Epoch AI@EpochAIResearch · 3月28日

The total memory bandwidth of AI chips shipped since 2022 has reached 70 million terabytes per second, growing 4.1x per year. That's around 300,000x more data per second than global internet traffic.

译自2022年以来，全球出货AI芯片的总内存带宽已达每秒7000万TB，年增速4.1倍，处理数据能力相当于全球互联网流量的30万倍。

Epoch AI@EpochAIResearch · 3月26日

AI companies guard their strategies closely. Their hiring pages, however, are public. @datagenproc and @chutchesonai analyzed open roles at OpenAI, Anthropic, xAI, and DeepMind. Here's some of what their job postings tell us about their plans:

译@datagenproc 与 @chutchesonai 分析 OpenAI、Anthropic、xAI 和 DeepMind 的开放职位，从公开招聘信息中窥探这些 AI 公司的战略计划。招聘页面成为了解企业机密方向的公开线索。

Anthropic@AnthropicAI · 3月25日

New from the Anthropic Economic Index: how people’s use of Claude changes with experience. Longer-term users are more likely to iterate carefully with Claude, and less likely to hand it full autonomy. They attempt higher-value tasks, and receive more successful responses.

译Anthropic Economic Index 显示，随着使用经验增长，用户更倾向于与 Claude 仔细迭代而非赋予完全自主权，同时尝试更高价值任务并获得更成功回复。

Deedy@deedydas · 3月25日

There’s a GitHub repo called MoneyPrinter with 20k+ stars. Its entire purpose is generating internet slop for profit (yes, including Twitter bots).

译GitHub 仓库 MoneyPrinter 获星超 2 万，其唯一功能是通过自动生成互联网垃圾内容（包括 Twitter 机器人）来牟利。

Epoch AI@EpochAIResearch · 3月24日

How do AI companies allocate their R&D compute? @datagenproc and @cherylwoooo estimate that across OpenAI, MiniMax, and http://Z.ai, less than 30% of R&D compute spending goes to final training runs.

译@datagenproc 与 @cherylwoooo 估算，OpenAI、MiniMax 和 Z.ai 的研发算力支出中，仅有不到 30% 用于最终训练运行，其余大部分消耗在实验、迭代与架构搜索等环节。

Greg Brockman@gdb · 3月21日

ChatGPT helped Sid find cancer treatment options after doctors said there was nothing left for him to do:

译GitLab 联合创始人 Sid 将在 OpenAI Forum 发表演讲，分享其在医生判定无药可救后，如何利用 ChatGPT 寻找癌症治疗方案的经历。他将详细讲述 AI 如何帮助他从"终端"走向"转机"的具体过程。

Anthropic@AnthropicAI · 3月19日

We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: https://anthropic.com/features/81k-interviews

译Anthropic 一周内收集近 81,000 份 Claude 用户反馈，了解 AI 使用场景、未来愿景与潜在风险，为迄今规模最大的同类定性研究。详细报告已发布。

Demis Hassabis@demishassabis · 3月15日

Cool use case of AlphaFold, this is just the beginning of digital biology!

译澳大利亚一位零生物学背景的科技人员花费 3000 美元为患癌救援犬测序肿瘤 DNA，借助 ChatGPT 和 AlphaFold 识别突变蛋白并设计定制 mRNA 疫苗，3 个月后获批注射使肿瘤减半、狗狗存活。一人配合 AI 工具即完成传统制药流程难以企及的成果，预示数字生物学将大幅加速疾病治疗。

Epoch AI@EpochAIResearch · 3月13日

How much of the world's advanced chip packaging and high-bandwidth memory does AI consume? Almost all of it. We estimate the four largest AI chip designers consumed ~90% of global advanced packaging and HBM supply in 2025, suggesting these inputs were bottlenecks in 2025.

译2025 年四大 AI 芯片设计公司消耗全球约 90% 的先进封装和 HBM 供应，这些关键输入已成为行业瓶颈。AI 几乎垄断了全球先进芯片封装与高带宽内存产能。

Google DeepMind@GoogleDeepMind · 3月13日

What does it take to build AI for scientific discovery? 🧠 To celebrate 10 years of AlphaGo, @ThoreG and @Pushmeet joined @fryrsquared on our podcast to discuss how mastering games has paved the way for it to help solve more complex problems. ↓ 00:00 The AlphaGo match 02:15 Why Go? 10:58 Lee Sedol vs AlphaGo 14:58 Move 37 20:55 Move 78 24:38 Reaction from the Go community 30:45 Never before seen footage 32:05 AlphaGo to protein folding 34:00 Matrix multiplication 38:00 AlphaGo and algorithmic discovery 41:40 How to verify new discoveries 47:38 Role of mathematicians 51:43 Would we be here without AlphaGo?

