矽穀獨角獸SCALE AI CEO ALEX WANG稱DEEPSEEK是earth-shattering AI模型

本帖於 2025-01-26 07:36:45 時間, 由普通用戶 顏陽 編輯

ALEX Wang 和OPEN AI的 SAM ALTMAN做了好幾年ROOMMATE。是ALTMAN的最好的朋友。。所以不要以為他是隨便說的。

https://www.cnbc.com/2025/01/23/scale-ai-ceo-says-china-has-quickly-caught-the-us-with-deepseek.html

Scale AI CEO says China has quickly caught the U.S. with the DeepSeek open-source model

Key Points
  • Alexandr Wang, whose company Scale AI provides training data to key artificial intelligence players including OpenAI, Google and Meta, said Thursday that the AI race between the U.S. and China is an “AI war.”
  • Wang said DeepSeek, the leading Chinese AI lab, released an “earth-shattering model” on Christmas Day.
  • Wang said he believes that it’ll take within two to four years to reach artificial general intelligence, or AGI, an oft-debated term and benchmark within the AI sector.
  • The U.S. may have led China in the artificial intelligence race for the past decade, according to Alexandr Wang, CEO of Scale AI, but on Christmas Day, everything changed.

    Wang, whose company provides training data to key AI players including OpenAI, Google and Meta, said Thursday at the World Economic Forum in Davos, Switzerland, that DeepSeek, the leading Chinese AI lab, released an “earth-shattering model” on Christmas Day, then followed it up with a powerful reasoning-focused AI model, DeepSeek-R1, which competes with OpenAI’s recently released o1 model.

     

    “What we’ve found is that DeepSeek ... is the top performing, or roughly on par with the best American models,” Wang said.

    In an interview with CNBC, Wang described the artificial intelligence race between the U.S. and China as an “AI war,” adding that he believes China has significantly more Nvidia H100 GPUs — AI chips that are widely used to build leading powerful AI models — than people may think, especially considering U.S. export controls.

    Wang also said he believes the AI sector will reach a trillion dollars, on par with estimates that the generative AI market is poised to top $1 trillion in revenue within a decade.

所有跟帖: 

看看現在世界 -讓股票飛- 給 讓股票飛 發送悄悄話 讓股票飛 的博客首頁 (111 bytes) () 01/26/2025 postreply 07:35:52

SCALE AI高管,Michael Kratsios, 原TRUMP的US CTO,已重新加入川普團隊,將被參議院確認 -顏陽- 給 顏陽 發送悄悄話 顏陽 的博客首頁 (0 bytes) () 01/26/2025 postreply 07:52:09

關鍵字:China has significantly more Nvidia H100 GPUs -BBL123- 給 BBL123 發送悄悄話 (0 bytes) () 01/26/2025 postreply 07:47:47

從這個文裏可以figure出,deepseek還是用的是Nvda芯片,而不是算法優化。至於芯片怎麽去的,不做評價 -傑西210- 給 傑西210 發送悄悄話 (0 bytes) () 01/26/2025 postreply 07:48:23

所以,成本低一些,但是不會那麽低。因為人工低,但是芯片一樣。有誰有機會看看deep seek機房,到底是不是nVDA芯 -傑西210- 給 傑西210 發送悄悄話 (0 bytes) () 01/26/2025 postreply 07:50:46

Never waste a crisis -越王劍- 給 越王劍 發送悄悄話 越王劍 的博客首頁 (300 bytes) () 01/26/2025 postreply 08:00:52

提醒我了,謝謝大佬. 感謝回歸投壇. -Hongmei20- 給 Hongmei20 發送悄悄話 (0 bytes) () 01/26/2025 postreply 08:07:08

越王高見!不過樓主及DeepSxx可能都是鐵掌水上漂裘千丈駕到,鋒芒太盛,我們隻能遠處觀望表演。 -八飛- 給 八飛 發送悄悄話 (61 bytes) () 01/26/2025 postreply 08:14:43

張口就胡說八道。5萬張H100?知道多少錢一張? 5萬張在中國需要多少錢? 完全不知道大陸情況的話 -評論2012- 給 評論2012 發送悄悄話 (0 bytes) () 01/26/2025 postreply 08:24:13

需要多少錢?如果很貴,成本就高。難道他用的國產芯片?如果這樣,那麽中國的確可以笑傲江湖了 -傑西210- 給 傑西210 發送悄悄話 (0 bytes) () 01/26/2025 postreply 08:45:11

可以帶上軟件去新加坡或者迪拜利用那裏的數據中心訓練 -桃花源裏人家- 給 桃花源裏人家 發送悄悄話 桃花源裏人家 的博客首頁 (0 bytes) () 01/26/2025 postreply 10:15:07

所以,美帝千方百計封鎖天朝的策略不一定奏效 -justbuy168- 給 justbuy168 發送悄悄話 justbuy168 的博客首頁 (0 bytes) () 01/26/2025 postreply 22:18:00

不用跑去,遠程雲端測試應該可以 -justbuy168- 給 justbuy168 發送悄悄話 justbuy168 的博客首頁 (0 bytes) () 01/26/2025 postreply 22:20:00

請您先登陸,再發跟帖!