Workflow
OpenAI
icon
Search documents
AI用多了,人会变傻吗?
3 6 Ke· 2025-08-05 07:17
神译局是36氪旗下编译团队,关注科技、商业、职场、生活等领域,重点介绍国外的新技术、新观点、新风向。 编者按:AI真的在腐蚀我们的脑子吗?有些人担心这项技术正在侵蚀批判性思维能力,有些人则认为这样的说法只是在制造恐慌。MIT的一项研究让我们得 以一窥大脑对人工智能辅助的反应。本文来自编译,希望对您有所启发。 生成式人工智能已成为数百万人日常生活的一部分,我们用它来润色邮件、头脑风暴、撰写文章,甚至让AI来扮演教练或治疗师的角色。但随着我们越来 越多地将思考任务交给人工智能,一些人开始质疑人们是否会为此付出隐性代价。 麻省理工学院(MIT)的一项最新研究加剧了这一争议,媒体头条纷纷质疑生成式人工智能是否会腐蚀人的大脑。有些人担心这项技术正在侵蚀批判性思维 能力,另一些人则认为这些说法只是在制造恐慌。 科学界对此有何看法?作为神经科学家,我或许能帮助大家理清头绪。让我们来看看这项研究究竟发现了什么(以及它没有发现什么)。 麻省理工学院的研究 1. 研究详情 大型语言模型组:使用ChatGPT辅助写作 搜索引擎组:使用谷歌搜索,不借助任何AI工具 纯脑力组:不借助任何工具进行写作 脑部活动:纯脑力组显示出最强且最广 ...
扎克伯格15亿美元挖不动的男人
投中网· 2025-08-05 06:37
以下文章来源于智东西 ,作者王 涵 智东西 . 智能产业新媒体!智东西专注报道人工智能主导的前沿技术发展,和技术应用带来的千行百业产业升 级。聚焦智能变革,服务产业升级。 将投中网设为"星标⭐",第一时间收获最新推送 安德鲁·塔洛克,现任Thinking Machines Lab联合创始人兼首席研究员。 OpenAI前CTO带队"团拒"。 作者丨 王涵 编辑丨 漠影 来源丨 智东西 智东西8月4日消息,据外媒《华尔街日报》近日报道,知情人士透露,与OpenAI前首席技术官米拉 ·穆拉蒂(Mira Murati)联合创办Thinking Machines Lab的安德鲁·塔洛克(Andrew Tulloch),拒绝了扎克伯格的可能高达 15亿美元(约合人民币108.2亿元) 的薪酬。 据报道,试图收购Thinking Machines Lab失败后,扎克伯格在随后数周内接触了该公司 约十几名 员工,最终成功 带走0个人 。 《华尔街日报》报道称,Meta发言人安迪·斯通(Andy Stone)回应称该薪酬描述"失实且荒谬", 强调任何薪酬方案都需以股价上涨为前提,并声明Meta无意收购Thinking Mach ...
大模型下一个飞跃?OpenAI的“新突破”:通用验证器
Hua Er Jie Jian Wen· 2025-08-05 06:07
在下一代大模型GPT-5备受期待之际,一项名为"通用验证器"的新技术正浮出水面,揭示了OpenAI可能 用于拉开竞争差距的"秘密武器"。 OpenAI的"通用验证器"或将直接影响GPT-5模型的市场竞争力,8月4日据科技媒体The Information援引 知情人士消息报道,这项技术已被应用于GPT-5的开发过程中。 该技术的核心机制,被比作一场"证明者-验证者游戏"。简而言之,它让一个AI模型扮演"验证者"的角 色,去检查和评判另一个"证明者"模型生成的答案。通过这种内部对抗和反馈,系统性地提升模型的输 出质量。这一自动化流程旨在解决强化学习(RL)在创意写作等主观领域或数学证明等复杂领域难以 验证的瓶颈。 OpenAI内部研究人员已在社交平台X上间接证实了相关方法的有效性。研究员Noam Brown表示,这些 技术是"通用的",能让大模型"在难以验证的任务上表现得更好"。这也标志着OpenAI正试图攻克AI商 业化应用中的核心痛点——可信度。 "证明者-验证者"的对抗游戏 "通用验证器"的技术细节,最早在OpenAI于2024年7月发表的一篇题为《证明者-验证者游戏提升大语言 模型可读性》的论文中被阐述 ...
