今天我们要聊的,不是AI模型又变大了多少,而是它们如何从内部变得更“聪明”。我们将看到,最新的论文如何教会AI从“指哪打哪”的工具,进化为能“懂你意思”的助手;又如何让强大的AI科学家学会“混圈子”,融入人类的协作生态。我们还会探讨,AI如何拥有“预算意识”,像个聪明的管家一样精打细算;以及当AI变小时,为什么最先退化的竟然是“眼力”而不是“脑力”。最后,我们还会揭秘AI“高考”中的乌龙事件,看看科学家们如何给AI的“评分尺”纠偏,这一切都指向了AI发展的新方向。00:00:42 让电脑学会“指哪打哪”之后,我们如何教它“看懂”?00:06:05 AI也能当科学家?关键要先学会“混圈子”00:11:20 聪明的AI,是如何学会“省钱”的?00:16:17 AI的“高考”,谁来检查试卷的错别字?00:21:14 AI变笨的秘密:为什么“眼力”比“脑力”更脆弱?本期介绍的几篇论文:[CV] SAM 3: Segment Anything with Concepts[Meta Superintelligence Labs]https://arxiv.org/abs/2511.16719---[AI] OmniScientist: Toward a Co-evolving Ecosystem of Human and AI Scientists[Tsinghua University]https://arxiv.org/abs/2511.16931---[LG] Budget-Aware Tool-Use Enables Effective Agent Scaling[Google Cloud AI Research & Google DeepMind & UC Santa Barbara]https://arxiv.org/abs/2511.17006---[LG] Fantastic Bugs and Where to Find Them in AI Benchmarks[Stanford University]https://arxiv.org/abs/2511.16842---[CV] Downscaling Intelligence: Exploring Perception and Reasoning Bottlenecks in Small Multimodal Models[Stanford University]https://arxiv.org/abs/2511.17487
No persons identified in this episode.
This episode hasn't been transcribed yet
Help us prioritize this episode for transcription by upvoting it.
Popular episodes get transcribed faster
Other recent transcribed episodes
Transcribed and ready to explore now
SpaceX Said to Pursue 2026 IPO
10 Dec 2025
Bloomberg Tech
Don’t Call It a Comeback
10 Dec 2025
Motley Fool Money
Japan Claims AGI, Pentagon Adopts Gemini, and MIT Designs New Medicines
10 Dec 2025
The Daily AI Show
Eric Larsen on the emergence and potential of AI in healthcare
10 Dec 2025
McKinsey on Healthcare
What it will take for AI to scale (energy, compute, talent)
10 Dec 2025
Azeem Azhar's Exponential View
Reducing Burnout and Boosting Revenue in ASCs
10 Dec 2025
Becker’s Healthcare -- Spine and Orthopedic Podcast