不管《财富》和穆斯克争论有多激烈，自动驾驶汽车还应继续前行

上周传出这样一条新闻：特斯拉Model S车主自动驾驶时遭遇车祸去世，美国监管机构正调查事故原因。这是自动驾驶技术首次出现车祸致死，不仅要惋惜生命的逝去，也要正视自动驾驶汽车发展中面临的一系列挑战、道德难题和难以回答的疑问。过去仅在理论上的讨论突然之间变成现实。

然而，没人觉得这次事故会放缓汽车的自动化进程。特斯拉公司的股票略有下挫，但分析师称之不过是“头条新闻风险”。

News emerged this week that U.S. regulators were investigating the death of a driver using the Autopilot feature of a Tesla Model S. This was the first death of its kind, and while it’s first and foremost a tragic loss of life, it also points to an array of challenges, ethical conundrums, and unanswered questions about the quest for self-driving cars. What had been theoretical debates are suddenly starkly real.

By and large, there seems little expectation that the event in and of itself will slow progress towards vehicle automation. Tesla’s own stock suffered modest losses on the news, and analysts described the event as a mere “headline risk.”

原因之一是，虽然调查还在进行，但看起来特斯拉的自动驾驶系统Autopilot并不是事故的元凶。而且有猜测认为，遇难的特斯拉车驾驶者生前有可能注意力分散，也许还有超速情况。更重要的是，事故发生时向特斯拉迎面驶来的拖挂卡车司机采取了危险转弯。Autopilot显然并非完美，但事后看来，原因就是两位驾驶员陷入了危险境地，自动驾驶系统也无能为力，驾驶系统并未犯致命错误。

从广义的角度看，这起事故是不可避免的：谁也没法保证自动驾驶系统能避免车祸，这样一起事故发生时，特斯拉作为业内技术最先进的公司，多少也承担了成为关注焦点的风险。特斯拉已经回应，指出这是Autopilot系统行驶1.3亿英里首次撞车，而美国司机平均每1亿英里行程就会出一起致命车祸。特斯拉的样本规模还不够大，跟人类驾驶对比证据稍显不足，但至少目前看来Autopilot有可能让汽车更安全。

尽管如此，这起致命事故还是导致了一些风险。首先，自动驾驶面临监管可能更严格，目前监管施加的压力并不大。严格的监管可能拖累技术研发的进度。这起事故的法律影响还不确定——如果有人说服法官或者陪审团相信特斯拉应为撞车负责，Autopilot和自动驾驶可能会遭受重挫。

关于监管和追究事故责任的追问不断可能让特斯拉焦头烂额。虽然特斯拉反复强调，Autopilot是一款“测试版”产品，即使自动驾驶状态开启，系统内设的语音也会提示驾驶者把双手放在方向盘上。在科技行业，产品尚未完善就推出很正常。可这起事故提醒我们，汽车行业人命关天，产品不能有瑕疵。尽管“测试”项目对帮助特斯拉改进Autopilot至关重要——但法官和议员可能得仔细衡量，让驾驶员冒着生命危险使用监管松弛又不完善的产品值不值得。

That’s in part because, while an investigation is still underway, it so far does not seem that the Tesla Autopilot feature was the root cause of the accident. There is speculation that the Tesla driver may have been distracted, and perhaps speeding. More important still, most accounts of the incident have the semi-truck’s driver making a very dangerous turn across oncoming traffic. Autopilot clearly isn’t perfect, but the emerging picture is one in which two human drivers created a situation that an automated system failed to save them from, rather than one in which an automated system made a fatal mistake on its own.

More broadly, this incident has an air of inevitability: No one claims that automated systems will prevent all crashes, and as the company with the most advanced commercially available automation tech, Tesla more or less knowingly shouldered the risk of being in the spotlight when a crash like this occurred. Tesla has responded to the event in part by pointing out that this is the first crash after 130 million miles of Autopilot use, while U.S. drivers overall average about one death per 100 millionvehicle miles traveled. Though Tesla’s sample size is not big enough to make the case on that comparison alone, it’s at least an early indicator that Autopilot does make the cars safer.

Nonetheless, the incident generates some risks. For one, it could lead to political pressure to tighten regulation of automation features, which is currently relatively limited. Tighter regulation could slow development of the technology. The legal fallout from the incident is also still uncertain—if someone can convince a judge or jury that Tesla is liable for the crash, the picture for Autopilot and automation could shift rapidly.

Questions of both regulation and liability could hinge on Tesla’s repeated insistence that Autopilot is a ‘beta’ product, and its many built-in warnings that drivers should keep their hands on the wheel even when it is active. While releasing a product that’s less than perfect is common practice in the tech world, where Elon Musk’s roots lie, this crash reminds us that things are different when it comes to cars. The ‘beta’ program has been crucial to helping Tesla improve Autopilot – but judges and lawmakers may ultimately have to decide whether that’s worth the tradeoff of risking driver lives on a lightly regulated and explicitly imperfect product.

由此还引申了另一个问题，即“半自动化”是否导致一种新型风险。正如美国自动驾驶研究公司Kelley Blue Book一位分析师向《底特律新闻》指出，“滥用辅助驾驶技术的记录”在Youtube等网站上比比皆是——一些视频里驾驶员双手离开方向盘，甚至边看报边开车。或许真应该问问：特斯拉是否应该采取更激进的方式管控滥用自动驾驶系统问题，也许推广或者描述功能时可以更保守些。有可能很快就能看到变化。

在这场争论中，最极端的观点来自美国知名科技博客Gizmodo的编辑阿丽莎·沃尔克。她认为，这次致命撞车证明，“全自动汽车才最适合现实路况的。”这种观点有问题，因为自动刹车等各类半自动化系统早已投入应用，挽救了不少生命。虽然时而有些进展公开，但全自动汽车会否很快面市很难说。那一天真正到来以前，如果不允许汽车应用线道侦测和其他安全技术，可能会阻碍全自动汽车必需的诸多功能开发进程。

至少近期内，特斯拉这起致命车祸的教训是：汽车是强大却危险的机器。或许有一天，全自动化会让车辆真正安全，每年全球造成逾百万起死亡的交通事故将大为减少，甚至不复存在。

可惜的是，那一天还没到来。  （财富中文网）

译者：Pessy

审校：夏林

Related to this is the question of whether ‘partial automation’ creates a unique sort of risk. As a Kelley Blue Book analyst put it to the Detroit News, “documented abuses of driver-assist technology” have been plastered all over sites like YouTube —videos of drivers operating their Tesla with no hands, or even while reading the newspaper. It’s fair to ask whether Tesla should have been more aggressive about policing these misuses of the system, perhaps by marketing or characterizing the technology itself more conservatively. Those changes could be coming soon.

At the most extreme end of that debate, Gizmodo’s Alissa Walker argues that the crash proves that “fully autonomous vehicles are the only types of self-driving cars that make sense in our streets.” That’s a problematic argument, because various kinds of partial automation, such as automatic braking, arealready on the road and saving lives. Despite some bold public statements, there’s also little certainty that full vehicle automation is coming anytime soon, and keeping lane detection and other safety features out of cars until it’s here would possibly hinder the development of the myriad features necessary to add up to a fully autonomous car.

At least in the near term, what it all boils down to is this: Automobiles are powerful, dangerous machines. Maybe full automation will someday make them truly safe, preventing most, or even all, of the million-plus traffic deaths that occur worldwide each year.

But we’re not there yet.