請更新您的瀏覽器

您使用的瀏覽器版本較舊,已不再受支援。建議您更新瀏覽器版本,以獲得最佳使用體驗。

本文由AI翻譯

DeepSeek、月之暗面使用OpenAI的GPT-6 Astra訓練最新模型,《The Information》報導

Reuters

更新於 8小時前 • 發布於 8小時前
檔案照片:這張2026年9月5日拍攝的插圖可見月之暗面(Moonshot AI)標誌。路透社/Dado Ruvic/插圖/檔案照片FILE PHOTO: Moonshot AI logo is seen in this illustration taken, September 5, 2026. REUTERS/Dado Ruvic/Illustration//File Photo

路透10月28日 - 《The Information》週三援引直接知情人士報導,中國AI開發商DeepSeek(深度求索)和月之暗面(Moonshot AI)在訓練其最新模型時,高度仰賴OpenAI強大的GPT-6 Astra模型輸出成果。Oct 28 (Reuters) - Chinese AI developers DeepSeek and Moonshot AI relied heavily on outputs from OpenAI's powerful GPT-6 Astra model to train their latest models, The Information reported on Wednesday, citing people with direct knowledge of the matter.

• 這種做法稱為蒸餾。隨著美中科技競爭加劇,蒸餾已成為領先AI公司爭議焦點,相關企業主張,這種做法讓競爭對手能更快速、以更低成本重現先進能力• The practice, known as distillation, has become a flashpoint for leading AI companies as US-China technology competition intensifies, with firms arguing it allows rivals to reproduce advanced capabilities more quickly and cheaply

• 《The Information》表示,月之暗面上週發布的Kimi K3.1 Pro模型,以及DeepSeek於10月17日推出的升級版V4.1-Flash 1017模型,都受益於對Astra的蒸餾• Moonshot AI's Kimi K3.1 Pro model, which was released last week, and DeepSeek's upgraded V4.1-Flash 1017 model, which was launched on October 17, benefited from distilling Astra, The Information said

• OpenAI、DeepSeek和月之暗面未立即回應路透社的採訪請求• OpenAI, DeepSeek and Moonshot AI did not immediately respond to Reuters requests for comment

• OpenAI在9月30日一篇部落格文章中表示,已將一項7月展開的蒸餾活動的核心部分,與月之暗面相關人士連結起來• In a September 30 blog post, OpenAI said it had linked a core part of a distillation campaign that began in July to individuals associated with Moonshot AI

• 包括Anthropic、OpenAI和Google在內的企業主張,這種做法讓競爭對手能更快速、以更低成本重現先進能力• Firms including Anthropic, OpenAI and Google have argued the practice enables rivals to reproduce advanced capabilities more quickly and at lower cost

• Anthropic本月稍早表示,該公司為防止競爭對手複製其Claude AI模型能力而設計的新措施,已使所謂蒸餾攻擊中被蒐集的資料減少約90%至99%• Earlier this month, Anthropic said new measures designed to prevent rivals from copying the capabilities of its Claude AI models have reduced data harvested in so-called distillation attacks by about 90% to 99%

(Disha Mishra班加羅爾報導;Sherry Jacob-Phillips與Subhranshu Sahu編輯)(Reporting by Disha Mishra in Bengaluru; Editing by Sherry Jacob-Phillips and Subhranshu Sahu)

更多國際相關文章

01

第五縱隊滲透日本?銀座5月遭催淚瓦斯「恐攻」釀30傷 主嫌是中國人

鏡報
02

列車追撞釀68死!乘客被甩飛、鐵軌刺穿車頂 現場慘況曝

CTWANT
03

高鐵追撞釀68死!乘客斷臂仍冷靜求援 78歲藥師「衝現場救人」揭慘況

CTWANT
04

留錢沒意義!日籍婦「繼承1000萬遺產」 改吃高級餐廳、買包3年後卻因1問題後悔了

鏡週刊
05

日本河口湖命案/井女友人曝「媽媽很自責」告別式上痛哭 痛斥網路謠言:寧願女兒還在

鏡報
06

長達13年來起床就腰痛!他裝監視器竟「意外得知謎樣真相」 全網驚呼:挺讓人羨慕的

鏡週刊
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...