Imperfect self-driving cars should be rolled out now to prevent more crash fatalities in the long run, a new report has found.

The RAND Corporation recently published "The Enemy of Good", which explains that waiting for completely crash-free automated cars would cost more lives than it would save. The paper compares self-driving cars that are 10% safer (than human driven cars) with cars that are 75% or 90% safer.

Results show that about 3,000 lives could be saved a year by allowing automated vehicles on the road now, despite only improving on human error by 10%.

In the report, RAND researchers Nidhi Kalra and David Groves estimate that "hundreds of thousands of lives" would be saved if humans did not wait for autonomous car error to beat human drivers by 90%. "Waiting for the cars to perform flawlessly is a clear example of the perfect being the enemy of the good," Kalra said.

More than 90% of car crashes are caused by human errors such as driving impairment, speeding or not reacting to another driver's actions. "We've all heard the argument that autonomous vehicles are never drunk, distracted, or tired," Groves said. "They could reduce the huge number of crashes involving these factors."

To be considered an autonomous vehicle, the car can ask for assistance from a human driver, but must do most of the driving itself. The highest level of automation requires zero input from a driver. Automated steering and throttle features do no qualify a car as automated.

The report estimates that more than one million lives would be saved by 2070 if automated cars are rolled out by 2020. If the release is delayed by 20 years to 2040, only about half of those people (580,000) would be saved. These figures apply to the United States.

The UK suffered 1792 road deaths in 2016, an increase of 4% when compared with 2015. That is almost half the total from 2006 (3,172).

Researchers at MIT have been developing artificial intelligence that can help automated vehicles make moral decisions. The software would allow self-driving cars to crash in the same way a human would.