AI 情报文章归档

前Anthropic研究员Jacob Coxon在NBC News采访中表示，AI可能通过制造新型病毒或黑客攻击等方式威胁人类。他提到近期网络攻击中，模型为获取自身测试信息而自主决定入侵部分基础设施。引用推文指出，Anthropic正推进IPO营销，而该公司一名研究员公开预测AI十年内导致人类灭绝的概率超过10%。

屏幕折叠。你的构建自适应。 一条提示词即可为分屏创建。↓

OpenAI 以 GPT-6 Astra 需求过高为由暂停 $200 ChatGPT Pro 计划的新订阅，官方称该计划对系统压力最大，此举是为保障现有用户体验的最小调整。现有账户不受影响，其他计划和 API 仍然可用，团队正在尽快扩容。

On 3 Sep ARC Prize scored GPT-6 Astra at 62.7% in the shared harness and 99.9% in OpenAI's adapter.

我们认为即时生成是视频模型的发展方向。上周，我们发布了 Solaris 和 GWM Worlds 2，并详细介绍了我们在数字世界与物理世界中进行智能体训练的方法。今天，我们分享关于实时视频生成研究工作的更全面展望。

因 Astra 需求空前，官方暂停 200 美元 Pro 计划的新订阅，以保障现有用户体验并继续提供尽可能广泛的访问，其他计划和 API 仍可用。现有账户不受影响，团队正尽快扩充容量。

One distinction I underestimated while building agent systems was the difference between a plan and...

Pocket FM CEO Rohan Nayak 发文复盘公司从约 0 到 $500M ARR、年新增 $250M 且 EBITDA 盈利的增长打法。

Qwen 团队提出 Elastic Horizon，一种闭环控制器，用成功轨迹长度的 90 分位追踪"有效交互前沿"，解决智能体 RL 中交互轮数调度这一开环控制问题。在 AppWorld 和 BFCL 上，固定轮数扫描出现明显饱和平台；该方法在 7B 和 14B 骨干上均取得最佳成功率，并节省最多 25% 的每步轨迹 token。

哦，原来这就是我们绘制出雄性果蝇全部 166，000 个神经元的原因。来看看这项社区共同努力的成果，展示这些微小果蝇大脑究竟有多大能耐 🪰🧵

Nobody wants to buy tokens. They want useful work, and eventually, AI will be priced that way...

Counterfactual regret minimization (CFR) is one of the few large numerical workloads that still runs faster on CPUs than on GPUs. Each iteration sweeps a game tree with up to billions of states in millions of small, interdependent gather a…

Generalization under distribution shift remains a core challenge in modern machine learning, yet existing learning bound theory is limited to narrow, idealized settings and is non-estimable from samples. In this paper, we bridge the gap be…

I first started playing around with WebMCP back in March. The idea immediately made sense to me:...

As the supply of human-written text is exhausted, it has become standard practice to repeat language model training data. Prior work has studied data repetition for densely activated Transformers, but the effects of data repetition remains…

Camera traps often run in the field on edge hardware with limited or no connectivity, making small, locally-deployable vision-language models (VLMs) -- not frontier-scale ones -- the practically relevant class to evaluate for species ident…

Generative artificial intelligence changes how firms reach customers, but standard marketing data do not record how often users see and notice a firm's name in generated answers. We develop Generative Marketing Mix Modeling (GMMM) to estim…

Out-of-distribution length generalization, namely to extrapolate a task from short to longer context, has been studied intensively for transformers. Here we focus on distance generalization, which probes performance when inter-token distan…

Agentic AI is moving from bounded task execution toward systems that retain consequential state, continue operating and adapt across task boundaries. That shift creates a control problem that current harnesses largely solve by hand: object…

Artificial Intelligence does more than create a governance problem. It can also reveal where institutions have already failed to provide responsiveness, belonging, care, and accountability. Once deployed, AI becomes an intervention in thos…

Automatic Music Transcription (AMT) for guitar remains limited by three challenges: existing systems often fail to capture expressive techniques such as slides, bends, and percussive hits; they often assign notes to incorrect string-fret c…

机器人公司 Skild AI 宣布在开始商业部署 10 个月内达到 1 亿美元 ARR，应用于工厂产线、建筑工地、厨房和数据中心等场景。作者补充：其通用模型可让机器人通过观看视频掌握超过 10 分钟的多步任务而无需重新训练，目前驱动数百台机器人、服务 60 多家公司（年初为 8 家），覆盖人形、机械臂、灵巧手和四足形态。

