大模型
www.aibase.com ·
DeepMind 首席科学家 Shane Legg 反驳黄仁勋：Astra 尚未达到 OpenAI 自定的 AGI 标准
针对 NVIDIA CEO 黄仁勋公开祝贺 OpenAI 发布 Astra 并称“AGI 已到来”，Google DeepMind 联合创始人兼首席 AGI 科学家 Shane Legg 在采访中公开反驳，认为现在谈进入 AGI 时代为时过早。他引用 OpenAI 章程中对 AGI 的定义——“在大多数具有经济价值的任务上超越人类的高度自主系统”，指出 GPT-6 Astra 虽能完成许多复杂任务，但是否真正具备完成大多数有认知价值且具经济意义任务的能力仍高度不确定。Legg 维持其长期谨慎预测：2028 年前实现“最小 AGI”的概率约为 50%；DeepMind 对 AGI 的定义更全面，视为具备人脑全部认知能力的系统。他同时警告 AI 能力提升速度过快，安全机制与控制框架不能落后。DeepMind 另宣布成立 DeepMind Institute，由 Shane Legg、董事长 Demis Hassabis 与 Google 高级副总裁 James Manyika 领导，研究 AGI 快速演进对科研、教育、社会结构和人类福祉的影响。
AI推荐理由
内容为 DeepMind 联合创始人 Shane Legg 对黄仁勋称 OpenAI Astra 已达 AGI 的公开反驳，并给出 OpenAI 章程中 AGI 定义、Legg 对 2028 年前实现最小 AGI 约 50% 概率的长期预测，以及 DeepMind 成立 DeepMind Institute 的信息，属于具体人物观点与机构动态。但来源为 AIbase 聚合转述，非原始采访，缺少直接引语与可核验出处，且标题带有对抗性炒作色彩，故质量分中等；AGI 判定标准与安全治理议题对行业有讨论价值，兴趣分略高。
62
正文
DeepMind Chief Scientist Confronts Huang Renxun: Stop the Hype, Astra Hasn't Even Reached OpenAI's AGI Standards
AIbase基地
Published inAI News · 4 min read · Sep 17, 2026
6
The AI industry has recently seen a heated debate on whether "General Artificial Intelligence (AGI)" has already arrived. In response to NVIDIA CEO Jensen Huang's public congratulations on OpenAI's launch of Astra and his statement that "AGI has arrived," Shane Legg, co-founder and chief AGI scientist at Google DeepMind, recently publicly refuted this view in an interview, pointing out that it is too early to talk about entering the AGI era.
Regarding the criteria for evaluation, Shane Legg pointed out that OpenAI had previously defined AGI in its bylaws as "a highly autonomous system that surpasses humans in most economically valuable tasks." Although the latest GPT-6Astra can perform many impressive complex tasks, it remains highly uncertain whether it truly has the ability to perform most cognitively valuable economically significant tasks. As one of the early proponents of the "AGI" concept, Legg maintains his long-term cautious prediction that the probability of humans achieving "minimum AGI" before 2028 is approximately 50%. In contrast, DeepMind's definition of AGI is more comprehensive, viewing it as a system with all cognitive abilities of the human brain.
Although he casts cold water on the idea that AGI is not yet truly here, Legg still keeps a high level of vigilance towards the rapid advancement of AI technology. He warned that the speed of improvement in AI capabilities is too fast, and the corresponding safety mechanisms and control frameworks must not lag behind. If there is no sufficient understanding and risk management, humans may be passively confronted with more powerful intelligent systems without any preparation.
To further address this industry trend, DeepMind officially announced the establishment of the DeepMind Institute. The institution will be led by Shane Legg, DeepMind Chairman Demis Hassabis, and Google Senior Vice President James Manyika. In the future, it will focus on studying the profound impact of the rapid evolution of AGI on scientific research, education systems, social structures, and the well-being of all humanity.
AGIOpenAIDeepMindArtificialIntelligence
This article is from AIbase Daily
Scan to view
Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications.
—— Created by the AIbase Daily Team
© Copyright AIbase Base 2024, Click to View Source -