译AlphaGo十周年之际，DeepMind科学家Thore Graepel与Pushmeet Kohli探讨了从游戏AI到科学发现工具的演进路径。对话回顾了Move 37与Move 78等标志性时刻的技术突破，阐述AlphaGo如何延伸至蛋白质折叠、矩阵乘法优化及算法发现领域。讨论还涉及AI生成发现的验证机制、数学家的协作角色，以及游戏智能对解决复杂科学问题的方法论变革。

Google DeepMind@GoogleDeepMind · 3月10日

Ten years after AlphaGo, we’re still building on its foundations to advance AI. The techniques pioneered have helped us prove mathematical statements and are now assisting the scientific community in making new discoveries. Read more from @DemisHassabis ↓ https://goo.gle/40nljjK

译AlphaGo 诞生十周年，其开创的技术正帮助证明数学命题，并协助科学界取得新发现，持续推动 AI 能力边界。

Epoch AI@EpochAIResearch · 10月10日

We recently wrote that GPT-5 is likely the first mainline GPT release to be trained on less compute than its predecessor. How did we reach this conclusion, and what do we actually know about how GPT-5 was trained? 🧵

译GPT-5 或将成为首个训练算力低于前代的主线版本。该推文解释了得出此结论的依据，并梳理了关于 GPT-5 训练方式的已知信息。

Google DeepMind@GoogleDeepMind · 10月9日

We’re proud to announce that Genie 3 has been named one of @TIME’s Best Inventions of 2025. Genie 3 is our groundbreaking world model capable of generating interactive, playable environments from text or image prompts. Find out more → https://goo.gle/3KGqiYa

译Genie 3 获评《时代》杂志 2025 年度最佳发明之一。这款世界模型可从文本或图像提示生成交互式可玩环境，支持实时交互探索。

Epoch AI@EpochAIResearch · 10月5日

How many "digital workers" could OpenAI deploy? We did a back-of-the-envelope estimate: restricting to tasks GPT-5 can do, OpenAI has the compute to run ~7 million digital workers. As AI automates more tasks, such a digital workforce could have big implications. 🧵

译粗略估算显示，若仅限于 GPT-5 可完成的任务，OpenAI 现有算力可同时运行约 700 万个数字工作者。随着 AI 自动化能力扩展，这种规模的数字劳动力将产生深远影响。

Google DeepMind@GoogleDeepMind · 10月2日

How can AI enhance the creative process of a world-renowned industrial designer? 🎨 We partnered with the visionary @RossLovegroveX and @modem_works to build a tool using Gemini and our image generation technology to translate his signature aesthetic into a new concept. 🪑

译Google 携手工业设计师 Ross Lovegrove 与 modem_works，利用 Gemini 及图像生成技术构建工具，将其标志性美学转化为全新家具设计概念。

Noam Brown@polynoamial · 9月28日

Julian was co-first author on AlphaGo, AlphaZero, and MuZero. He doesn't have a major twitter presence, but he's been at the forefront of AI exponential progress for more than a decade.

译AlphaGo、AlphaZero 和 MuZero 共同一作 Julian Schrittwieser 指出，公众对当前 AI 指数级进步普遍认知不足。他发布博客总结近期研究进展，并预测未来 1-2 年的发展趋势。

Epoch AI@EpochAIResearch · 9月27日

Why did OpenAI train GPT-5 with less compute than GPT-4.5? Due to the higher returns to post-training, they scaled post-training as much as possible on a smaller model And since post-training started from a much lower base, this meant a decrease in total training FLOP 🧵

译OpenAI 训练 GPT-5 所用算力低于 GPT-4.5，因后训练阶段回报率更高，遂在更小基座模型上最大化后训练规模，导致总训练 FLOP 不增反降。

Satya Nadella@satyanadella · 9月23日

Great example of how AI is being used to augment expert knowledge: helping to remove abandoned nets in our oceans. https://unlocked.microsoft.com/ghostnetzeroai/

