Three recent autopilot/cruise control accidents

[Boatguy]

When you are sampling data, you need controls. When the entire population (Tesla drivers) are all reporting data, you don't need controls. It would be interesting to know, since they have ALL the data, is there a preference for people in various regions to engage/not engage autopilot. But in the meantime, they are perfectly able to say that Teslas on Autopilot are 50% less likely to be in accidents than Teslas not on Autopilot..

I beg to disagree. Both populations are a non-random uncontrolled sample of all miles driven. Trying to draw an inference about one sample compared to the other is not statistically valid.

Imagine you give a 10 people each a rifle and send them into the forest to hunt for deer in two groups. Five are experienced hunters, five are not, we have no idea who is in which group. We also know nothing about the terrain into which they are sent or the deer populations in each forest. The first group returns with three bucks, the other with zero, we can't infer anything about one group over the other because there are no controls on all the other variables.

Before Elon can make that inference, he needs to at least control the groups for the driving conditions (traffic, speed, road type, etc.).

 

[Papafox]

Boatguy, you have a valid point that the driving done with AP on is not going to closely reproduce the type of driving done without AP, and an error is of course present in such a comparison. For example, we don't see snow driving with autopilot on and that driving is more hazardous than general driving. Your point is well taken.

Nonetheless, a statistic of being twice as safe as without autopilot on does still tell us something important. What percentage of accidents occur in conditions too severe for AP use? I would guess less than 10-15%. The AP on safety statistics are so much better than driving without AP that I would guess that AP on driving is indeed considerably safer than driving in the same conditions without AP. Maybe it's not twice as safe, but it is still considerably safer.

I had the experience a couple months ago to be driving on a portion of I-80 in Northern California that was very well painted. Unlike coming over the mountains on I-80 (where water on the highway made AP use unworkable), this segment allowed for autopilot use with wet roads and the autopilot was performing beautifully. I was amazed at how much better the TACC could detect changes in the speed of the car in front of me before my eyes detected the change. The weather was bad, the roads were wet, but on this particular highway I could use AP and it added immensely to the safety of the drive.