，將與 AI 公司 EvolutionaryScale 合作，利用 AI 與 GPU 運算能力加速人類疾病研究。 生物樞紐在聲明中指出，計畫到 2028 年將運算能力提升至 10,000 張 GPU。目前 CZI 的運算叢集已有 1,000 張 GPU，Priscilla…
Moore Threads Technology, a Chinese graphics processing unit (GPU) start-up founded by a veteran of Nvidia’s China operations, has become one of the most closely watched semiconductor players in the country, as domestic AI chip designers race to displace the US chip giant. The Beijing-based company last week received approval from the Shanghai Stock Exchange…
When Chinese artificial intelligence start-up DeepSeek unveiled an updated foundational model late in August, investors in Nvidia were stunned. Shares of the US chip giant slid, as market watchers grappled with news that the two-year-old start-up, which has developed models rivalling the world’s best, was shifting towards supporting domestically produced chips. Adding to their concerns,…
Nvidia’s RTX6000D, its newest artificial intelligence chip tailored for the Chinese market, has seen only lukewarm demand with some major tech firms opting not to place orders, two people with knowledge of procurement discussions said. The RTX6000D, designed mainly for AI inference tasks, is seen as expensive for what it does, according to the two…
Nvidia’s earnings aren’t just about Nvidia anymore. The $4 trillion chipmaker’s quarterly financials have become a litmus test for the AI boom—and, by extension, for the whole stock market. Constituting 8% of the market cap-weighted in S&P 500 Index and with an unrivaled grip on the chips that power generative AI, Wall Street now treats…
The sales resumption of Nvidia’s H20 processors and Advanced Micro Devices’ (AMD) MI308 chips in China is expected to ease the anxiety of local artificial intelligence companies, according to analysts. Speaking at the China International Supply Chain Expo in Beijing on Wednesday, Nvidia CEO Jensen Huang said he hoped to introduce more advanced chips to…
Meta 最近豪砸數千億美元成立「超級智慧部門」，開始對業界瘋狂挖角 AI 人才大概有幾個禮拜了，他們預計將投入數千億美元開發「超越人類能力」的 AI 系統。 挖角行動包括找來 Scale AI 前執行長 Alexandr Wang 擔任首席 AI 長來領導該部門，同時網羅 11 名來自 OpenAI、Google 等競爭對手的頂尖人才。 對此，Meta 執行長祖克柏（Mark Zuckerberg）最近在接受媒體訪問時透露，在招聘頂尖 AI 人才時，除了優渥薪酬外，還有兩項關鍵因素特別受到重視：那就是這些 AI 人才最看重最精簡的管理架構，以及是否有最豐富的 GPU 運算資源！ 「過去找人時，求職者關心的是職責範圍有多大，」祖克柏說「但這群頂尖 AI 研究人才的要求恰恰相反——他們希望管理負擔最輕，同時握有豐沛的 GPU 資源。」 你可把祖克柏這番言論視為 GPU（等於算力）成為科技業搶奪人才的最重要籌碼之一。另一方面，Nvidia 的 H100 晶片在 AI 熱潮中供不應求，更凸顯了運算資源的稀缺性。Perplexity 執行長 Aravind Srinivas 曾分享另一個反例，有次他試圖從 Meta 挖角一位 AI 人才時對方毫不客氣：「等你有 1 萬張 H100 GPU 時再來談。」 Meta 前兩天才宣布將投入數千億美元建設多座大型人工智慧（AI）資料中心，第一座…