经济学人132:当代哲学 "直觉泵"入门（打印版）

　　“THINKING is hard,” concedes Daniel Dennett. “Thinking about some problems is so hard that it can make your head ache just thinking about thinking about them.” Mr Dennett should know. A professor of philosophy at Tufts University, he has spent half a century pondering some of the knottiest problems around: the nature of meaning, the substance of minds and whether free will is possible. His latest book, “Intuition Pumps and Other Tools for Thinking”, is a precis of those 50 years, distilled into 77 readable and mostly bite-sized chapters.

　　丹尼尔·丹尼特承认：“思考是困难的。思考某些问题之难，几乎到了只要一想到要思考这些问题就会头疼的地步。”对此，丹尼特深有体会。作为塔夫斯大学的一名哲学教授，在过去半个世纪里，他一直在试图解开我们身边那些最难以解决的问题：语言含义的本质，人类心灵是由什么组成的以及自由意志是否可能等。日前，他把自己50年来的思考浓缩为77篇既有可读性又短小精悍的文章，汇集成册后取名为《思考的工具：直觉泵及其他》

　　“Intuition pumps” are what Mr Dennett calls thought experiments that aim to get at the nub of concepts. He has devised plenty himself over the years, and shares some of them. But the aim of this book is not merely to show how the pumps work, but to deploy them to help readers think through some of the most profound conundrums.

　　“直觉泵”是丹尼特用来称呼思想实验的一个词语。所谓思想实验就是一种意在探究概念核心内容的实验。在书中，他同读者一起分享了这些年来为自己所设计的许多此类实验。但是，他写作本书的目的不仅仅是要展示“直觉泵”的工作流程，更重要的是通过使用“直觉泵”，帮助读者去思考那些最深奥的难题。

　　As an example, take the human mind. The time-honoured idea that the mind is essentially a little man, or homunculus, who sits in the brain doing clever things soon becomes problematic: who does all the clever things in the little man's brain? But Mr Dennett offers a way out of this infinite regress. Instead of a little man, what if the brain was a hierarchical system?

　　下面，就让我们以“人类的心灵”为例来介绍一下丹尼特的“直觉泵”理论。传统的观点认为，从本质上说，心灵是一个住在人类大脑里从事智力活动小人，或者侏儒。不过，在丹尼特看来这种观点是站不住脚的：既然如此，那么，住在这个侏儒的大脑中，从事智力活动的人又是谁呢？对此，丹内特提供了一种走出此类无限回归的方法。在他看来，如果大脑不是一个侏儒，而是一个分级系统，情况会如何呢？

　　This pump, which Mr Dennett calls a “cascade of homunculi”, was inspired by the field of artificial intelligence. An AI programmer begins by taking a problem a computer is meant to solve and breaking it down into smaller tasks, to be dealt with by particular subsystems. These, in turn, are composed of sub-subsystems, and so on. Crucially, at each level down in the cascade the virtual homunculi become a bit less clever, to a point where all it needs to do is, say, pick the larger of two numbers. Such homuncular functionalism replaces the infinite regress with a finite one that terminates at tasks so dull and simple that they can be done by machines.

　　丹内特把这种泵称之为“级联侏儒”，其灵感自于人工智能领域。在人工智能领域，当一名程序员接受了一项准备交给计算机去解决的任务时，他首先会把这项任务分解为一些更小的任务，然后再把这些更小的任务交给一些特定的子系统去解决，而这些子系统又是由更小的子子系统组成的，以此类推。这一过程的关键在于，在这个串联在一起的每一个较低层次的子系统中，那个虚拟侏儒会渐渐地变得不再那么聪明，直至到达这样一个点为止。在这个点上，需要那个侏儒来做的全部事情就是在两个数字中选出较大的那一个。这样，这种侏儒技能主义就用有限回归代替了原来的无限回归，而这种有限回归能够在任务变得非常单调又非常简单以至于只需机器就可以完成的时候自动停止。

　　Of course the AI system is designed from the top down, by an intelligent designer, to perform a specific task. But there is no reason why something similar couldn't be built up from the bottom. Start with nerve cells. They are certainly not conscious, at least in any interesting sense, and so invulnerable to further regress. Yet like the mindless single-cell organisms from which they have evolved, each is able to secure the energy and raw materials it needs to survive in the competitive environment of the brain. The nerve cells that thrive do so because they “network more effectively, contribute to more influential trends at the [higher] levels where large-scale human purposes and urges are discernible”.

