每日下载:中国AI引发白宫分歧,以及创纪录的版权和解金

MIT Technology Review 新闻

摘要

本期通讯涵盖白宫内部对如何应对中国强大的开源AI模型Kimi的政治分歧,以及Anthropic创纪录的15亿美元版权和解协议的批准。

<p><em>这是今天的</em><a href="https://forms.technologyreview.com/newsletters/briefing-the-download/?_ga=2.179569122.736533416.1649661040-405833893.1649413289"><em>每日下载</em></a><em>,我们的工作日通讯,为你每日带来科技世界的最新动态。</em><br></p> <h3 class="wp-block-heading"><strong>中国AI模型让特朗普的AI阵营内部分裂</strong><strong></strong></h3> <p>上周末,总统唐纳德·特朗普的几位现任和前任AI顾问公开向美国领先AI公司恶语相向。David Sacks称Anthropic的模型“脑叶切除”且“觉醒”。五角大楼高级官员Emil Michael则称OpenAI的新战略未来主管是“超级乡村白痴”。</p> <p>这一切源于各方对中国AI公司Moonshot上周发布的免费开源模型Kimi看法不一。该模型的智能水平似乎与OpenAI和Anthropic的模型相当,而后两者绝非免费。&nbsp;</p> <p>每当中国发布一个新的智能免费模型,美国公司就更没有理由为Anthropic或OpenAI的模型掏钱。这给总统带来了经济和政治问题,也分裂了他身边的顶级AI战略家。&nbsp;</p> <p><a href="https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*">阅读完整报道,了解为何各方对Kimi的处理方式无法达成一致</a>。</p> <p><em>——James O&#8217;Donnell</em></p> <p><strong>这篇</strong><a href="https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*"><strong>文章</strong></a><strong>来自我们的每周AI通讯《算法》。</strong><a href="https://forms.technologyreview.com/newsletters/ai-demystified-the-algorithm/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*"><strong>订阅</strong></a><strong>后每周一将在你的收件箱中收到。</strong></p> <p class="has-medium-font-size"><strong>必读文章</strong></p> <p><em>我浏览了互联网,为你找到今天最有趣/重要/恐怖/迷人的科技故事。</em></p> <p><strong>1 Anthropic创纪录的15亿美元版权和解协议获批</strong><br>原告称Anthropic使用盗版作品训练Claude。(<a href="https://www.reuters.com/world/us-judge-approves-anthropics-15-billion-settlement-copyright-lawsuit-2026-07-20/">Reuters</a> $)<br>+ <em>并赢得了历史上已知的最大版权赔付金。</em>(<a href="https://www.engadget.com/2219475/judge-approves-anthropic-1-5-billion-settlement-authors/">Engadget</a>)<br><em>+ 然而许多作者和创作者仍不认为这是胜利。</em>(<a href="https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/">TechCrunch</a>)<br><em>+ 但AI版权焦虑可能限制创造力。</em>(<a href="https://www.technologyreview.com/2025/06/17/1118898/ai-copyright-anxiety-will-hold-back-creativity/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*">MIT Technology Review</a>)<br>&nbsp;<br><strong>2 特朗普政府正在权衡禁止中国AI模型</strong><br>Kimi K3的发布重新引发了限制呼声。(<a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi">Axios</a>)<br><em>+ 但官员们对这些提案存在分歧。</em>(<a href="https://www.fastcompany.com/91576757/trumps-proposed-ban-on-chinese-ai-models-could-strengthen-beijings-hand">Fast Company</a>)<br><em>+ 中国对开源的下注正在获得回报。</em>(<a href="https://www.technologyreview.com/2026/04/21/1135658/china-open-source-models-ai-artificial-intelligence/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*">MIT Technology Review</a>)<br>&nbsp;<br><strong>3 中国正在考虑对AI模型和芯片实施更严格的出口管制</strong><br>它希望阻止西方获取其技术和初创公司。