TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Debate May Help AI Models Converge on Truth

15 点作者 sonabinu6 个月前

2 条评论

gildandstain6 个月前
Sounds like they applied insights from Mercier and Sperber's "The Enigma of Reason", namely that perceiving truth is hard, but sorting arguments by quality is much easier. Or -adversarial reason-giving as the engine of collective knowledge
emptiestplace6 个月前
Just wait until they gain meaningful context persistence and the consequent feedback loops.