Autonomous Fatalities

When it comes to the future of transportation I do all my math on napkins, because they clean hype the best. If you manage to stay awake over the next 2 paragraph of numbers, you’ll understand the industry better than 99% of the people.

Waymo objectively shown to be the market leader in the last disengagement report where the driver had to avoid a crash every 11,017 miles [1], half less frequent than the first loser. Project prometheus, in 1986 had a disengagement rate of once every 100 miles [2]. That’s an improvement of two orders of magnitude in 30 years. Impressive, right?

Human drivers get into accident every 165,000 miles. That’s 15 times less frequent than the 11,017 miles reported by Waymo. Eight Billion miles are driven in the US every day. If you were to made all cars autonomous, there would be 740,000 crashes every day. It is estimated that 0.7% of car crashes results in a fatality, that means 5000 autonomous fatalities every day [3]. There are 109 fatalities every day in the states, if you multiply that by 15 (how better are humans than robots) you get 1635 instead of 5000, our math isn’t exact but is approximately correct [4].

Just think about this for a second, if we release our best car on the street, 5000 people would die daily. Uber’s program would disintegrate if they killed another pedestrian, Waymo’s positive image would similarly flip and neither of them are willing to take any risks. The safety driver stays.

The bottom line, for Waymo to have a single fatality a day, the disengagement rate would have to be 5000 time less, or once every 55 million miles. This would avoid 108 of 109 fatalities every day. Will society allow a company to have a fatality each day? Will venture capitalist fund the research until we reach this level of safety?

I've been try to explain in simple words the state of the autonomous industry. Read my other post on why autonomy is hard.