[minhazm]

> Hard to compare that to a level 4 company where the car can go hundreds of thousands of miles without disengagement

This is nowhere close to being true. The latest numbers I could find for 2022 are 17,060 miles per disengagement[1] for Waymo. And even then the definition is not quite clear, because these are safety disengagements from what I can tell, and not disengagements for things like getting stuck because the road is closed or something else.

[1] https://www.eetimes.com/waymo-cruise-dominate-av-testing/

[danans]

> This is nowhere close to being true. The latest numbers I could find for 2022 are 17,060 miles per disengagement[1] for Waymo

Tesla doesn't reveal it's miles/disengagement stats, but according to crowdsourced data it's 690-828 miles between critical disengagement [1], which is 2 orders of magnitude worse than the Waymo number you posted, and far below their goal of besting human drivers' abilities.

1. https://electrek.co/2024/10/24/elon-musk-just-said-some-wild...

[bushbaba]

Regardless we are probably about 10-50x the disengagement rate needed. It’s about 1 accidents every 500k miles for a human driver.

[danans]

Your 1 accident per 500k miles stat is an average across urban and rural areas. But accident rates in urban areas are likely significantly higher than in non-urban areas, due to higher vehicle density/congestion, despite the fact that vehicle accident fatality rates are higher in rural areas[1]. Urban areas tend to have a lot more low-speed no-injury or mild-injury collisions. When you account for all of those, AVs are in shooting distance of human driver accident rates in urban areas. Also, the accident rates are much higher for drunk, sleepy, or distracted drivers, so to the extent AVs reduce those types of driving, they bring the overall accident rates down much further.

1. https://crashstats.nhtsa.dot.gov/Api/Public/Publication/8135....