返回全部资讯
GREG ISENBERG (@gregisenberg)T2·基座大模型

Greg Isenberg 调侃早期 LLM 局限:静态数据、无记忆与多模态缺失将成未来笑谈

原标题:we’ll tell our kids that the first AI we used was basically a giant scrape of the internet frozen in...
35AI 研判分

科技评论人 Greg Isenberg 指出,2021 年冻结的互联网数据、缺乏长期记忆、无法处理多模态输入以及仅能文本回复等限制,构成了早期大语言模型的核心特征。他认为这些如今被视为基础缺陷的功能缺失,将在未来被后代视为“石器时代”的技术瓶颈。该观点强调了 AI 从纯文本静态模型向具备实时感知、记忆和多模态能力的智能体演进的历史断层感。

报道正文

GREG ISENBERG (@gregisenberg) · 2026-10-13 13

we’ll tell our kids that the first AI we used was basically a giant scrape of the internet frozen in 2021, couldn’t go online to check anything, forgot who we were every time we closed the tab, couldn’t see a photo or hear our voice, couldn’t read more than a few pages at once, couldn’t do a single thing for us besides type back, and was “at capacity” half the time, and they’ll think we were living in the stone age.