Altman: world should accept some AI harms in exchange for huge benefits
In the first episode of Politico's new podcast "Decoded," OpenAI CEO Sam Altman said the world should accept some negative events from AI in exchange for the technology's benefits and user autonomy. Asked how OpenAI's approach to AI safety differs from Anthropic's, he said one key disagreement with strict safety advocates is that OpenAI favors a lighter-touch regulatory stance, arguing "some bad things" will happen as society explores resilience. NYU professor emeritus Gary Marcus criticized the remarks as telling the public to put up with harm. DoNews also cites TechSpot listing AI-driven job losses, data-center pollution and noise, copyright infringement, online spam, declining critical thinking, and surging power and water demand, which Altman still says are outweighed by the benefits.
Why it matters: It publicly frames the split between OpenAI and Anthropic over AI safety and regulation, and keeps the "accept risk for upside" argument at the center of the debate between frontier labs and their critics.
![]()
![]()
首页快讯商业 消费游戏3C家电汽车文娱
体育专栏直播专题APP
DoNews>商业>OpenAI 奥尔特曼:世界应接受 AI 带来的部分坏处以换取巨大红利
OpenAI 奥尔特曼:世界应接受 AI 带来的部分坏处以换取巨大红利
杨亮 2026-10-08 15:58:09
14136
分享到

10月8日消息,在接受 Politico 新播客《Decoded》首期节目采访时,OpenAI 首席执行官萨姆 · 奥尔特曼(Sam Altman)表示,世界应接受人工智能带来的部分“坏事”,以换取技术带来的巨大益处与用户自主权。
在被问及在人工智能安全方面,OpenAI 和竞争对手 Anthropic 的方法有何不同时,奥尔特曼表示:
我认为其中有很多值得商榷的地方。OpenAI 与部分严格 AI 安全倡导者的关键分歧之一,就是我们认为世界应接受技术带来的部分负面事件,以换取 AI 红利与用户自主权。
我们倡导较轻监管立场(lighter touch regulatory stance),社会在探索系统韧性(resilience)过程中,一些不好的事情将会发生。
奥尔特曼认为凡事都有两面性,但从推动 AI 的收益来看,AI 好事数量级远超坏事。科技媒体 TechSpot 称尽管 AI 已引发大规模失业、数据中心污染与噪音、版权侵权、网络垃圾、批判性思维下降、电力与水资源需求激增等问题,他仍坚持收益将远超弊端。
纽约大学荣誉教授加里 · 马库斯(Gary Marcus)批评奥尔特曼“把不该说的话大声说了出来:忍着点,这样你们才能让我们变得富有和强大。”大量公众对奥尔特曼再次强调 AI 益处远大于弊端表示不满。

标签:OpenAI萨姆·奥尔特曼Anthropic人工智能安全
关于我们| 电子协议| 合作联系| 京ICP备2025120072号
网站信息![]()
Copyright © DoNews 2000-2026 All Rights Reserved
京ICP备2025120072号
联系地址:北京市海淀区宝盛东路兴华绿色产业楼3层307室(东升地区)
邮箱:[email protected]
网上有害信息举报专区: www.12377.cn
Copyright © DoNews 2000-2026 All Rights Reserved
京ICP备2025120072号


获取验证码 获取验证码
登录
密码登录 >
验证码登录 >
#热门搜索
热门文章
最新文章
推荐文章
How we got here
- Center for Humane Technology lays off half its staff in 'founder-led' pivotWIRED AI · OpenAI
- UniPat AI's PaperBenchX benchmark: GPT-6 Astra fully reproduces only 13.98% of papers量子位(原生 RSS) · OpenAI
- OpenAI withdraws three math papers over a sign errorHacker News · AI(100+ 分) · OpenAI
- hackingtool: AI-guided all-in-one security testing toolkit trends on GitHubGitHub Trending · Python · OpenAI
- gpt-instruct publishes Codex jailbreak prompt pack with new pre-releasesGitHub Trending · Python · OpenAI
- PFN launches PLaMo 3 Translate 31B, claiming edge over GPT-6 on translationAIbase AI新闻 · OpenAI

京公网安备11010802023059号