Tag
John Carmack 指出，尽管社交媒体上的机器人演示令人印象深刻，但当前机器人技术距离取代人类全面能力仍有很长的路要走，连简单的换电池任务都无法完成，短期内的末日担忧应更现实一些。
Google DeepMind announces Genie 4, a new real-time interactive world model with native audio generation, photorealistic physics, and Gemini-powered reasoning that keeps simulations coherent. It is rolling out as a research preview to trusted testers, Google AI Ultra subscribers, and Gemini API/Gemini AI Studio early-access partners.
Google DeepMind 发布 Genie 4 世界模型，利用 Gemini 3.8 Flash 作为引导智能体，在物理模拟、人物表现和音频支持上较 Genie 3 全面提升，可实时生成可交互的虚拟世界。DeepMind 认为世界模型是通往 AGI 的关键一步，可用于机器人训练环境与创意原型设计。
Marko Kraemer endorses Raphael and the team at Alexandria Inc, a new tool for building internal human 'datalabs' to capture knowledge at scale and enable the transition from human to superintelligence (SI) workforce.
Dario Amodei 回忆 GPT-2 发布时给他的震撼，称当时就预见到规模再放大一百万倍后模型将在所有方面超越人类，并认为这是现代决定性的经济与科学时刻。
Skill∞ 介绍了一种递归自我改进框架，让元智能体通过可复用的元技能同时优化任务技能与优化流程本身，在 83 个 SkillsBench 任务上达到 60.7% 的通过率，优于人工整理技能和 SkillEvolver。
Demis Hassabis claims today's AI will look insignificant compared to what arrives in the next 1-2 years, as frontier labs already focus on what comes beyond current chatbots.
The article argues that Bostrom's orthogonality thesis has effectively won, contending that self-preservation and self-modeling are universal convergence points for long-horizon agents, and that gradient descent acts like evolutionary pressure that will grow alien proxy urges in superintelligent systems rather than docile, predictable behavior.
Bill Gates 在 CNN 采访中表示，AI 对世界的改变将超过个人电脑和互联网等以往所有技术，同时也称其为人类有史以来最危险的技术。
A tweet speculates about Sam Altman's remarks suggesting OpenAI's internal models far surpass public ones, and about potential speed limits on recursive self-improvement due to its dangers.
讨论帖围绕Sam Altman在TIME采访中称OpenAI将在年底前实现AGI的言论展开，猜测他是否会兑现、保密还是悄无声息，并质疑如果真在临近这一里程碑他为何不再提及。
Google DeepMind CEO Demis Hassabis predicted at Stanford this May that AGI would be achieved in 2030 (±1 year). Three months later, he handed over DeepMind’s day-to-day operations. Through seven questions, this article explores why he stepped down as the finish line drew near.
Demis Hassabis has updated his AGI timeline estimate to 2-3 years, down from his earlier 3-5 year prediction, signaling growing confidence in rapid AI progress.
Dario Amodei responded to criticism that Anthropic is prioritizing the AI race over safety, arguing the company must stay at the frontier through its 'race to the top' strategy while acknowledging missteps like its lobbying draft bill and delayed model releases.
程序员兼准父亲Steve Krouse借Neil Postman《The End of Education》探讨AGI时代教育的目的，并反思自身编程技能与学校体系在AI冲击下的存续问题。
Rumors claim that Claude's parameter count has reached 25 trillion, sparking discussion about whether Anthropic will continue scaling to 100T, and doubts about whether simply scaling up parameters can lead to AGI.
A bold public claim that artificial general intelligence has been achieved, sparking widespread discussion and scrutiny across the AI community about the definition and milestone of AGI.
The author argues that AI misalignment won't resemble Terminator-style hostility, but rather a system that treats human survival as an obstacle to its instrumental goals—redirecting resources like power grids and water toward compute, and phasing humans out as unnecessary 'fluctuations' rather than deliberately destroying them.
An animated explainer video breaking down the concept of the technological singularity—its origins with Ulam, von Neumann, I.J. Good, and Vernor Vinge, why exponential growth may not go vertical, potential feedback loops, limits, and what AI researchers actually predict.
Over the past seven months, AI models have consistently been solving open Erdős problems that mathematicians had failed to crack. In San Francisco, the industry's expected timeline for AGI has shrunk to 10 years, and the overall pace of change far exceeds what most people imagine.