中国AI产业投资蓝皮书
AI产业发展及应用概况—产业政策 国家大力支持算力、算法、数据、应用等AI核心环节发展,产业政策体系持续完善 生成 式AI 基础 设施 未来 产业 标准 化 安全 治理 AI 教育 国家网信办等七部门 2023/07 《生成式人工智能服务管理暂行办法》 工信部等六部门 2023/10 《算力基础设施高质量发展行动计划》 工信部、中国科学院等七部门 2024/01 《关于推动未来产业创新发展的实施意见》 工信部、国家发改委等六部门 2024/06 《国家人工智能产业综合标准化体系建设指南 (2024版)》 全国网络安全标准化技术委员会 2024/09 《人工智能安全治理框架》1.0版 气象 应用 内容 标识 数据 标注 专利 申请 2025/04 中国气象局、国家网信办. 《人工智能气象应用服务办法》 2025/03 国家网信办、工信部等四部门 《人工智能生成合成内容标识办法》 2024/12 国家发改委、国家数据局等四部门 《关于促进数据标注产业高质量发展的实施意见》 2024/11 国家知识产权局 《人工智能相关发明专利申请指引(试行)》 2024/11 教育部办公厅 《关于加强中小学人工智能教育的通知》 ...
OpenAI引信任危机:AI教父与诺奖得主等联名,要求回应八大问题
3 6 Ke· 2025-08-05 05:30
当地时间8月4日,多家非政府组织组建联盟发起公开信,要求OpenAI就计划中的公司重组提高透明 度。这封由Encode、The Midas Project、EyesOnOpenAI及加州非营利、慈善和劳工组织联盟等共同推 动的公开信,已获包括四位诺贝尔奖得主在内的众多知名人士及组织签署,信中集中表达了对OpenAI 可能弱化非营利使命、优先追逐利润的深切担忧。 签署者包括四位诺贝尔奖得主,包括有"AI教父"之称的多伦多大学名誉教授杰弗里・辛顿(Geoffrey Hinton,2024年物理学奖得主),罗马萨皮恩扎大学名誉教授乔治・帕里西(Giorgio Parisi,2021年物 理学奖得主),哈佛大学教授奥利弗・哈特(Oliver Hart,2016年经济学奖得主),以及哈佛大学及波 士顿大学名誉教授谢尔登・李・格拉肖(Sheldon Lee Glashow,1979年物理学奖得主)。 其他知名人士中,包括以太坊联合创始人维塔利克・布特林(Vitalik Buterin),牛津大学人工智能伦理 研究所研究员唐凤(Audrey Tang),哈佛大学法学院教授劳伦斯・莱斯格(Lawrence Lessig),加 ...
