通用人工智能(AGI)

Search documents
通用人工智能何时到来?
腾讯研究院· 2025-05-12 08:11
闫德利 腾讯研究院资深专家 一、AI已在诸多任务领域超越人类 AI发展日新月异,在许多任务上已经陆续超越人类基线水平。如2015年图像分类,2018年中等水平阅读 理解,2020年视觉推理、英语语言理解,2023年多任务语言理解、竞赛级数学,2024年博士级科学问 题。下图所示的8项关键任务技能中,AI仅在多模态理解和推理能力上还略逊人类一筹,但从2023年开 始就加速提升。我们有望很快见证AI 能力在现有主流基准上"全部超越人类水平"的奇点时刻。 图 选定的 AI 指数技术性能基准与人类表现对比 二、AGI的终极目标或于年内实现 我们已经构建了无数在特定任务上超越人类水平的AI系统,但它们缺乏通用性,无法应对超出预定任务 之外的问题,尚处于"狭义人工智能 (Narrow AI) "阶段。随着AI性能的大幅提升,具备跨领域能力、在 多个方面媲美甚至超越人类的、更强大的AI被提上日程。 人们常将之命名为"通用人工智能(AGI)" 。 各国高度重视AGI。2023年4月28日中共中央政治局会议提出:"要重视通用人工智能发展";英国《国家 人工智能战略》 (2021 ) 对AGI进行了专门强调,指出"必须认真对待A ...
刘强东穿猪猪侠T恤现身日本:背后还有二维码;消息人士称马云回归绝不可能,其从来没离开过;王兴兴回应人形机器人产业泡沫化丨邦早报
创业邦· 2025-05-11 01:06
完整早报音频,请点击标题下方小耳机收听 【消息人士称马云回归是绝不可能的,马云是创始人也从来没离开过】 5月10日,记者从多位阿里巴巴内 部人士处获悉,阿里巴巴已全面打通了内网(内部论坛)权限,并调整员工跨业务流动机制、启动工牌 焕新。5月9日,针对"马云5月10日回归阿里巴巴,将要重启大集团模式"相关消息,记者向阿里巴巴集团 求证,截至发稿,暂无回应。阿里巴巴内部人士向记者表示,"马云回归是绝不可能的,并且马云是创始 人也从来没离开过"。(贝壳财经) 【苹果渠道官方调价:iPhone16 Pro最高降价176美元,Pro Max全容量降价160美元】 苹果5月10向渠道 商下发了调价通知,这也是苹果首次在周六宣布调价,给相当大一部分批发商打了个措手不及。苹果 iPhone16 Pro Max所有容量版本降价160美元,对应1313.06元人民币;而 iPhone16 Pro的128GB版本降价 176美元(对应1445.27元人民币),其他版本同样降价160美元。今年的618活动将于5月13日开始,因此 有渠道商认为苹果此次调价就是在为即将到来的618大促做准备。(IT之家) | | | SI Rebate ...
雷军隔月首发声:创立小米以来最难的日子;Altman向马斯克服软「AGI比恩怨重要」;酷玩等明星抗议AI滥用
Sou Hu Cai Jing· 2025-05-11 00:11
根据美国联邦通信委员会 (FCC) 发给 SpaceX 公司代表的电子邮件,SpaceX 即将发射星际飞船 9 号 (Starship Flight 9)。 来自 FCC 代表的信函显示 SpaceX 星际飞船第九次测试发射即将启动 雷军时隔月余首度发声:系小米创办以来最艰难的一段时间,感谢大家的关心和支持 自上个月底在测试场测试了星际飞船上面级火箭并两次点火以来,该公司加快了 9 号飞行的发射速度。SpaceX 已将星际飞船再次运回测试场,这似乎是在 为 9 号飞行前的另一次静态点火做准备。 5 月 10 日消息,小米集团创始人、董事长兼 CEO 雷军在其个人微博发布动态,这是其自小米 SU7 事故后首次公开表态(平时会有运营人员打理账号), 也是其停止互动一个多月后的明牌回归。 雷军表示,这一个多月的时间是他创办小米以来最艰难的一段时间,情绪长期低落,所以也暂停了在社交媒体上的互动。好在有大家的关心和支持,他也因 此逐渐找回勇气和信心、恢复状态。(来源:IT 之家) 当地媒体的画面显示,SpaceX 已将上面级飞船运送至其地面测试设施,很可能进行静态点火测试。如果静态点火测试成功,SpaceX 将能够在星 ...
