Gary Marcus

There are aprox 260M cars in the US and 32K fatalities a year. So 0.0123%. There are about 200K Teslas and 2 fatal accidents. So 0.001%. I’ve googled all data. Maybe more fatal crashes and maybe more teslas. Also not all of 260M car are driven. Still an order of amgnitude safer.

you are computing the wrong stats. lots more than 2 tesla fatal crashes, i would imagine; you are comparing autopilot deaths to human-aided miles. what tesla won’t tell us is how many autopilot miles were driven without user intervention. that’s the denominator we would need.

Have you been involved in a car accident ?

They say that to sell you a car, but after an accident, it's "we told you to keep your eyes on the road."

Consider perhaps changing it around to time (hrs)driving: Tesla Autopilot engaged hrs = X Fatalities occurred in X = Y X Y = safety/risk metric Human driving hrs = A Fatalities occurred in A =B A B = safety/Risk metric Which one do you think has the most deaths per hrs driving?

"The driver is expected to remain alert and ready to take over if something comes up that the vehicle cannot handle." This is the real issue. Is that a reasonable expectation? Same HMI prob that arose in Tempe L4 crash, & that Waymo skipped over in L3, is still present in L2.