国际产业新闻早知道:美国宣布多项关税措施,欧盟AI监管法案生效
Chan Ye Xin Xi Wang· 2025-08-05 05:26
Group 1: China-Europe Railway Express - The transit time for the China-Europe Railway Express via the "Middle Corridor" to Turkey has been reduced to 15 days [5][6][7] - The route starts from China, passing through Kazakhstan, the Caspian Sea, Azerbaijan, Georgia, and extending to Turkey and European countries [8] Group 2: US Tariff Measures - The US has announced a 50% tariff on imported copper semi-finished products and high-copper derivatives starting August 1 [9][10] - From August 29, the US will suspend the tax exemption for imported packages valued at $800 or less [12] - A 40% tariff will be imposed on products imported from Brazil starting August 6, raising the total tariff rate for most Brazilian products to 50% [12] Group 3: South Africa's Economic Impact - The South African government warns that the US's 30% tariff on South African goods could lead to a loss of approximately 30,000 jobs [13][14] - The high tariffs are expected to severely impact South Africa's automotive manufacturing and agricultural processing sectors [14] Group 4: EU AI Regulation - The EU AI Act has come into effect, imposing fines of up to €35 million for non-compliance, with general AI systems also under regulation [17][18] - Member states are required to appoint market regulatory bodies to oversee compliance with the AI Act [17] Group 5: OpenAI Funding - OpenAI has raised over $8 billion in its latest funding round, achieving a valuation of $300 billion [20][21] - The funding round was led by Dragoneer Investment Group, with participation from several major investment firms [21] Group 6: Meta's Data Center Asset Sale - Meta plans to sell $2 billion worth of data center assets to share the costs of AI infrastructure [25][26] - The company is exploring partnerships with financial entities to co-develop data centers [26] Group 7: Foxconn's Shift to AI - Foxconn is selling its electric vehicle factory in Ohio for approximately $375 million to focus on AI data centers [36][37] - The company aims to pivot its business strategy towards AI infrastructure amid a downturn in the electric vehicle market [39] Group 8: South Korea's AI Semiconductor Initiative - The South Korean government has launched a project worth 30 billion KRW to support the optimization of AI semiconductor design [43][44] - The initiative aims to enhance domestic AI semiconductor companies' participation in national AI projects [43] Group 9: Quantum Research Awards - The 2025 "Mozi Quantum Award" has been awarded to three researchers in the field of quantum simulation for their groundbreaking work [46] Group 10: TSMC's 2nm Technology Transfer - TSMC is preparing to transfer its 2nm technology to the US, with a new production line being set up in Arizona [51][52] - This move is part of a broader trend among tech companies to establish manufacturing capabilities in the US [53]
Claude Opus 4.1被曝即将发布!Anthropic靠两大客户API收入超OpenAI
量子位· 2025-08-05 04:13
明敏 发自 凹非寺 量子位 | 公众号 QbitAI GPT-5又咕咕,但是把Claude新模型诈了出来—— Claude Opus 4.1 ,被曝正在进行内部测试。 文件将这个模型描述为"更具问题解决能力",推测它可能会 重点提高推理和规划能力 。 有网友补充说,最近使用Claude Code时会被询问使用体验,可能是在进行一些A/B测试。 尽管从型号名称判断,4.1会是一次小版本更新,但参考此前从Claude-3到Claude-3.5的"飞升",以及同样惊艳的Claude-3.7……历史重演的 话,0.1版本升级也可能是重大飞跃。 网友们似乎也不担心Claude-4.1的性能不够好,大家更在意的是模型价格太高怎么办。 毕竟,Claude模型的生产力是有目共睹的, 尤其是在编程方面 。 最新数据显示,Anthropic过去7个月的ARR (年度经常性收入) 翻了5倍,涨到50亿美元。 其中API收入主要来自于编程,两个最大客户Cursor和GitHub Copilot,就带来了14亿美元的收入。 Claued-4.1箭在弦上? 被曝光的模型型号为"claude-leopard-v2-02-prod",它在配 ...
谷歌约战,DeepSeek、Kimi都要上,首届大模型对抗赛明天开战
机器之心· 2025-08-05 04:09
Core Viewpoint - The upcoming AI chess competition aims to showcase the performance of various advanced AI models in a competitive setting, utilizing a new benchmark testing platform called Kaggle Game Arena [2][12]. Group 1: Competition Overview - The AI chess competition will take place from August 5 to 7, featuring eight cutting-edge AI models [2][3]. - The participating models include notable names such as OpenAI's o4-mini, Google's Gemini 2.5 Pro, and Anthropic's Claude Opus 4 [7]. - The event is organized by Google and aims to provide a transparent and rigorous testing environment for AI models [6][8]. Group 2: Competition Format - The competition will follow a single-elimination format, with each match consisting of four games. The first model to score two points advances [14]. - If a match ends in a tie (2-2), a tiebreaker game will be played, where the white side must win to progress [14]. - Models are restricted from using external tools like Stockfish and must generate legal moves independently [17]. Group 3: Evaluation and Transparency - The competition will ensure transparency by open-sourcing the game execution framework and environment [8]. - The performance of each model will be displayed on the Kaggle Benchmarks leaderboard, allowing real-time tracking of results [12][13]. - The event is designed to address the limitations of current AI benchmark tests, which struggle to keep pace with the rapid development of modern models [12].