【中国那些事儿】俄专家:中俄人工智能合作跨越“小院高墙”,构建公平世界科技新秩序
Huan Qiu Wang Zi Xun· 2025-05-10 05:18
科洛宁还提到,人工智能的飞速发展引发了人们对滥用人工智能和通用人工智能的担忧。一些国家利用 其在人工智能领域的主导地位,对他国进行胁迫,阻挠它们与被视为威胁的国家开展合作。鉴于此,那 些希望建立公平世界秩序的国家需加深彼此间的合作,例如在金砖国家框架下,秉持互惠互利的原则, 共同推动全球科技治理体系的完善。 科洛宁强调,俄罗斯科学界对与中国以及其他志同道合的国家携手,共同推动全球在人工智能和通用人 工智能领域的协调发展与有效治理持开放态度。欢迎其他国家参与俄罗斯AGI社区研讨会等开放活动, 以及数学AI等联合会议,并期待各方逐步完善人工智能技术管理的联合战略。 另据相关报道,由外国顶尖专家组成的"瓦尔代"国际辩论俱乐部(Valdai Discussion Club)项目主任季 莫费·博尔达切夫(Timofei Bordachev)同样指出,人工智能是前沿科技领域,中国和俄罗斯都具备相 应技术和人才,两国可以通过在这一领域的合作,树立起科技合作的典范,并为全球南方国家在科学、 文化和教育领域的解放贡献力量。这不仅将为两国开辟全新的合作领域,还将切实推动南南合作,这对 于构建一个更加平衡、公正的世界秩序至关重要。 ...
资本研·观|AI代理的概念及其在金融领域的发展
野村东方国际证券· 2025-05-09 09:37
AI代理的概念及其在金融领域的发展 -应用于财富管理的潜力- 作者:野村资本市场研究所 坂上 聖奈 要点 1. 随着大型语言模型在自然语言处理领域取得显著突破,AI代理受到了广发关注。例如,摩根 士丹利在2025年至2027年中长期经营计划中指出,AI代理的潜力是其技术战略的一个重点领域。 2. 简而言之,AI代理就是一种能够自主决策的AI模型。本文以穆迪和Coinbase为例,介绍AI代 理的应用。尽管两家公司根据其业务战略对AI代理的应用采取了不同的方式,但在增强AI自主性方 面却有着共通之处。 3. 穆迪将AI代理视为强化现有服务的工具,通过数据驱动的决策效率来提升客户体验。另一方 面,Coinbase则将AI代理定位为区块链交易的手段,旨在通过交易的自动化和自主化来提升客户体 验,甚至改变经济活动的性质。 4. 随着客户需求的多样化、依赖特定人员的业务流程优化、以及降低人力和成本等因素的驱使 下,金融机构对AI代理的需求在不断扩大。在财富管理业务中,AI代理的应用也被认为是一个重要 的候选方向。具体包括:(1)为客户提供最佳投资建议以实现特定的财务目标;(2)提升托管服 务水平;(3)扩大数字资产交易 ...
