Researcher Claims That Self Driving Cars Only Need To Be Marginally Safer Than The Average Driver

Imagine a scenario in which a control group has ONLY autonomous vehicles, and in this group, the number and severity of accidents decrease at a *slightly* faster rate than the national average, which is comprised virtually exclusively of human drivers.

What argument do you have now against the autonomous cars in the control group? (Assuming they don't introduce new issues, such as more gridlock, no-movement-stalemates, more stranded motorists, etc.)

This is an example of autonomous cars needing to be only slightly better than their human counterparts. If they are, in all aspects, then we'll have a slightly better driving reality.

Some folks argue "change nothing unless the new state is perfect." Other people believe "better is better... go for the incremental improvement, but that does not mean we should stop improving."

It's just a matter of which perspective you choose.

BobM: You bring up a valid point, of course, in that an autonomous car being *theoretically* safer than a human driver is different from being *actually* safer.

In my post, I was speaking of a (fictitious) scenario, in which autonomous vehicles are *actually* a little safer than their human-driven counterparts, and not just theoretically safer.

This distinction between actual and theoretical is one reason why I am concerned about recent discussions by US lawmakers to loosen laws pertaining to autonomous vehicle safety. I obviously don’t support that.

The problem could be that the "measuring stick" if you wish is a little different, it is not objective. It is hard to be when people's well-being or lives are at stake. A machine killing humans is not acceptable (even if it was deployed by humans). Humans killing other humans?! Happens all the time, excuses are given, empathy is misplaced or missing.

^^^
vdiv: I get your point! :-) Today, there is very little outcry about the approximately 38,300 people killed and 4.4 million injured on US roads in 2015. We, Joe Public, just accept this as normal (or we’re not aware of it). But when an autonomous car messes up, it’s all over the news, in print, radio and TV, getting repeat coverage.

If the numbers PROVE that autonomous cars have meaningfully fewer accidents, and less severe ones, per 100,000 miles traveled as compared to the existing human-driver numbers, then we need to take notice. That fact needs to be well advertised to help shift public perception away from the highly sensationalized and fear-perpetuating media, and onto the facts.

While this is not base on imperial evidence I would dare say that an autonomous car should be safer just because the driver cannot be distracted by screaming kids, the accident in the lane next to the car, temperature, sun brightness, radio, texting or talking to someone in the car or on the phone, eating, doing a crossword, doing one's nails etc. However from what I have read autonomous vehicles need to be a bunch better at the driving part before they can be trusted to take the wheel in all circumstances, as active winter weather is still their downfall.

Waymo is starting to deploy full Level 5 autonomous service in Phoenix. The "safety driver" is only there to push a button in the case of an emergency which tells the car to pull over. Self driving cars are here, whether you are a fan or not.