昨天午饭时,一位朋友问我:有什么技术趋势是他应该关注、但很可能正在忽略的。 Yesterday at lunch a friend asked me what tech trend he should pay attention to but was probably ignoring. ¶
我不假思索地回答「人工智能」,事后又多想了一下,我认为这个答案大概是对的。 Without thinking much I said “artificial intelligence”, but having thought about that a bit more, I think it’s probably right. ¶
需要说明的是,AI(按照通行的科学定义)很可能行不通。这句话对任何新技术都可以这么说,而且大体上都是正确的。但我认为大多数人对 AI 的前景过于悲观了——AI 失败的时间太久,已经背上了坏名声。计算机科学教授提起它时总带着一丝讥笑。这套逻辑是:神经网络第一次就没成,所以这次也不会成。 To be clear, AI (under the common scientific definition) likely won’t work. You can say that about any new technology, and it’s a generally correct statement. But I think most people are far too pessimistic about its chances - AI has not worked for so long that it’s acquired a bad reputation. CS professors mention it with a smirk. Neural networks failed the first time around, the logic goes, and so they won’t work this time either. ¶
但通用人工智能(AGI)或许能成,而如果它成了,那将是有史以来最重大的技术进展。 But artificial general intelligence might work, and if it does, it will be the biggest development in technology ever. ¶
我认为我们在许多具体领域已经走得更近了——例如,计算机如今在许多令人印象深刻的事情上比人类做得更好,比如下国际象棋和驾驶飞机。但我们并不把这些称为 AI 的例子,而只是说它们本来就没那么难。平心而论,这些确实没有一个让人感觉像是一台能像人一样思考的计算机。 I’d argue we’ve gotten closer in lots of specific domains - for example, computers are now better than humans at lots of impressive things like playing chess and flying airplanes. But rather than call these examples of AIs, we just say that they weren’t really that hard in the first place. And to be fair, none of these really feel anything like a computer that can think like a human. ¶
有一些私营(或新近被收购的)公司,加上几家大型上市公司,正在通向通用人工智能的道路上取得令人瞩目的进展,但其中做得好的都对此守口如瓶。 There are a number of private (or recently acquired) companies, plus some large public ones, that are making impressive progress towards artificial general intelligence, but the good ones are very secretive about it. ¶
当然也有一些乐观的理由。吴恩达(Andrew Ng)——曾参与或正在参与 Google 的 AI 工作——说过,他相信学习来自单一算法:你大脑中处理耳朵输入的那个部分,也有能力学会处理眼睛的输入。如果我们能弄清这一种通用算法,程序或许就能学会通用的事情。 There are certainly some reasons to be optimistic. Andrew Ng, who worked or works on Google’s AI, has said that he believes learning comes from a single algorithm - the part of your brain that processes input from your ears is also capable of learning to process input from your eyes. If we can just figure out this one general-purpose algorithm, programs may be able to learn general-purpose things. ¶
这类工作已经发表了一些有希望的早期成果,但因为大脑是一个如此复杂、如此依赖涌现行为的系统,很难说我们离目标到底有多近。我们已经相当了解单个神经元如何工作,而这有可能就是我们建模智能工作原理所需的全部知识。但 1000 亿个神经元按同样原理协同工作时涌现出的行为会变得极其复杂,难以用软件建模。或者,用 Nick Sivo 的话说,「这就像只凭晶体管工作原理的基础知识去逆向工程最新的英特尔处理器。」也有可能智能背后另有某种现象,而研究这个方向的人走错了路。 There have been promising early results published from this sort of work, but because the brain is such a complex system so dependent on emergent behavior it’s difficult to say how close to the goal we really are. We understand how individual neurons work pretty well, and it’s possible that’s all we need to know to model how intelligence works. But the emergent behavior of 100 billion of them working together on the same principles gets extraordinarily complex, and difficult to model in software. Or, as Nick Sivo says, "it's like reverse engineering the latest Intel processor with only the basic knowledge of how a transistor works." It’s also possible that there’s some other phenomenon responsible for intelligence, and the people working on this are on the wrong track. ¶
对我来说,最大的问题不在于人工智能,而在于人工意识,或者说创造力、欲望——随你怎么称呼它。我相当确信,我们能做出把特定的复杂任务完成得非常好的计算机程序。但我们如何做出一个能自己决定想做什么的计算机程序?我们如何让一台计算机自己决定去关心学会开车这件事?或者去写一部小说? The biggest question for me is not about artificial intelligence, but instead about artificial consciousness, or creativity, or desire, or whatever you want to call it. I am quite confident that we’ll be able to make computer programs that perform specific complex tasks very well. But how do we make a computer program that decides what it wants to do? How do we make a computer decide to care on its own about learning to drive a car? Or write a novel? ¶
有可能——甚至很可能——这种创造力将以某种非直觉的方式成为学习的涌现属性。在进化的进程中发生了某件事,让人脑不同于爬行动物的脑——后者更接近一台玩乒乓游戏的计算机。(我本来想说一台下国际象棋的计算机,但计算机下棋没有直觉或本能——它们只是非常快地搜索一个巨大的解空间。) It’s possible--probable, even--that this sort of creativity will be an emergent property of learning in some non-intuitive way. Something happened in the course of evolution to make the human brain different from the reptile brain, which is closer to a computer that plays pong. (I originally was going to say a computer that plays chess, but computers play chess with no intuition or instinct--they just search a gigantic solution space very quickly.) ¶
而且,也许我们并不想造出在这种意义上有意识的机器。我能想到的最正面的结局是:计算机变得非常擅长「做」,人类变得非常擅长「想」。如果我们始终弄不清如何让计算机拥有创造力,那么人与机器之间将会有一种非常自然的分工。 And maybe we don't want to build machines that are concious in this sense. The most positive outcome I can think of is one where computers get really good at doing, and humans get really good at thinking. If we never figure out how to make computers creative, then there will be a very natural division of labor between man and machine. ¶