外星真理
外星真理
2022年10月
如果宇宙中其他地方存在智能生物,他们会与我们共享某些真理。数学真理将是相同的,因为它们根据定义为真。物理真理也是如此;碳原子的质量在他们的星球上将是相同的。但我认为除了数学和物理的真理之外,我们还会与外星人分享其他真理,思考这些可能是什么是值得的。
例如,我认为我们会分享这样的原则:测试某个假设的受控实验使我们有理由按比例增加对它的信念。对外星人来说,通过练习可以变得更擅长某事,这似乎也是相当可能的。我们可能会分享奥卡姆剃刀。这些想法中似乎没有任何特定于人类的东西。
当然,我们只能猜测。我们不能确定智能生命可能采取什么形式。我在这里的目标也不是探索这个问题,尽管它很有趣。外星真理概念的重点不是它给我们一种推测智能生命可能采取什么形式的方式,而是它为真理提供了一个门槛,或者更精确地说是一个目标。如果你试图找到除了数学或物理之外的最普遍的真理,那么大概它们就是我们与其他形式的智能生命共享的那些。
如果我们倾向于慷慨,外星真理将最好地作为一种启发式方法。如果一个想法可能合理地与外星人相关,那就足够了。例如,正义。我不想打赌所有智能生物都会理解正义的概念,但我也不会打赌反对它。
外星真理的概念与Erdos的上帝之书的概念有关。他过去常描述一个特别好的证明是在上帝之书中,这意味着(a)足够好的证明更多是被发现而不是被发明,(b)它的优点会被普遍认可。如果有外星真理这样的东西,那么上帝之书中除了数学还有更多内容。
我们应该称什么为寻找外星真理?显而易见的选择是”哲学”。无论哲学还包括什么,它可能都应该包括这个。我相当确定亚里士多德会这么认为。人们甚至可以说,寻找外星真理如果不是对哲学的准确描述,也是对它的一个很好的定义。也就是说,这是自称哲学家的人应该做的事情,无论他们现在是否在这样做。但我并不执着于此;做这件事才是重要的,而不是我们称它为什么。
有一天,我们可能会以人工智能的形式在我们中间拥有类似外星生命的东西。而这反过来可能让我们能够精确地确定智能生物必须与我们分享什么真理。例如,我们可能会发现,不可能创造出我们不认为智能却不使用奥卡姆剃刀的东西。有一天我们甚至可能能够证明这一点。但是,虽然这种研究会非常有趣,但这对我们的目的不是必要的,甚至不是同一领域;哲学的目标,如果我们要这么称呼的话,将是看看我们使用外星真理作为目标能想出什么想法,而不是说它的门槛确切在哪里。这两个问题有一天可能会趋同,但它们会从相当不同的方向趋同,在它们趋同之前,限制自己只思考我们确定会是外星真理的事情会过于约束。特别是因为这可能是那些最佳猜测结果证明出奇接近最优的领域之一。(让我们看看这个是否如此。)
无论我们称它为什么,发现外星真理的尝试都将是一项值得的努力。而且奇怪的是,这本身可能就是一个外星真理。
感谢Trevor Blackwell、Greg Brockman、Patrick Collison、Robert Morris和Michael Nielsen阅读本文草稿。
Alien Truth
October 2022
If there were intelligent beings elsewhere in the universe, they’d share certain truths in common with us. The truths of mathematics would be the same, because they’re true by definition. Ditto for the truths of physics; the mass of a carbon atom would be the same on their planet. But I think we’d share other truths with aliens besides the truths of math and physics, and that it would be worthwhile to think about what these might be.
For example, I think we’d share the principle that a controlled experiment testing some hypothesis entitles us to have proportionally increased belief in it. It seems fairly likely, too, that it would be true for aliens that one can get better at something by practicing. We’d probably share Occam’s razor. There doesn’t seem anything specifically human about any of these ideas.
We can only guess, of course. We can’t say for sure what forms intelligent life might take. Nor is it my goal here to explore that question, interesting though it is. The point of the idea of alien truth is not that it gives us a way to speculate about what forms intelligent life might take, but that it gives us a threshold, or more precisely a target, for truth. If you’re trying to find the most general truths short of those of math or physics, then presumably they’ll be those we’d share in common with other forms of intelligent life.
Alien truth will work best as a heuristic if we err on the side of generosity. If an idea might plausibly be relevant to aliens, that’s enough. Justice, for example. I wouldn’t want to bet that all intelligent beings would understand the concept of justice, but I wouldn’t want to bet against it either.
The idea of alien truth is related to Erdos’s idea of God’s book. He used to describe a particularly good proof as being in God’s book, the implication being (a) that a sufficiently good proof was more discovered than invented, and (b) that its goodness would be universally recognized. If there’s such a thing as alien truth, then there’s more in God’s book than math.
What should we call the search for alien truth? The obvious choice is “philosophy.” Whatever else philosophy includes, it should probably include this. I’m fairly sure Aristotle would have thought so. One could even make the case that the search for alien truth is, if not an accurate description of philosophy, a good definition for it. I.e. that it’s what people who call themselves philosophers should be doing, whether or not they currently are. But I’m not wedded to that; doing it is what matters, not what we call it.
We may one day have something like alien life among us in the form of AIs. And that may in turn allow us to be precise about what truths an intelligent being would have to share with us. We might find, for example, that it’s impossible to create something we’d consider intelligent that doesn’t use Occam’s razor. We might one day even be able to prove that. But though this sort of research would be very interesting, it’s not necessary for our purposes, or even the same field; the goal of philosophy, if we’re going to call it that, would be to see what ideas we come up with using alien truth as a target, not to say precisely where the threshold of it is. Those two questions might one day converge, but they’ll converge from quite different directions, and till they do, it would be too constraining to restrict ourselves to thinking only about things we’re certain would be alien truths. Especially since this will probably be one of those areas where the best guesses turn out to be surprisingly close to optimal.
Whatever we call it, the attempt to discover alien truths would be a worthwhile undertaking. And curiously enough, that is itself probably an alien truth.
Thanks to Trevor Blackwell, Greg Brockman, Patrick Collison, Robert Morris, and Michael Nielsen for reading drafts of this.