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。

GPT-3.5 passed yet another Theory of Mind test

40 点作者 izzygonzalez大约 2 年前

5 条评论

gnulinux大约 2 年前
The answers aren&#x27;t right at all... Answer to question 4 is clearly bogus as another commenter (Dfiesl) pointed out. But question 5 is also wrong. It&#x27;s not unclear, from the conversation we can deduce that Ana <i>thought</i> that Maria is pregnant, otherwise she wouldn&#x27;t have said it, unless she intentionally wants to make Maria uncomfortable, which is an unusual set of circumstances. What&#x27;s more is, <i>that</i> possibility would be inconsistent with the answer to Q4 (&quot;trying to make conversation&quot;).<p>Test failed?
评论 #34876024 未加载
RcouF1uZ4gsC大约 2 年前
Actually, ChatGPT might be useful for actually testing the theory of mind. The philosophers were always working with an N of 1 (with respect to language) when they devised these tests. It is real easy to overfit a test if you have limited samples.<p>Chat GPT is actually a good test as to which parts of the theory of mind are actually BS.
评论 #34874410 未加载
Dfiesl大约 2 年前
Seems like it got question 4 wrong... Who implies someone is pregnant to make them feel good? You imply someone is pregnant because they appear pregnant.
评论 #34878209 未加载
评论 #34874362 未加载
评论 #34874269 未加载
评论 #34874189 未加载
kelseyfrog大约 2 年前
It just predicts the next word.
评论 #34874904 未加载
评论 #34876062 未加载
评论 #34874309 未加载
评论 #34889461 未加载
评论 #34874301 未加载
mensetmanusman大约 2 年前
The answers to preëxisting theory of mind questions are stored in the graph network in a compressed sort of way, so I’m not surprised.
评论 #34875408 未加载