Could more brain-like chips provide a path to consciousness?
更像大脑的芯片，能走向意识吗？
从存储与计算分离的普通电脑，到忆阻器、活神经元和类器官，文章依次介绍几条接近大脑的路径。关键是读清：工程机制、已有演示和关于意识的可能性判断，各自说明了什么。
原文来源：The Economist
学习目标
- 分清三类近脑计算方案
- 读懂压缩修饰与多层强调结构
- 准确还原比较对象和代词
- 保留科学报道中的条件与不确定性
中英对照阅读
THE CURRENT crop of AI models looks to some scientists like a dead end when it comes to consciousness, regardless of how powerful they become. For some, this is because of their lack of wet, biological stuff; for others, because of their design structure. But they do not rule out the potential for consciousness in future versions of AI built either with living cells, or on different, more brain-like, computational architectures.
在一些科学家看来，谈到意识，目前这一批人工智能模型似乎走进了死胡同，不论它们变得多么强大。 对一些人而言，这是因为它们缺少湿润的生物物质；对另一些人而言，则是因为它们的设计结构。 但他们并不排除未来版本的人工智能拥有意识的可能性，这些版本或由活细胞构建，或采用不同的、更接近大脑的计算架构。
Modern computers are based on a blueprint invented by John von Neumann, a Hungarian polymath, in 1945. They feature a processing unit (for making calculations) and a memory unit (for storing instructions and data). Information has to zoom back and forth between these units and, for AI applications, the energy requirements can mount up.
现代计算机基于匈牙利博学家约翰·冯·诺依曼于1945年提出的一个设计蓝图。 它们有一个处理单元，用来进行计算，还有一个存储单元，用来存放指令和数据。 信息必须在这两个单元之间快速来回传递，而在人工智能应用中，能源需求可能累积起来。
Brains do not work this way. Neurons process and store data in the same place, making networks of them much more energy-efficient at computations than their digital-computer equivalents. A typical brain consumes around 20 watts whereas an artificial neural network carrying out an equivalent number of parallel calculations would require a data centre that guzzles millions of watts.
大脑并非这样工作。 神经元在同一位置处理和存储数据，使它们构成的网络在计算时比相应的数字计算机系统节能得多。 一个典型的大脑消耗约20瓦功率，而执行相同数量并行计算的人工神经网络，则需要一个耗电数百万瓦的数据中心。
“Neuromorphic” computing aims to be more efficient and brain-like, using devices called memristors to perform computations and store information in the same place. Apply a voltage to a memristor and its conducting properties change; remove that voltage and that change persists, acting as a memory. Such traits make computing far more efficient. But they are, separately, intriguing to some philosophers who believe that it is not computational prowess but the way in which computers “think” that will make it possible for them to become conscious.
“神经形态”计算旨在更加高效，也更加接近大脑：它使用称为忆阻器的器件，在同一位置执行计算并存储信息。 给忆阻器施加电压，它的导电特性就会改变；撤去该电压，这一变化仍会保留，发挥记忆的作用。 这些特征使计算高效得多。 但另一个方面是，它们也让一些哲学家感到兴趣；这些人相信，使计算机有可能获得意识的，将不是计算本领，而是计算机“思考”的方式。
Another idea is to use real brain cells. Cortical Labs, an Australian startup, has developed a computer using hundreds of thousands of human neurons (grown from donated stem cells) mounted on silicon. Their “biological computer” has been taught to play “Pong”, a simple computer game, and, in later versions, “Doom”, a first-person shooter game.
另一个想法是使用真正的脑细胞。 澳大利亚初创企业Cortical Labs已经开发出一种计算机，它使用数十万个安装在硅上的人类神经元，这些神经元由捐赠的干细胞培养而来。 他们的“生物计算机”已经学会玩简单的电脑游戏《Pong》，后来的版本还学会了玩第一人称射击游戏《Doom》。
Further in the future are neural organoids, three-dimensional clusters of brain cells grown in labs. Though still a relatively new technology, used in labs to understand better how brains grow and behave, they could also one day be used to perform computations. The largest organoids grown in labs today contain more brain cells than there are in a honey bee. They are big enough to potentially develop consciousness, suggests Peter Godfrey-Smith, a philosopher at the University of Sydney who works on animal minds, but they are just not organised in the right way.
更遥远的未来方向是神经类器官，即在实验室培养的三维脑细胞团。 虽然这仍是一项相对较新的技术，目前用于实验室中更好地了解大脑如何生长和活动，但有朝一日它们也可能被用于计算。 如今在实验室培养出的最大类器官，所含脑细胞比一只蜜蜂拥有的还多。 悉尼大学研究动物心智的哲学家彼得·戈弗雷－史密斯认为，它们已大到有可能发展出意识，但只是尚未以恰当方式组织起来。
The common factor in all these ideas is that they are closer analogues with the human brain (in some cases because they are made of the same stuff) than the digital computers of today.
所有这些想法的共同点，是它们比今天的数字计算机更接近人脑的对应物；在某些情况下，这是因为它们由与人脑相同的物质构成。
“How brain-like does AI have to be to move the needle on the credence that it might be conscious?” says Anil Seth, a neuroscientist at the University of Sussex. He believes that silicon, no matter how cleverly it is constructed, probably will not achieve consciousness on its own. “If you start to build systems that share more and more properties with brains, given that we don’t know which of those properties matters, then you’re more likely to move the needle a bit.”
萨塞克斯大学神经科学家阿尼尔·塞思问：“人工智能要像大脑到什么程度，才能让我们对它可能具有意识的相信程度有所提升？” 他认为，硅无论被多么巧妙地构造，很可能都无法仅凭自身实现意识。 “如果你开始构建与大脑共有越来越多特征的系统，考虑到我们并不知道这些特征中哪些起作用，你就更有可能让这种判断向前移动一点。”