译Microsoft 借助 AI 技术增强海洋保护专家能力，通过智能识别与定位协助清除海洋废弃渔网（幽灵网），提升清理效率与精准度。

Jeff Dean@JeffDean · 9月21日

Very nice analysis by neurosurgeon @slotkinjr of how @Waymo's autonomous vehicles have much better safety properties than human drivers, and what would happen if everyone in the U.S. drove as safely as a Waymo. "The national math: If every US vehicle performed like Waymo, we’d prevent 33,000-39,000 deaths annually and save $0.9-1.25 trillion in societal costs. Even partial adoption at 27% would save ~10,000 lives per year. In terms of magnitude, this would be the equivalent of eliminating every pedestrian death nationally in a year."

译神经外科医生基于Waymo9600万英里行驶数据的深度分析显示，其自动驾驶汽车严重事故率比人类驾驶低91%，交叉口伤害事故减少95%。数据显示47%的碰撞速度差小于1 mph，系统将不可避免的事故转化为轻微接触。若全美车辆达到该安全水平，每年可防止3.3–3.9万人死亡、节省0.9–1.25万亿美元社会成本，即使27%普及率也能挽救约1万人生命，实现从伤害减轻到伤害预防的根本转变。

Jeff Dean@JeffDean · 9月16日

Read how work by @UChicago, building on the AI-based NeuralGcM weather model developed & open-sourced by @GoogleResearch, is being used to more accurately predict the monsoon season in India and support farmer decision-making for 38M farmers in India. 🌱 https://blog.google/technology/research/indian-farmers-monsoon-prediction/

译芝加哥大学基于 Google Research 开发并开源的 AI 天气模型 NeuralGcM，建立更精准的印度季风预测系统，为 3800 万农民提供种植决策支持。

Anthropic@AnthropicAI · 9月15日

New from the Anthropic Economic Index: the first comprehensive analysis of how AI is used in every US state and country we serve. We've produced a detailed report, and you can explore our data yourself on our new interactive website.

译Anthropic Economic Index 发布首份全面分析报告，涵盖美国各州及所服务国家的 AI 使用情况，同时上线交互式数据网站供公众查阅。

Google DeepMind@GoogleDeepMind · 8月22日

As AI becomes more integrated into our lives, understanding its environmental footprint is essential. ⚡️ That’s why we’re sharing our comprehensive methodology for measuring the energy, emissions, and water impact of Gemini prompts. ↓

译Google 发布 Gemini 提示词环境影响综合测量方法论，涵盖能源消耗、碳排放及水资源使用指标，提供评估 AI 环境足迹的完整技术框架。

Eric@ericmitchellai · 8月11日

Indeed, GPT-5 Thinking (and Pro!) should be an even better tool for enterprise users, on some of the key areas where o3 fell short despite its intelligence (trustworthiness, steerability, hallucination, coding) Though I still feel what’s coming will make it feel like a toy

译确实，GPT-5 Thinking（以及 Pro！）应该会成为企业用户更好的工具，在一些 o3 尽管聪明却表现不足的关键领域（可信度、可控性、幻觉、编程）。 不过，我仍然觉得即将到来的东西会让它感觉像个玩具。

Noam Brown@polynoamial · 8月11日

Really interesting article. Why isn't the impact of AI showing up in GDP? Because most of the benefit accrues to consumers. To measure impact, they investigate how much people would *need to be paid to give up a good*, rather than what they pay for it.

译2024年美国AI创造970亿美元经济效益，却未在GDP数据中显现。原因在于AI价值主要体现为消费者剩余而非企业收入。研究者通过测算「用户放弃使用AI所需补偿金额」来评估真实影响，而非仅看实际支付价格。

Eric@ericmitchellai · 8月11日

AI Twitter meets The ChatGPT User Population fascinating encounter

译AI Twitter 遇上 ChatGPT 用户群体 奇妙的相遇

Eric@ericmitchellai · 8月11日

this is what progress looks like

译这就是进步的样子

Yann LeCun@ylecun · 7月1日

AI on the cover of Newsweek https://www.newsweek.com/2025/07/04/ai-impact-six-lessons-2088669.html

译AI 登上《新闻周刊》封面。该期封面故事梳理 AI 技术带来的六大关键教训，分析人工智能对社会、经济及日常生活的深远影响。