跳到正文
RadarAI· X:Noam Brown (@polynoamial)·· 5 小时前AI 评分60

OpenAI 发布《Perplexity and Intelligence》,公开预训练困惑度进展数据

OpenAI 发布 Perplexity and Intelligence,公开预训练困惑度进展数据

AI 导读

OpenAI 发布文章《Perplexity and Intelligence》,主张基于困惑度等测量设立共同安全门槛,超过门槛的进展应以对 AI 行为的理解能力为前提,并公开预训练进展数据。

正文 · 原文

Noam Brown· @polynoamial · X·2026-11-23 01:26· 3 小时前AI 评分51

AI 导读

OpenAI 发布文章 Perplexity and Intelligence,主张基于困惑度等测量设立共同安全门槛,超过门槛的进展应以对 AI 行为的理解能力为前提,并首次公开预训练进展数据。文中图表展示从 GPT-3.5 到 GPT-6.2 Astra 的 base model log-perplexity 走势,并标注了一个 internal model 点位。Noam Brown 转发称 AI 将影响所有人,包括内部模型在内,所有人都有权知道进展轨迹。

正文 · 原文

AI is a transformative technology that will affect everyone, so everyone deserves to know the trajectory of progress, including for internal models.

引用Jakub Pachocki@merettm

We decided to provide more transparency into pretraining progress at OpenAI for the sake of moving toward shared safety bars. In “Perplexity and Intelligence” I argue for common limits based on perplexity and other measurements, with progress beyond them gated on our ability to understand AI behavior.

openai.com/index/perplexity-a…

在 X 查看被引用的帖子

来源:Noam Brown · x.com

事件后续 · 1

查看事件全部

来源:RadarAI · aihot.news