　　人工智能系统当然是由一位有智慧的设计师，为了完成某项特殊的任务，按照由上至下的原则设计出来的。既然如此，那么就没有理由不能按照从下至上的原则建造一个同样的系统。以神经细胞为例，这些细胞当然是没有意识的，他们几乎没有任何有趣的感觉，它们非常脆弱，再也经不起进一步的回归。然而，就像它们从中进化而来的、没有思维能力的单细胞器官一样，这些神经细胞中的每一个都具备获取所需的能量和原料的能力，这种能力确保它们能在互相竞争的大脑环境中生存下去。那些茁壮成长的神经细胞之所以这样做，是因为他们“串联在一起时的工作效率更高，对[更高]层次那些更有影响力的趋势做出了贡献，而这些更高的层次可以辨识大量的人类意识。”

　　From this viewpoint, then, the human mind is not entirely unlike Deep Blue, the IBM computer that famously won a game of chess against Garry Kasparov, the world champion. The precise architecture of Mr Kasparov's brain certainly differs from Deep Blue's. But it is still “a massively parallel search engine that has built up, over time, an outstanding array of heuristic pruning techniques that keep it from wasting time on unlikely branches”.

　　由此来看，心灵并不是完全不同于“深蓝”的东西。“深蓝”是由IBM公司设计建造的一台计算机，它因在一场比赛中战胜了国际象棋世界冠军加里·卡斯帕罗夫而名噪一时。卡斯帕罗夫大脑的精确架构肯定有别于“深蓝”。不过，即便如此，“深蓝”的大脑仍旧是“人类到目前为止按照启发式剪枝技术阵列而建造的一台能够进行大规模并行运算的搜索引擎，启发式剪枝技术阵列是一种先进的技术，它能够让搜索引擎不把时间浪费在不可能的选择之上。“

　　Those who insist Deep Blue and Mr Kasparov's mind must surely be substantially different will balk at this. They may well be right. But the burden of proof, Mr Dennett argues, is on them, for they are in effect claiming that the human mind is made up of “wonder tissue” with miraculous, mind-moulding properties that are, even in principle, beyond the reach of science—an old-fashioned homunculus in all but name.

　　读到此处，那些坚持“深蓝”与卡斯帕罗夫的心灵一定是由不同物质构成的人定会咆哮不已。他们也许是对的。但是，丹内特指出，如果如此的话，那么证明这一理论的重担就落在了这些人的身上。这是因为，他们的观点实际上是在宣称，人类心灵是由具有不可思议的、能够影响心灵的性能的“非同一般的组织”所组成的。但是，从理论上来说，这些组织已经超出了科学的认知范畴—实际上，它们不过是一种已经过时的侏儒而已。

　　Mr Dennett's book is not a definitive solution to such mind-benders; it is philosophy in action. Like all good philosophy, it works by getting the reader to examine deeply held but unspoken beliefs about some of our most fundamental concerns, like personal autonomy. It is not an easy read: expect to pore over some passages more than once. But given the intellectual gratification Mr Dennett's clear, witty and mercifully jargon-free prose affords, that is a feature, not a bug.

　　丹内特的书并不是对此类让头脑发晕的问题的终极答案。相反，他的书是对此类问题的终极思考。这就如同那些有意义的终极思考一样，该书通过让读者对那些深藏于我们内心却又无法言说的信念进行检讨，从而达到自己的目的。这不是一本容易读懂的书。因此，我们希望读者在阅读某些篇章时多读几遍。不过，考虑到丹内特那种明晰、诙谐以及充分为读者考虑而不使用术语的文风所带来的智力上的满足，多读几遍才能读懂正是本书的一大特色，而不是一个瑕疵。

经济学人132:当代哲学 "直觉泵"入门（在线收听）