OpenAI人事大调整,技术理想主义又回来了
2 1 Shi Ji Jing Ji Bao Dao· 2025-05-08 13:03
21世纪经济报道记者孔海丽 北京报道 六年试错,技术理想主义正在试图寻回现实土壤。 北京时间5月8日,OpenAI宣布重大人事调整,话事人塞姆·奥特曼(Sam·Altman)将应用业务的CEO权 杖转交给菲吉·西莫(Fidji Simo),他自己则把更多精力投入到研究、算力和安全等核心技术领域。 就在2天之前,OpenAI摇摆许久的公司架构,也终于明朗化,终止之前转向营利性公司的计划,而是引 入PBC(公益公司)控制模式,仍由非营利实体掌控核心治理权。 在经历商业化与理想主义的激烈博弈后,这家以"确保AGI造福全人类"为使命的明星AI公司,似乎找到 了新的平衡点,愿景得以与商业化并存,回归创业初心。 非营利初心重新觉醒 根据声明,OpenAI原本为应对资金黑洞而设立的营利性有限责任公司(LLC)将转变成PBC,PBC由非 营利组织控制,即形成"非营利组织-PBC"的架构,非营利组织保留绝对控制权。 在OpenAI之前,Anthropic、X.ai等其他明星AI公司以及Patagonia等许多目标驱动型公司已经采纳了PBC 架构。OpenAI认为,这对他们也很有意义,既避免了完全商业化带来的伦理风险,也能够通过 ...
OpenAI悬着的心终于“死了”
Hu Xiu· 2025-05-08 11:31
Core Viewpoint - OpenAI is undergoing a structural change to maintain control by its non-profit organization while transitioning its for-profit entity into a Public Benefit Corporation (PBC), emphasizing its mission to benefit humanity through artificial intelligence [4][10][15]. Group 1: Structural Changes - OpenAI's for-profit entity will transform into a PBC, with the non-profit organization retaining significant ownership and control [4][15]. - The previous plan to separate the non-profit and for-profit entities has failed, leading to a re-emphasis on the non-profit's control over OpenAI [4][5]. - The non-profit organization aims to ensure that the PBC operates in alignment with its mission, which includes providing AI benefits to a broad audience [16][17]. Group 2: Financial Context - OpenAI completed a $40 billion private funding round, achieving a valuation of $300 billion, with a condition to restructure into a for-profit company by the end of the year [5][6]. - Microsoft, a major investor with $13.75 billion invested, is a significant player in the restructuring discussions, seeking to protect its interests [6][10]. - OpenAI plans to reduce the revenue-sharing percentage with Microsoft from 20% to 10% by 2030, indicating a shift in financial strategy [6][10]. Group 3: Industry Pressures - OpenAI's commercialization efforts have faced legal challenges and opposition from figures like Elon Musk and other industry stakeholders [5][10]. - The restructuring has been met with academic pressure, as notable figures in AI have called for a halt to the transition plans [5][10]. - The company is navigating complex dynamics between maintaining its mission and addressing investor expectations and regulatory requirements [10][16]. Group 4: Mission and Vision - OpenAI's mission is to ensure that artificial general intelligence (AGI) benefits all of humanity, with a focus on democratizing access to AI tools [12][13]. - The organization aims to provide powerful AI tools while ensuring that their development aligns with ethical considerations and societal benefits [12][14]. - OpenAI envisions a future where its AI capabilities can significantly enhance productivity and address global challenges across various sectors [14][17].