(<a href="https://www.ft.com/content/6049a031-9e9b-464c-97bb-414da04d5a6a?syn-25a6b1a6=1">FT</a> $)<br><em>+ 北京已与科技公司就潜在限制进行会谈。</em>(<a href="https://www.reuters.com/world/asia-pacific/china-considers-tighter-export-controls-ai-models-chips-ft-reports-2026-07-21/">Reuters</a> $)<br>&nbsp;<br><strong>4 特朗普的AI安全负责人上任仅三个月后辞职</strong><br>Chris Fall自4月起领导联邦AI安全研究所CAISI。(<a href="https://www.axios.com/2026/07/20/trump-ai-security-agency-head-resigns">Axios</a>)<br><em>+ 未给出辞职原因。</em>(<a href="https://www.cnbc.com/amp/2026/07/20/trumps-head-of-ai-safety-agency-caisi-resigns-after-months-on-job.html">CNBC</a>)<br>&nbsp;<br><strong>5 谷歌正在研发新芯片以更高效运行Gemini模型</strong><br>这款名为“Frozen V2”的芯片可能在2028年部署。(<a href="https://www.theinformation.com/articles/google-plans-new-frozen-chip-run-ai-models-efficiently">Information</a> $)<br><em>+ 该报道推动Alphabet股价上涨。</em>(<a href="https://www.cnbc.com/2026/07/20/alphabet-googl-stock-ai-chip-report.html">CNBC</a>)<br><br><strong>6 新奥尔良警方曾探索为无人机配备武器</strong><br>一份无人机手册草案为武装四旋翼机铺平了道路。(<a href="https://www.404media.co/new-orleans-cops-published-policy-document-allowing-weaponized-drones/">404 Media</a>)<br><em>+ 扒手可能很快会被无人机追赶。</em>(<a href="https://www.technologyreview.com/2025/09/25/1124088/shoplifters-could-soon-be-chased-down-by-drones/?utm_source=the_download&amp;utm_medium=email&amp;utm_campaign=the_download.unpaid.engagement&amp;utm_term=*%7CSUBCLASS%7C*&amp;utm_content=*%7CDATE:m-d-Y%7C*">MIT Technology Review</a>)<br>&nbsp;<br><strong>7 欧盟因销售不安全产品对速卖通处以创纪录罚款</strong><br>5.5亿欧元罚款是《数字服务法》实施以来的最大金额。(<a href="https://www.bbc.com/news/articles/cp8714286dpo">BBC</a>)<br><em>+ 阿里巴巴誓言将上诉。</em>(<a href="https://www.scmp.com/tech/policy/article/3361255/alibaba-vows-appeal-us629m-eu-fine-breaches-digital-services-act?utm_source=feedly_feed">SCMP</a>)<br><br><strong>8 AI聊天机器人的选举建议“不准确且不可靠”</strong><br>这是今年早些时候在匈牙利测试得出的结论。(<a href="https://www.theguardian.com/technology/2026/jul/21/election-voting-advice-ai-chatbots-inaccurate-unreliable-hungary">Guardian</a>)<br><br><strong>9 红光疗法在愈合和健康老龄化方面展现前景</strong><br>改善皮肤和减少视力损失也在预期之中。(<a href="https://www.economist.com/science-and-technology/2026/07/20/the-surprising-benefits-of-red-light">Economist</a> $)<br><br><strong>10 Neill Blomkamp的新恐怖短片完全由AI生成——而且很糟糕</strong><br>这位著名导演想“以这种形式制作一部长片”。(<a href="https://gizmodo.com/neill-blomkamps-new-horror-clip-is-entirely-ai-generated-sucks-2000788103">Gizmodo</a>)</p> <p class="has-medium-font-size"><strong>今日语录</strong></p> <h2 class="wp-block-heading"><strong>“这将是一种非常自毁的干涉”</strong></h2>
查看原文
查看缓存全文

