As many advances as Google's autonomous cars are making, they still haven't exceeded the performance of human drivers.
On Tuesday, the company announced that its engineers had to take control of its self-driving vehicle 341 times between September 2014 and November 2015, as reported by Wired. Google's autonomous cars covered 423,000 miles over that duration, making the amount of times humans had to step in not as much as it may seem.
Taking the data further, as reported by Wired, 272 of the 341 instances that an engineer had to take control of Google's self-driving car was due to the "overall stability of the autonomous driving system" such as communication and/or system failures.
The other 69 takeovers were "related to safe operation of the vehicle" or occasions when the self-driving car possibly made a poor decision on the road — something that should be more concerning to Google than the previous instances.
Google told the California DMV that 13 of those 69 incidents would have led to accidents had a human not taken over the wheel.
Google's technical lead behind its driverless cars, Chris Urmson, told Wired that he's concerned with how the vehicles make decisions and that the company must make its autonomous driving technology — software and hardware — bolstered before the cars hit the road.
That being said, it's not as simple as a quick coding change.
"We're not really in a situation where we say, 'We should have put a one here and we had a two,'" Urmson told Wired.
Still, 341 total instances over a year's time is not a high number and Google continues to improve its autonomous technology toward its 2020 target date to hit the road.
