Google’s Self-Driving Car Crash

Google released a 32-page report last month containing data during 15 months of testing their fleet of self-driving cars on California’s public roads, which showed that there have been 272 situations where the Google car disengaged from its autonomous drive mode and immediately turned control of the car over to the Google employee driving it.  In total there were 13 instances that would have resulted in a crash left up to the automated driving software.

Google leads the way in terms of their search engine, advertising and pay per click advertising, but this recent data undermines their recent foray into building cars without manual controls. Fortunately all the near-misses recorded in the new report resulted in a safe outcome, but only thanks to the human driver intervention in each case, which surely discredits the premise of the experiment.

Chris Urmson, the Director of Google Self-Driving Cars stated that “software detected an anomaly in the system somewhere that could have had possible safety implications, and in these cases it immediately handed control of the car to the test driver.”

The 32-page report released in January says that during 15 months of tests:

  • Google’s vehicles drove 424,331 miles without manual control.
  • In 72 instances, the car’s software registered a “failure” that resulted in emergency control being given to the manual driver.
  • In a further 69 cases, the driver took control without being alerted by the system because they anticipated an accident themselves.
  • Computer simulations designed to replicate each incident found that in 13 of the cases where drivers intervened, a crash would have occurred if they had failed to do so.

There is no way of knowing how many of those incidents might have ended in a serious collision where someone could have been hurt or killed.

The question is, should Google continue to develop this kind of technology, or will it go the same way as Google Glass? Google insist that these cars are “safer than humans”, which may be true on average, but without the necessary data we can’t make that comparison. If the experiment was repeated with human drivers instead of Google software, perhaps we would have seen a higher number of near-misses or actual accidents. All we know for sure is that in these cases, the human knew better than the car.

Many think that driver-less cars will be beneficial to the economy, which is an interesting point. The idea is that our working lives will be made more efficient since we won’t have to concentrate on driving, which takes up valuable time we could spend working! We should also be less tired when we get to work since we didn’t have to concentrate on the road. Except, of course, that if Google’s test drivers took this attitude, 11 of them would have crashed.

In any case, the safety of this new mode of transport needs to be tested much more rigorously given these results. It does serve as a reminder, though, of Google’s continuous push towards a future where automated software understands the human brain better than we do ourselves, a trend we’re also seeing in search engine technology. What do you think about Google’s plans – are they pushing the limits too far too fast, or do we need that attitude to make progress?