2021年经济学人 自动驾驶汽车（2）（在线收听）

Similar techniques are used to train self-driving cars to operate in traffic.

自动驾驶汽车在进行上路训练时也会应用类似技术。

Cars thus learn how to obey lane markings, avoid other vehicles, hit the brakes at a red light and so on.

汽车会学习如何遵守车道标志、避开其他车辆、在红灯时刹车等。

But they do not understand many things a human driver takes for granted—that other cars on the road have engines and four wheels, or that they obey traffic regulations (usually) and the laws of physics (always).

但他们理解不了许多人类司机认为理所当然的事情，比如为什么路上的汽车有引擎和四个轮子，或者为什么这些车(通常)遵守交通规则、(总是)遵循物理定律。

And they do not understand object permanence.

它们无法理解“物体恒存性”。

In a recent paper in Artificial Intelligence, Mehul Bhatt of Orebro University, in Sweden, who is also the founder of a firm called CoDesign Lab which is developing his ideas commercially, describes a different approach.

近期，瑞典厄勒布鲁大学的梅于尔·巴特在《人工智能》杂志上发表了一篇论文，描述了一种独特的方法。梅于尔·巴特也是一家名为CoDesign Lab的公司的创始人，该公司正着手将他的想法商业化。

He and his colleagues took some existing AI programs which are used by self-driving cars and bolted onto them a piece of software called a symbolic-reasoning engine.

他和他的同事采用了现有的自动驾驶汽车使用的部分人工智能程序，同时将一个叫做“符号推理引擎”的软件嵌入其中。

Instead of approaching the world probabilistically, as machine learning does, this software was programmed to apply basic physical concepts to the output of the programs that process signals from an autonomous vehicle's sensors.

该软件不像机器学习那样以概率的方式认知世界，而是将基本的物理概念应用到处理自动驾驶汽车传感器信号的程序输出中。

This modified output was then fed to the software which drives the vehicle.

然后将修改后的输出输入驱动车辆的软件当中。

The concepts involved included the ideas that discrete objects continue to exist over time, that they have spatial relationships with one another— such as "in-front-of" and "behind"—and that they can be fully or partly visible, or completely hidden by another object.

所涉及的概念有：互相独立的物体会一直存在，它们之间有空间关系，比如“在前面”、“在后面”，它们可以完全可见、部分可见、或者完全被另一个物体挡住。

And it worked. In tests, if one car momentarily blocked the sight of another, the reasoning-enhanced software could keep track of the blocked car, predict where and when it would reappear, and take steps to avoid it if necessary.

它是有用的。在测试中，如果一辆车暂时挡住了另一辆车的视线，该推理增强软件可以追踪被挡住的车，预测它将在何时何地再次出现，并在必要时采取措施避开它。

The improvement was not huge. On standard tests Dr Bhatt's system scored about 5% better than existing software.

不过进步并不大。在标准测试中，巴特博士的系统评分比现有的软件仅高出约5%。

But it proved the principle. And it also yielded something else.

但它证明了原理，还产出了其他东西。

For, unlike a machine-learning algorithm, a reasoning engine can tell you the reason why it did what it did.

因为，推理引擎和机器学习算法不同，它可以告诉你为什么这么做。