[人人能懂] AI的五项修炼:挂挡、瘦身、听风、量尺、育苗 episode artwork

EPISODE · Jan 30, 2026 · 32 MIN

[人人能懂] AI的五项修炼:挂挡、瘦身、听风、量尺、育苗

from AI可可AI生活

我们总在惊叹AI变得多聪明,但你有没有想过,我们该如何从根基上,打造一个学得更快、身形更巧、感知更敏锐、评价更科学,甚至能自我进化的AI呢?本期节目,我们将通过五篇最新的AI论文,一次性揭开这些秘密。我们会聊聊AI学习速度原来只有四个“档位”;探讨如何给大模型“减肥”却不牺牲效果;见证AI如何拥有“听声辨位”的超能力;学习如何给眼花缭乱的AI科学地“排座次”;最后,我们还会看到一个“博士生”AI是如何手把手教出一个更聪明的“小学生”AI的。准备好了吗?让我们即刻出发,探索AI的底层构造蓝图。00:00:45 人工智能学习的速度,原来只有四档00:07:40 AI减肥记,如何不花钱还把活干好?00:13:35 AI的“听声辨位”,我们离《三体》里的智子还有多远?00:19:43 给AI大模型排座次,你信的榜单可能用错了尺子00:26:35 让AI自己教自己,我们如何从根上培养一个更聪明的模型?本期介绍的几篇论文:[LG] A Theory of Universal Agnostic Learning[Purdue University & Technion and Google Research]https://arxiv.org/abs/2601.20961---[CL] ECO: Quantized Training without Full-Precision Master Weights[Google Research & ISTA]https://arxiv.org/abs/2601.22101---[AS] PhaseCoder: Microphone Geometry-Agnostic Spatial Audio Understanding for Multimodal LLMs[Google DeepMind & Google AR]https://arxiv.org/abs/2601.21124---[LG] Nonparametric LLM Evaluation from Preference Data[LMU Munich & CMU & University of Cambridge]https://arxiv.org/abs/2601.21816---[CL] Self-Improving Pretraining: using post-trained models to pretrain better models[FAIR at Meta]https://arxiv.org/abs/2601.21343在小宇宙查看该单集文稿

Episode metadata supplied by the publisher feed · Published Jan 30, 2026

Embed this episode

NOW PLAYING

[人人能懂] AI的五项修炼:挂挡、瘦身、听风、量尺、育苗

0:00 32:58

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

Frequently Asked Questions

How long is this episode of AI可可AI生活?

This episode is 32 minutes long.

When was this AI可可AI生活 episode published?

This episode was published on January 30, 2026.

Can I download this AI可可AI生活 episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.
URL copied to clipboard!