跳到主要内容

最新播客 知识。

Podwise 处理的新节目,随时可搜索、浏览,并在完整应用中继续。

刚刚摘要

在按下播放前,先读读每集讲了什么。

Risky Business

Risky Business #859 -- ARTEX damage approaches crisis levels

AI 摘要 The rapid proliferation of AI-driven hacking frameworks, specifically ARTEX, is fundamentally altering the threat landscape by automating complex exploitation workflows and lowering the barrier to entry for attackers. This operational shift enables indiscriminate, high-tempo…

  • ARTEX
  • LLM (Large Language Model)
  • NTLM Relay Attack

1 h 10 minEnglish摘要 · 转录

ChinaTalk

EMERGENCY POD: Jensen and His Spy

AI 摘要 The alleged relationship between NVIDIA CEO Jensen Huang and a Chinese intelligence operative highlights systemic failures in U.S. export control enforcement. Chris McGuire, a Senior Fellow at the Council on Foreign Relations and former NSC official, details how the…

  • NVIDIA
  • Jensen Huang
  • Blackwell

38 minEnglish摘要 · 转录

The Lawfare Podcast

Lawfare Archive: Keith Whittington on the Law, Politics, and Purpose of the Impeachment Power

AI 摘要 The impeachment power functions as an extraordinary constitutional tool that grants Congress the unique authority to remove members of the executive and judicial branches. Keith Whittington, a professor at Yale Law School, argues that while legal standards like "high crimes and…

  • The Impeachment Power
  • High Crimes and Misdemeanors
  • Separation of Powers

47:25English摘要 · 转录

Asia Specific

Is North Korea creating world class hackers?

AI 摘要 North Korea is increasingly utilizing sophisticated cyber warfare to generate revenue and gather intelligence, exemplified by the alleged $6.4 billion theft from the cryptocurrency exchange Binance. This incident, the largest virtual heist in history, aligns with a broader…

  • Binance
  • Lazarus Group
  • DMZ

21 minEnglish摘要 · 转录

跨国串门儿计划

#830.Yann LeCun:世界模型与下一场 AI 革命 | 南洋理工大学高等研究所公开讲座

AI 摘要 Yann LeCun 在南洋理工大学提出 “世界模型” 是实现下一代人工智能的关键。当前大语言模型受限于文本训练,缺乏物理常识与预测能力,难以匹配人类智能。他主张采用联合嵌入预测架构(JEPA),通过自监督学习在抽象表征空间进行预测,而非像素级生成。该架构支持分层规划,使机器能预判行动后果,在无需海量标注数据的情况下高效解决新任务。LeCun 强调,AI 发展应回归第一性原理,通过构建具备世界模型的自主系统,克服强化学习样本效率低下的局限,从而迈向更通用、安全的智能形态。

  • Yann LeCun
  • 图灵奖 (Turing Award)
  • 卷积神经网络 (Convolutional Neural Networks)

1 h 41 minChinese摘要 · 转录

跨国串门儿计划

#829. Dwarkesh Patel|RSI:AI怎么从数学题进步到超越顶尖研究员?

AI 摘要 AI 研发正处于自动化临界点,预计到 2028 年左右,AI 将在 AI 研究领域达到人类顶尖水平。随着 AI 在数学证明和代码编写等可验证任务中表现出超越人类的能力,其递归自我改进将大幅压缩技术迭代周期。通过构建大规模自动化强化学习环境,AI 能够以远超人类的速度进行实验、分析并优化自身架构。这种进步不仅依赖算力增长,更源于 AI 在算法创新、数据筛选及实验设计上的自主性。尽管持续学习能力和实验算力瓶颈仍存,但 AI 研发的自动化将推动技术跨越式发展,在短短一年内实现相当于过去数年的进步,最终催生出在各领域均具备专家级能力的超级智能。

  • 超级智能
  • 递归自我改进
  • 强化学习

44 minChinese摘要 · 转录

更多新节目

  1. 摘要43 min
  2. Bitcoin, A Reasonable View of The World, and Passion Is The Inverse of Information

    The Psychology of Money with Morgan Housel

    摘要00:24:58
  3. 摘要2 h 37 min
  4. 摘要00:35:35
  5. 摘要34 min
  6. 摘要21 min
  7. 摘要1 h 10 min
  8. The Case of the $4 Million Gold Coffin (Update)

    Freakonomics Radio

    摘要00:54:10
  9. 摘要26:20
  10. 摘要55 min
  11. 摘要00:33:37
  12. 摘要01:05:31
  13. 摘要56 min
  14. What if America Followed the Rules?

    The Ezra Klein Show

    摘要00:54:15
  15. Heterodox Economists: Rosa Luxemburg

    Ones and Tooze

    摘要34 min
  16. 摘要43:54
  17. 摘要56 min
  18. 摘要49 min
  19. 摘要9 min
  20. How to Make Friends as an Adult

    On Purpose with Jay Shetty

    摘要34 min
  21. 100 Objects #13: The Sand Letters

    99% Invisible

    摘要00:48:40
  22. 摘要01:30:22
  23. How Not to Read the Story of the Syrophoenician Woman

    Bishop Barron’s Sunday Sermons - Catholic Preaching and Homilies

    摘要16 min
  24. Does Google even want to win in AI?

    Decoder with Nilay Patel

    摘要39 min
  25. 摘要26 min
Podwise

每一集新节目 都有摘要。

在 Podwise 关注你的节目,在决定收听前先读每集新节目。