缓存时间: 2026/07/21 12:37

# 下载:中国AI白宫分歧与版权赔偿创纪录 来源:https://www.technologyreview.com/2026/07/21/1140685/the-download-chinese-ai-divides-white-house-anthropic-copyright-settlement *今天是**《The Download》*(https://forms.technologyreview.com/newsletters/briefing-the-download/?_ga=2.179569122.736533416.1649661040-405833893.1649413289)*的最新一期,这是我们每周发布的工作日通讯,为你带来科技世界的最新动态。* ### **中国的AI模型让特朗普的AI世界内讧** 上周末,美国总统唐纳德·特朗普的多位现任及前任AI顾问公开抨击了该国领先的AI公司。David Sacks称Anthropic的模型“被阉割”且“觉醒”。五角大楼高官Emil Michael则把OpenAI的新任战略未来主管称为“超级村夫”。 起因是没人能就中国AI公司月之暗面上周发布的免费开源模型Kimi达成一致。Kimi的智能似乎足以与OpenAI和Anthropic的模型匹敌,而后者远非免费。 每当中国发布一款新的、智能且免费的模型,美国公司就少了一个为Anthropic或OpenAI模型掏钱的理由。这给总统带来了经济和政治上的麻烦,也让他身边的顶级AI战略家们产生了分歧。 阅读全文了解为何没人能就Kimi达成一致(https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*) *—James O'Donnell* **本文**(https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*)**来自《The Algorithm》,我们每周的AI通讯。** **订阅**(https://forms.technologyreview.com/newsletters/ai-demystified-the-algorithm/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*)**即可每周一收到。** **必读** *我搜遍了互联网,为你找到今天最有趣/重要/可怕/引人入胜的科技故事。* **1 Anthropic创纪录的15亿美元版权和解获批** 原告称Anthropic使用盗版作品训练Claude。(路透社(https://www.reuters.com/world/us-judge-approves-anthropics-15-billion-settlement-copyright-lawsuit-2026-07-20/)$)+ *并获得历史上已知的最大版权赔偿。*(Engadget(https://www.engadget.com/2219475/judge-approves-anthropic-1-5-billion-settlement-authors/))*+ 但许多作者和创作者仍不认为这是胜利。*(TechCrunch(https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/))*+ 但AI版权焦虑可能限制创造力。*(MIT Technology Review(https://www.technologyreview.com/2025/06/17/1118898/ai-copyright-anxiety-will-hold-back-creativity/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*)) **2 特朗普政府正在考虑禁止中国AI模型** Kimi K3的发布重新引发了限制呼声。(Axios(https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi))+ *但官员们对这些提案存在分歧。*(Fast Company(https://www.fastcompany.com/91576757/trumps-proposed-ban-on-chinese-ai-models-could-strengthen-beijings-hand))+ *中国押注开源正在获得回报。*(MIT Technology Review(https://www.technologyreview.com/2026/04/21/1135658/china-open-source-models-ai-artificial-intelligence/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*)) **3 中国正在考虑对AI模型和芯片实施更严格的出口管制** 它希望阻止西方获取其技术和初创公司。(FT(https://www.ft.com/content/6049a031-9e9b-464c-97bb-414da04d5a6a?syn-25a6b1a6=1)$)+ *北京已与科技公司就潜在限制措施进行了会谈。*(路透社(https://www.reuters.com/world/asia-pacific/china-considers-tighter-export-controls-ai-models-chips-ft-reports-2026-07-21/)$) **4 特朗普的AI安全负责人上任仅三个月后辞职** Chris Fall自4月起领导联邦AI安全研究所CAISI。(Axios(https://www.axios.com/2026/07/20/trump-ai-security-agency-head-resigns))+ *未给出辞职原因。*(CNBC(https://www.cnbc.com/amp/2026/07/20/trumps-head-of-ai-safety-agency-caisi-resigns-after-months-on-job.html)) **5 Google正在开发新芯片以更高效地运行Gemini模型** 该芯片名为“Frozen V2”,可能于2028年部署。