对话阶跃星辰CEO姜大昕:两年发布16款多模态模型,DeepSeek证明投流模式不成立|钛媒体AGI
Tai Mei Ti A P P· 2025-05-08 08:33
Core Insights - The CEO of Leap AI, Jiang Daxin, announced the upcoming release of the full version of the inference model Step R1 and a more advanced Step image editing model within the next two to three months [2] - Leap AI emphasizes the importance of "multi-modal understanding and generation integration" as a key path towards developing a world model and progressing towards Artificial General Intelligence (AGI) [2][3] - Jiang Daxin highlighted that traditional traffic investment logic in AI product growth needs reevaluation, as demonstrated by the performance of DeepSeek and other AI products [2] Company Overview - Leap AI, founded in April 2023, is a leading startup focused on developing general AI models and has released the Step series of foundational models [5] - The company has raised several hundred million dollars in its B-round financing, with key investors including Shanghai State-owned Capital Investment Co., Tencent Investment, and Qiming Venture Partners [5] - Leap AI has launched 22 self-developed foundational models, with over 70% being multi-modal models, establishing itself as a leader in the multi-modal AI space [5] Product Development - The company has made significant advancements in multi-modal models, covering various applications such as image understanding, video generation, and music generation [5][7] - Leap AI has established deep collaborations with industry leaders in automotive, mobile, and IoT sectors, enhancing its product capabilities [7] - Recent product releases include the Step R-mini inference model and open-sourced video models, indicating a commitment to expanding its model capabilities [7] Strategic Focus - Leap AI is concentrating on developing intelligent terminal agents that enhance user experience by understanding environmental contexts [11] - The company believes that the integration of pre-trained foundational models with reinforcement learning can significantly improve reasoning capabilities [12] - Jiang Daxin asserts that achieving AGI requires a multi-modal approach, as human intelligence is diverse and relies on various modalities [8] Competitive Positioning - Leap AI differentiates itself from competitors like OpenAI and Google by focusing on foundational model development and multi-modal capabilities [13] - The company aims to create an ecosystem that integrates models with intelligent agents, bridging cloud and edge computing [13]
美怎么也没料到,中方动真格了?阿里开源模型发布,特朗普慌了
Sou Hu Cai Jing· 2025-05-08 01:05
Core Viewpoint - Alibaba's announcement of the open-source Qwen3 model marks a significant milestone in the global AI landscape, showcasing China's strong capabilities in AI innovation and potentially shifting the competitive dynamics with the U.S. [1][6][9] Industry Summary - The Qwen3 model integrates "fast thinking" and "slow thinking" capabilities through a "Mixture of Experts (MoE)" architecture, allowing for efficient processing of both simple and complex tasks while reducing computational costs [3][5]. - Following the release of DeepSeek's R1 model, several Chinese tech companies have launched cost-effective AI models, including Baidu's Wenxin Yiyan 4.5 and Volcano Engine's Doubao 1.5, contributing to a wave of AI model upgrades in the domestic market [3][5]. - Qwen3 has demonstrated impressive performance in benchmark tests, achieving a score of 81.5 in the AIME25 assessment and outperforming competitors like Grok3 and OpenAI's models in various evaluations [5][6]. Company Summary - Alibaba is strategically positioning itself towards achieving Artificial General Intelligence (AGI), with plans to invest over 380 billion RMB in cloud and AI hardware infrastructure over the next three years, surpassing the total investment of the past decade [6]. - The open-sourcing of Qwen3 is a crucial step in Alibaba's journey towards AGI, with over 200 models already open-sourced and a global download count exceeding 300 million [6][9]. - The release of Qwen3 enhances China's standing in the global AI arena, providing robust technical support for developers and businesses, and potentially narrowing the gap with the U.S. in AI technology [9].
扎克伯格深度专访:怼苹果,夸DeepSeek,聊AI开源痛点
Sou Hu Cai Jing· 2025-05-07 15:28
智东西 编译 | 金碧辉 编辑 | 程茜 智东西5月7日消息,据Stratechery报道,4月28日下午,社交媒体平台Stratechery的创始人、记者本・汤普森(Ben Thompson)在Meta总部对 Meta的创始人、董事会主席兼CEO马克・扎克伯格(Mark Zuckerberg)进行了专访。 从访谈中得知Meta在AI领域的布局以开源大语言模型Llama为核心,实现了生成文本、数学推理、代码生成等能力的跃升,其关键创新在于仅 使用公开数据集训练,并通过1.4万亿tokens的数据量弥补参数规模的不足。这种"小模型大智慧"的技术路径,印证了扎克伯格"效率优先"的AI 哲学。 这是继2021年10月和2022年10月之后,Stratechery对扎克伯格的又一次访谈。访谈前,Stratechery已了解了LlamaCon(Meta的新开发者大会) 的部分公告,并体验了新的Meta AI应用。 一、元宇宙概念持续推进,LlamaCon因开源模型需求诞生 Stratechery在体验Meta AI应用后,扎克伯格透露了这个新应用的更多细节,Meta AI已实现月活跃用户近10亿,成为全球用户规模最大的 ...