As we enter the age of autonomy, traffic fatalities in the United States have reached their lowest level since 2014, according to annual data released by NHTSA’s Fatality Analysis Reporting System.
NHTSA reported 36,096 fatalities in motor vehicle traffic crashes in 2019, down 2 percent from 36,835 in 2018 even as travel rose 0.8 percent.
Auto News
Remember, this is in 2019 –– before any of the behavioral changes associated with the global pandemic. 36,096 people killed on the road is about 98 American lives lost every single day. This decade, we have the opportunity to bring that number down to single digits, eliminating 90%+ of fatal accidents with advancements in automation and advanced driver assistance.
This is a big deal. Think about a room full of 98 people, and the damage to their family, their friends, their work, if all of their lives were suddenly lost violently and unexpectedly. That happens every single day, and those are just the numbers for the United States. Globally, 1.35 million people are killed on the road every year –– around 3,699 people every day. And let’s not even get started on how many people are injured. So far, we’ve accepted fatal errors as part of human nature. This decade, we no longer have to. Drivers will now start to be updated over the air.
The fatality rate fell to 1.1 per 100 million vehicle miles traveled, the lowest rate since 2014, compared with 1.14 in 2018. Alcohol-impaired driving fatalities decreased to the lowest percentage since 1982.
Auto News
Great to see drunk driving deaths going down, and 100 million miles traveled between every death. Working backward, that means about 3.28 trillion miles were traveled in the United States in 2019. Even if Tesla Autopilot travels 2 billion miles a year, that’s just 0.06% of miles traveled in the U.S. Right now, autonomy and driver assistance are a drop in the bucket in terms of their market share of miles traveled. The total distance driven by computers will have to grow 1000x – 2000x to overtaken human-driven miles. That poses a lot of challenges.
When you launch autonomy or driver assistance systems at scale, you start to hit some pretty big numbers. That’s going to pose some PR problems, especially since there are a lot of people who want to generate sensational headlines to capitalize on the public’s fear of automation (including through short-selling bets). Consider Tesla Autopilot: Lex Fridman’s latest pre-pandemic estimates put the average number of Autopilot miles traveled per day for 2020 at 7.12 million miles a day.
If Tesla’s FSD software had the exact same fatality rate as the U.S. average –– essentially, if you have software that “does nothing” to make things better or worse –– someone would die with FSD engaged roughly every two weeks. That’s not going to look good in the press, even if this hypothetical software really didn’t “do anything” and the number of fatalities would have been unchanged whether it was on or off. This is a really tough minefield for the entire industry to navigate, and there are no easy answers here. Autonomy will benefit all of humankind immensely, but someone is really going to have to stick their neck out to get us there.
My guess is that the current FSD beta software would have a lower fatality rate than the U.S. average when monitored attentively by a human. Since the software generally chooses to drive more cautiously than a human would, the rating should improve while human monitors catch any potentially dangerous mistakes made by the software, disengaging as needed. However, at this time the software left unmonitored by a human would probably have a higher fatality rate than the U.S. average. It will get better than a human over time with over the air updates, but it isn’t there yet. I don’t have any data to back this up, but I’m just trying to guess based on my experience.
So really, Tesla’s FSD Software with a human monitor must be 10x better than a human just to get to 140 days without a fatality involving FSD. That’s a high bar, and it’s going to be a tough sell to convince the public that this is normal when they read about these deaths all over the news.
Let me know in the comments if I got any of my napkin math wrong.