(Information(https://www.theinformation.com/articles/google-plans-new-frozen-chip-run-ai-models-efficiently)$)+ *Alphabet股价因该报道上涨。*(CNBC(https://www.cnbc.com/2026/07/20/alphabet-googl-stock-ai-chip-report.html)) **6 新奥尔良警方探索为无人机配备武器** 一份无人机草案为武装四轴飞行器铺平了道路。(404 Media(https://www.404media.co/new-orleans-cops-published-policy-document-allowing-weaponized-drones/))+ *扒手可能很快会被无人机追逐。*(MIT Technology Review(https://www.technologyreview.com/2025/09/25/1124088/shoplifters-could-soon-be-chased-down-by-drones/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*)) **7 欧盟因销售不安全产品对速卖通处以创纪录罚款** 5.5亿欧元的罚款是《数字服务法》实施以来最大的一笔。(BBC(https://www.bbc.com/news/articles/cp8714286dpo))+ *阿里巴巴誓言要上诉。*(SCMP(https://www.scmp.com/tech/policy/article/3361255/alibaba-vows-appeal-us629m-eu-fine-breaches-digital-services-act?utm_source=feedly_feed)) **8 来自AI聊天机器人的选举建议“不准确且不可靠”** 这是今年早些时候在匈牙利进行测试得出的结论。(卫报(https://www.theguardian.com/technology/2026/jul/21/election-voting-advice-ai-chatbots-inaccurate-unreliable-hungary)) **9 红光疗法在愈合和健康老龄化方面显示出希望** 改善皮肤和减少视力丧失也在预期之中。(经济学家(https://www.economist.com/science-and-technology/2026/07/20/the-surprising-benefits-of-red-light)$) **10 Neill Blomkamp的新恐怖短片完全由AI生成——而且很差** 这位著名导演想“以这种形式制作一部长片”。(Gizmodo(https://gizmodo.com/neill-blomkamps-new-horror-clip-is-entirely-ai-generated-sucks-2000788103)) **每日语录** ## **“如果这种干预真的发生,那将是一种极其自我挫败的行为。”** ——科技投资者Chamath Palihapitiya在X上(https://x.com/chamath/status/2079236756365348888)抨击限制中国AI模型的计划。 **还有一件事** 深蓝色天空背景下电线杆上的监控摄像头AKILAH TOWNSEND #### **走进芝加哥的监控全景监狱** 2024年9月2日凌晨,四人在一起芝加哥西行列车枪击案中身亡。警方迅速启动数字天罗地网——一个连接全市数千个摄像头的监控网络——仅90分钟后就逮捕了嫌疑人。 执法部门和安全倡导者认为,这个庞大的监控系统保护了公共安全,运行良好。但活动人士和许多居民表示,这是一个监控全景监狱,对行为产生寒蝉效应,侵犯了隐私和言论自由保障。 走进这个分裂芝加哥的监控网络(https://www.technologyreview.com/2026/02/23/1132740/inside-chicago-surveillance-panopticon/?utm_source=the_download&utm_medium=email&utm_campaign=the_download.unpaid.engagement&utm_term=*%7CSUBCLASS%7C*&utm_content=*%7CDATE:m-d-Y%7C*) *—Rod McCullom* **我们仍能拥有美好事物** *一个为你带来舒适、乐趣和消遣的地方,照亮你的一天。(有想法?*[*给我发邮件*](mailto:[email protected])*。)* + NASA分享了一段令人惊叹的延时视频(https://www.jpl.nasa.gov/news/nasas-psyche-mission-delivers-mars-flyby-data-time-lapse-video/),展示了Psyche探测器眼中的火星。 + 这个美国与欧洲城市主义(https://www.instagram.com/p/DZzaXrjiIaS?img_index=3)的对比显示,良好的城市设计是一种选择。 + 音乐家Luca Stricagnoli最近表演了一段精彩的Prodigy歌曲原声吉他串烧(https://www.youtube.com/watch?v=rtbjZTAzZhc)。 + 两名澳大利亚桨板运动员在沙袋鼠被冲入大海后将其救起——并拍下了整个救援过程(https://x.com/depressionlesss/status/2066473290114208125)。

相似文章

白宫正在设法应对中国AI

Wired

白宫内部对于如何应对中国快速进步的AI存在分歧,争论焦点在于实施更严格的管控还是制定可行的限制措施,尤其是针对美国模型(如Anthropic的Claude Fable 5)遭受的蒸馏攻击。

下载:你在OpenAI中的股份,以及财政部的AI警告

MIT Technology Review

本期通讯涵盖两大AI新闻:山姆·奥特曼提议美国政府获得OpenAI 5%股份(价值每户320美元),以及一份泄露的财政部报告警告AI市场泡沫。还包括三星AI芯片利润更新和CISA使用Anthropic的Mythos。