上周五，谷歌的一輛用于測試無人駕駛技術(shù)的雷克薩斯測試車在加州的山景城遭到了一輛闖紅燈的汽車撞擊。有目擊者稱，這起事故是谷歌無人駕駛汽車遭遇的幾起事故中最嚴(yán)重的一起，好在據(jù)稱現(xiàn)場并無人員傷亡。

從攝于事故現(xiàn)場的一張照片中可以看出（這張照片可以在9TO5Google網(wǎng)站上查看），撞擊谷歌路試車的這輛州際電池公司（Interstate Batteries）的面包車明顯應(yīng)負(fù)全責(zé)。據(jù)谷歌公司的一份聲明稱：“在我們的測試車駛?cè)胧致房谥?，綠燈至少已經(jīng)亮了6秒鐘?！?

目擊者表示，在事故發(fā)生后，“幾名暈頭轉(zhuǎn)向的谷歌員工”坐在路邊等待拖車。

此次事故與特斯拉公司的無人駕駛技術(shù)的幾起事故形成了鮮明對比。畢竟特斯拉的事故還造成了一起人員傷亡，這說明特斯拉的監(jiān)測系統(tǒng)在態(tài)勢感知和響應(yīng)方面還存在一些不足。這次谷歌的路試車雖然也被撞得挺慘，但我們也不能責(zé)怪它，畢竟事故的責(zé)任方是那輛闖了紅燈的面包車。

不過此次事故也指出了圍繞著無人駕駛汽車的另一個(gè)問題——未來幾十年里，無人駕駛汽車可能還得與不靠譜的人類司機(jī)們共享路權(quán)。高盛公司的最近一份報(bào)告就指出，以當(dāng)前的車輛置換率和擁車模式來看，北美地區(qū)要想全面落地?zé)o人駕駛汽車，至少得等到2060年。

在這種情況下，又會有一種新的問題出現(xiàn)——一旦發(fā)生了交通事故，擁有預(yù)判和避免人類司機(jī)的駕駛錯(cuò)誤能力的無人駕駛系統(tǒng)，又該在事故中承擔(dān)多少責(zé)任呢？ （財(cái)富中文網(wǎng)）

譯者：樸成奎

On Friday, a Lexus outfitted with Google autonomous driving technology was struck by a vehicle that ran a red light in Mountain View, California. Some observers say it is the worst crash that Google’s autonomous vehicles have been involved in. There were no reported injuries.

A photo of the aftermath (which you can see at 9TO5Google) shows the Interstate Batteries van apparently at fault. According to a statement from Google, the autonomous vehicle’s “l(fā)ight was green for at least six seconds before our car entered the intersection.”

Witnesses said that after the crash, “dazed Google employees” sat waiting for a tow truck.

The crash is in significant contrast to recent incidents involving Tesla vehicles driving in Autopilot mode. Those crashes, including one that was fatal, suggested at least some failure of the vehicles’ detection systems to properly read and respond to their environment. But it’s hard to lay much blame on Google’s system for being t-boned by a van running a red light.

Instead, the crash shows a much different problem for driverless cars—that they will continue to share the road with fallible human drivers for decades to come. A recent Goldman Sachs report points out that, at least using current replacement rates and ownership models, it could be 2060 before the North American auto fleet reaches full autonomy.

Under those circumstances, a new sort of question arises—how much responsibility will autonomous systems have for anticipating and avoiding the errors of old-fashioned human drivers?