<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>前沿AI on gulucaptain's Blog</title><link>https://gulucaptain.github.io/DailyPaper/tags/%E5%89%8D%E6%B2%BFai/</link><description>Recent content in 前沿AI on gulucaptain's Blog</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Mon, 12 Oct 2026 08:55:00 +0800</lastBuildDate><atom:link href="https://gulucaptain.github.io/DailyPaper/tags/%E5%89%8D%E6%B2%BFai/index.xml" rel="self" type="application/rss+xml"/><item><title>刚刚，Karpathy称全球只有5000人，真正接触过最前沿AI</title><link>https://gulucaptain.github.io/DailyPaper/clips/karpathy-capability-funnel/</link><pubDate>Mon, 12 Oct 2026 08:55:00 +0800</pubDate><guid>https://gulucaptain.github.io/DailyPaper/clips/karpathy-capability-funnel/</guid><description>Karpathy 用四层漏斗（60 亿人没碰过 LLM / 10–20 亿只用免费版 / 2000 万专业用户 / 5000 人有实验室内部访问权）说明：同一个 AI 在不同人眼里的样子完全不同，而决定差距的三个因子是问题规模、能力锯齿状（集中在数字化可验证且赚钱的领域）和访问权限。</description></item></channel></rss>