[gamblor956]

Hint: one of those has driven at least two if not three orders of magnitude more miles unaided by humans.

Yes we all know that Waymo has done this but this thread is about Tesla.

3-D spatial awareness component of Tesla’s stack has been solved problem for a couple of years now

A lot of people have died because their self driving Teslas did not have spatial awareness and crashed into stationary objects at full speed. Clearly we have different definitions of construction "solved".

[simondotau]

It's critical to note that Autopilot is not FSD, and has nothing in common with FSD, other than running on the same physical hardware. Autopilot is a brand name for lane centering / adaptive cruise control, similar to Nissan ProPilot and Ford Co-Pilot. Deaths attributed to the misuse of lane centering / adaptive cruise control features are commonplace among all makes of car, but deaths in Fords and Toyotas rarely make media reports, because the media is in the business of generating clicks.

A lot of people have died because their self driving Teslas

Two is not "a lot".

In the USA in 2023, the fatality rate of human drivers was 1.26 fatalities per 100 million vehicle miles in the USA.

Over the 60 million miles driven by Waymo vehicles, the number of fatalities involving a Waymo is one. This is a fatality rate of 1.67 deaths per 100 million miles. (It should be noted that this isn't a statistically meaningful number because Waymo vehicles have driven so few miles.)

Over the 3.8 billion miles driven by Tesla FSD, the number of fatalities involving a Tesla running the FSD software stack is two. This a fatality rate of 0.05 deaths per 100 million miles. If those 3.8 billion miles were driven by humans, you would expect 46 more deaths to have occurred.

[tzs]

One big difference is that those two fatal accidents involving Tesla FSD were both cases of the Tesla driving into someone (a pedestrian in one and a motorcyclist in the other).

The one fatal accident involving a Waymo on the other hand was an empty Waymo that was stopped at a red light along with several other cars when another car (a Tesla, BTW) going just under 100 mph slammed into that group of stopped cars from behind, killing a passengers and a dog in one of them, and injuring people in several of them.

[SR2Z]

> Two is not "a lot".

There have been zero from the current players.

> Over the 3.8 billion miles driven by Tesla FSD, the number of fatalities involving a Tesla running the FSD software stack is two.

Then why doesn't Tesla publish a full accounting of their crash reports?

[gamblor956]

All of your statistics are wrong.

Tesla advanced driving features (AP and FSD) have killed more people than every other automaker‘s comparable systems. Combined.

Tesla FSD killed 2 people this year alone. No other automakers self driving systems have killed or even seriously injured anyone.

Elon Musk isn't going to give you a job for shilling for him on HN so stop embarrassing yourself.

[simondotau]

All of your statistics are wrong.

Your first claim is baseless, as no statistics exist to make that claim. Tesla is the only major manufacturer which voluntarily provides all crash information to regulators. No other major manufacturer collects and discloses comparable statistics. Therefore it cannot be asserted that ADAS-involved deaths are more frequent in Tesla vehicles.

It's often said that Autopilot is uniquely dangerous because it gives drivers a false sense of confidence. This is a "feels true" claim. However, many other manufacturers have ADAS systems which exceed the technical AND subjective performance of Tesla Autopilot. On that basis, we should assume that ADAS-involved fatality rates are higher with those brands.

Your second claim exceeds the data published by both reputable AND biased sources. Even if true, that still places FSD well over an order of magnitude more safe than human drivers, with a supposed 3 deaths compared to the expected 48 deaths with human-piloted vehicles.

As for your last sentence, overuse of cutting comments makes them dull.