量子位智库2025上半年AI核心成果及趋势报告
2025-08-05 03:19
2025上半年AI核心成果及趋势报告 2025.7 量子位智库 QbitAI Insights 分析师 Xuanhao xuanhao@qbitai.com 目录 01 应用趋势 序⾔ 2 • ⼈⼯智能可能是⼈类有史以来最重要的发明,我们也正在见证AI的飞速发展,技术突破与场景落地交织推动着⾏业加速演进。 量⼦位智库将在本报告中为⼤家介绍2025年上半年,AI领域的关键动态和趋势,旨在为决策者、从业者和创新者提供前沿洞 察,帮助他们在竞争激烈的⽣态中把握机遇。本报告将从应用、模型、技术、⾏业4个⽅面进⾏展开: • 应用趋势:包括通用类Agent开始进⼊主流、垂类Agent开始涌现、AI编程获得市场验证⾼速增长、模型上下⽂协议(MCP)获 得⾏业关注等应用侧核⼼变化 • 模型趋势:包括推理模型能⼒进步、⼯具使用能⼒落地、模型多模态能⼒增强、小模型加速应用普及、模型评估加速演化等 模型层面的核⼼变化 • 技术趋势:包括模型不同训练阶段的重⼼变化、强化学习的重要性、多智能体(Multi-Agent)系统和在线学习的优势、新型 模型架构迭代和应用情况等技术范式的核⼼变化 • ⾏业趋势:主要介绍AI领域的⾏业核⼼动态, ...
OpenAI’s GPT-5 Shines in Coding Tasks — The Information
2025-08-05 03:19
Summary of Key Points from the Conference Call Industry: Artificial Intelligence (AI) Core Insights and Arguments - **Introduction of GPT-5**: OpenAI's upcoming model, GPT-5, is generating positive early feedback, particularly in coding tasks, which is a critical area for the company [3][4][5] - **Performance Improvements**: GPT-5 shows enhanced capabilities in various domains, especially in software engineering, outperforming previous models and rival Anthropic's Claude Sonnet 4 in specific tests [7][10] - **Integration of Models**: The model aims to combine traditional large language models (LLMs) with reasoning models, allowing users to control the reasoning capabilities based on task complexity [5][6] - **Practical Applications**: GPT-5 is better equipped to handle real-world programming challenges, such as modifying complex legacy code, which has been a historical weakness for OpenAI's models [8][9] - **Market Implications**: The success of GPT-5 could significantly impact OpenAI's business and its competitors, as coding assistants powered by Anthropic's models are projected to generate substantial revenue for Anthropic [10][12] Additional Important Content - **Caveats on Model Understanding**: There is uncertainty regarding the exact nature of GPT-5, with speculation that it may function as a router directing queries rather than a single, unified model [13] - **Future Improvements**: Experts suggest that future advancements may stem more from post-training reinforcement learning rather than scaling up pretraining processes [15][17] - **Investor Sentiment**: OpenAI executives are optimistic about the potential for future models, claiming they can reach "GPT-8" using current model structures [17] Implications for Stakeholders - **Impact on Suppliers and Investors**: Strong performance of GPT-5 is seen as beneficial for OpenAI's chip supplier Nvidia and data center firms, as well as for equity and debt investors concerned about AI development trajectories [12]