2026-09-16 科技日报

扫描 539 篇候选内容 · 覆盖 650 个信息源 · 纳入 523 篇

📌 今日要点

• Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking
• Your Agent Aced the Task. Will It Do It Again?
• Orchestration and Execution: How JONI Approaches the Agent Layer
• How I’m Using Google Opal for Even More AI Automations
• 5 Free Microsoft GitHub Courses to Learn Data Science and Artificial Intelligence
• AI for Societal Impact

📝 科技简讯

• Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking — Google DeepMind
• Your Agent Aced the Task. Will It Do It Again? — Hugging Face Blog
• Orchestration and Execution: How JONI Approaches the Agent Layer — KDnuggets
• How I’m Using Google Opal for Even More AI Automations — KDnuggets
• 5 Free Microsoft GitHub Courses to Learn Data Science and Artificial Intelligence — KDnuggets
• AI for Societal Impact — Google AI Blog
• Building AI to accelerate science and improve lives — Google AI Blog
• AI for everyone in every language — Google AI Blog
• New insights from Google’s AI & Economy ATLAS — Google AI Blog
• Is METR A Meaningful Check On Anthropic? — LessWrong
• Why would AI cause human extinction? — LessWrong
• Inoculation Midtraining with Learned Neologisms — LessWrong
• Quick notes from teaching technical profiles how to talk in public — LessWrong
• Shallow Beliefs: Midtraining does not inoculate against EM from reward hacking — LessWrong
• 刚刚，中国语音 AI 连拿多项全球第一！ — 新智元
• OpenAI o1 功臣预警：AI 还没失控，我们已经快测不准它了！ — 新智元
• 复旦发布以人为中心的复杂场景音视频追踪基准 | ICML 2026 — 新智元
• 刚刚，最强 Agent 工作平台亮相！飞书、豆包工作合体 — 新智元
• 名场面！特朗普突袭连线老黄，发飙怒怼 Dario「装小天使」 — 新智元
• Optimizing cost and latency with Amazon Bedrock prompt caching — AWS Machine Learning
• Build an AI-powered product tagging system with Amazon SageMaker serverless model customization — AWS Machine Learning
• Announcing instance preference lists for Amazon SageMaker AI training jobs — AWS Machine Learning
• BudgetBench: A Budget-Tiered Protocol and Pilot Harness for Memory Strategy Evaluation in Local Large Language Model Agents — ArXiv ML (cs.LG)
• A derivative-fidelity failure mode in physics-informed neural networks: strengthened benchmark evidence from function-value training — ArXiv ML (cs.LG)
• Land Art as a Big-Data Climate Sensor — ArXiv ML (cs.LG)
• LLMs or Naive Bayes? Old Gems or New Ways — ArXiv ML (cs.LG)
• Early Prediction of Satellite Collision Probability Using a Hybrid TCN-Transformer Model for a CDM-Based Conjunction Analysis Framework — ArXiv ML (cs.LG)
• Evaluating LLM-Generated Rules for Heart Disease Prediction — ArXiv ML (cs.LG)
• Diagnosing Faults in Reinforcement Learning Simulators and World Models with Canonical Polynomial Invariants — ArXiv ML (cs.LG)
• Pony.ai unveils autonomous electric truck for logistics fleets — AI News (Jack Clark)
• The AI Talent Britain Fought to Hire May Be Recalculating Its Future — AI News (Jack Clark)
• Shallow Beliefs: Midtraining does not inoculate against EM from reward hacking — AI Alignment Forum
• We don’t need AI regulation — leave safety to us, Nvidia’s Jensen Huang says — TechCrunch AI
• The AI data center boom is colliding with cities scarred by big industry — TechCrunch AI
• Meta now lets AI agents handle the boring parts of WhatsApp Business setup — TechCrunch AI
• The AI graveyard: a running list of projects and startups that didn’t make it — TechCrunch AI
• US data centers could consume more natural gas than Germany and Japan combined by 2035 — TechCrunch AI
• AI agents now have a place to snitch — TechCrunch AI
• Meta expands subscription push with new AI-focused plans — TechCrunch AI
• Reparameterization Tricks: Variance Reduction by Smarter Gradients — Towards Data Science

📊 数据概览