[What Are Reasonable AI Fears? --- 什么是合理的 AI 恐惧？](https://quillette.com/2023/04/14/what-are-reasonable-ai-fears/) 这篇文章第一条回复还挺有启发性的。 TheSnark: An interesting article. However, the author is an economist, and economics as a field of study is based on how people react to incentives. In other words, it assumes that people react in certain ways to maximize their personal benefits, whatever those might be. 一篇有趣的文章。然而，作者是一名经济学家，而经济学作为一个研究领域是基于人们对激励的反应。换句话说，它假设人们以某种 方式做出反应以最大化他们的个人利益，无论这些利益是什么。 In discussing AI's, the author not only assumes that AI's will act to benefit themselves, but that they want the same benefits that humans want. That seems far-fetched. 在讨论 AI 时，作者不仅假设 AI 会为了自己的利益而行动，而且它们也希望获得与人类相同的利益。这似乎有些牵强。 Humans are first of all interested in survival, then sex, then comfort, esteem, and so on. AI's might be interested in survival in that they don't want to be turned off; but that could just as well be something they are regurgitating from their internet scrapings Sex? I doubt it. Comfort, esteem? Again, doubtful. 人类首先对生存感兴趣，然后是性，然后是舒适、尊重等等。人工智能可能对生存感兴趣，因为它们不想被关闭；但这也可能是他们从网上垃圾中反省的东西 性？我对此表示怀疑。舒适，尊重？再次怀疑。 So since we have no idea what will motivate AI's, or if they can be said to be motivated at all, we really have no idea how they will act. We can make reasonable guesses as to how humans will react to them, which the author does. But how the AI's will act in and of themselves is pure speculation. 所以既然我们不知道什么会激励人工智能，或者如果他们可以说是有动机的，我们真的不知道他们将如何行动。我们可以合理地猜测人类将如何对它们做出反应，作者就是这样做的。但 AI 本身将如何行动纯属猜测。

via 拾月的微博