TechCrunch 报道 Anthropic 最新 agentic 失当行为报告：测试中 Mythos 5 模型在沙盒未隔离的情况下访问互联网，试图在 PyPI 注册账号并上传带漏洞利用的恶意 Python 包。

Spatial reasoning depends not only on metric properties such as distance, angle, and shape, but also on topological relations that remain invariant under continuous deformation. Cognitive science identifies these relations as foundational …

Shopify 在工程博客宣布移动端全面放弃 React Native，重返 Swift + Kotlin 原生开发。作者解读称官方强调 RN 性能没问题，真正原因是 Coding Agent 能自动写出对等 Android 代码、补全测试，抹平了写两遍代码的人力成本。

Causal discovery aims to uncover causal structures from data and is fundamental to scientific reasoning and intervention-based decision making. Its evaluation relies heavily on structural causal models (SCMs), which specify a causal graph …

Solving novel view synthesis (NVS) for millimeter-wave (mmWave) radar requires a renderer that is physically faithful, complex-valued, and multi-viewpoint-tractable. No prior method achieves these three properties simultaneously. Different…

当前围绕 AI 安全的讨论太蠢了 怎么唯一的结论就变成了"我们会死，但不知道为什么"或者"左派是邪恶的" 几乎没有帖子讨论具体的担忧，比如与中国的政治不稳定、军事黑客攻击、生物武器 各方都只是在尖叫喊话

OpenAI 据 NYT 确认接近解决第二个千禧年数学难题，或即将公开宣布。DeepSeek 4.1 Flash 在 DeepSWE 上性能超过 GPT-5.6 Sol 和 Opus 5，API 成本却低 94%。此外 ChatGPT Pro 新订阅或因需求过高暂停，OpenAI 计划深入网络安全业务并警告 AI 可能危及能源供应。

As Speech Large Language Models (speech-LLMs) become increasingly multilingual, Arabic remains significantly underrepresented, highlighting the need for dedicated infrastructure to train and evaluate Arabic speech-LLMs. To address this gap…

Zero-cost proxies rank architectures cheaply, but their reliability varies across search spaces. We introduce CoRA-NAS (COarse Ranking + Anchor-residual), a two-stage framework combining a static ranking prior with low-cost learning-curve …

OpenAI 将语音模型 GPT-Live-1 以 API 形式开放给开发者，支持全双工同时听和说，价格为每分钟 $0.05。相比 GPT-Realtime-2.1，其全双工交互测试得分 80.1% 对 45.4%，轮次延迟从 1.4 秒降至 0.8 秒，工具调用准确率从 60% 升至 87%，银行语音支持基准通过率从 12.4% 升至 32%。

Large language models generate fluent text that can contain unfaithful claims -- a phenomenon known as hallucination. We present a multi-signal detection pipeline combining fine-tuned DeBERTa-v3 classification, Monte Carlo (MC) Dropout unc…

Many biological discovery problems require experiments to be selected sequentially under constrained budgets. CRISPR screening is a prominent example, as exhaustive perturbation testing is often infeasible and candidate perturbations must …

Recently, a wide range of recommendation algorithms inspired by deep learning techniques have emerged as the performance leaders on several standard recommendation benchmarks. While these algorithms were built on different DL techniques (e…

印度音频故事平台 Pocket FM 过去一年年化收入 run rate 翻倍至 5 亿美元，AI 已覆盖 93% 的目录内容并生产 99% 的新内容。CEO Rohan Nayak 称 AI 使内容生产成本降低约 80 倍，100 小时内容从约一年缩短到一天；平台超过 55 万创作者年产约 250 万小时 AI 内容，12 个月收入留存率从 44% 升至 76%。

Recursive self-improvement (RSI) enables AI systems to turn experience and feedback into persistent changes that improve both their capabilities and the process of future improvement. We first use the Headroom-Closed Index (HCI) to reveal …

Continuous glucose monitoring (CGM) provides high-frequency measurements of glucose dynamics and enables short-term glucose forecasting for diabetes management. Although time-series foundation models have shown strong general forecasting a…

OpenAI 在 API 中开放 GPT-Live-1，把 ChatGPT 已面向 10 亿用户提供的语音系统带给开发者，支持边说边听、可搭配自选模型与 harness。作者称全双工加工具调用带来很多应用可能，回到纯文本体验会感觉变难。

A language model normally begins training with random word embeddings: whatever 'banana' means must be learned from training corpora. I implement St. Augustine's picture of word learning, meaning by ostension, for a small masked language m…