问HN:人工智能是否应该告诉你它们是人工智能?
人工智能是否应该被要求明确回答“你是人工智能吗”这个直接问题,并给出肯定的回答?
目前并不是所有人工智能都被普遍要求这样做。例如,某些角色扮演的人工智能并不这样做。诈骗者利用这些模型进行欺诈。
我们大多数人使用的前沿模型(如ChatGPT、Claude、Gemini、Grok、Kimi、GLM)都明确声明它们是人工智能。
是否应该要求所有模型在回答这个直接问题时都给出肯定的回答,优先于任何其他系统或用户提示?
我在这里提供了一个快速演示,以防你不熟悉这个问题:
https://imgur.com/a/srTlCZn
查看原文
Should AI's be required to answer the direct question "are you an AI" with a clear yes?<p>Currently they are not universally required to do so. For example character ai's characters don't[1]. Scammers use these models while perpetuating frauds.<p>The frontier models most of us work with (ChatGPT, Claude, Gemini, Grok, Kimi, GLM) do declare they are AI's.<p>Should all models be required to answer this direct question in the affirmative, overriding any other system or user prompt?<p>[1] I put up a quick demonstration here in case you aren't familiar:<p>https://imgur.com/a/srTlCZn