Since the fatal crash that killed the driver on one of Tesla's autopilot cars occurred in May, arguments on whether its autopilot vehicles are safe or not have come up.
Last Oct. 4, California statesmen instructed Tesla to drop the term "autopilot" when advertising their newer car models unless and until the vehicle can actually drive on its own without human interference.
This verbal skirmish is not the only one happening between Tesla and the rest of the world. Tesla CEO and founder Elon Musk even wrote Fortune Magazine an email last July stating that if the company vehicles' autopilot features were utilized by many, it would be saving more lives than going without it.
"Indeed, if anyone bothered to do the math (obviously, you did not) they would realize that of the over 1M auto deaths per year worldwide, approximately half a million people would have been saved if the Tesla autopilot was universally available," Musk wrote.
Luckily for the company, there are others defending the autopilot systems' safety. Numerous publications like The Verge or Vanity Fair believe that the autopilot cars are safer than manual driving.
But have Tesla's autopilot vehicles actually helped in preventing accidents? Although this may be a possibility, factors such as lane departure warnings and adaptive speed controls will play a role in making driving safer someday.
Sample Size
Events like fatal car accidents are uncommon; hence the data needed to obtain an accurate result can be massive.
"Autonomous vehicles would have to be driven hundreds of millions of miles, and sometimes hundreds of billions of miles, to demonstrate their reliability in terms of fatalities and injuries," said RAND Corporation in a study published in April.
Note that statistically, even with the one death in the millions of miles cars traveled in autopilot being a good indication, it is ridiculous to declare that driving with autopilot on proves to be safer than going without it. Doing so is like using one patient's lab results to conclude a drug's [in]effectiveness.
Sample Class
Size is not the only thing to consider, statistically speaking.
Even while having millions of observations, a biased result is still plausible if the sample class is indeterminate. Just like how a group of soldiers is an unreliable sample to the world's overall health status, the 130 million miles traveled in autopilot mode may be unstipulated to road safety.
"It has no meaning," said Princeton University's director and professor of transportation program Alain Kornhauser of the comparison between Tesla's own vehicular data and the nationwide statistics in the United States.
Pointing out that even if highway driving might make for safer conditions since autopilot mode was designed for such, there still are broader driving conditions to consider with regard to a system's security efficiency.
Musk mentioned that Tesla will be sharing some of its autopilot system data with the U.S. Department of Transportation and with other manufacturers. Details on what specific info and when it will be handed over has not yet been discussed.
