[ImPostingOnHN]

The verified, raw data of at least 1,000 other people's worth of data, so the data has a chance of being statistically significant, rather than 1 random dude out of billions posting on the car company CEO's website (on which said CEO is infamous for moderating content to suit his views and ego).

[ggreer]

That’s what it will take for you to believe that someone has gone thousands of miles without intervening? I don’t think that even Waymo (which has thousands of vehicles that have gone much farther without human intervention) has met the criteria you’ve set.

[UltraSane]

If one person can do this then thousands should also be able to.

[ggreer]

> If one person can do this then thousands should also be able to.

That’s not at all true. Very few people take the time to drive thousands of miles in a short period of time and document it while never intervening for any reason (not even accidentally bumping the wheel).

Anyways, I remember you. You claimed that Tesla would never remove safety drivers from their robotaxis.[1] I tried to get us to bet on this prediction but you never replied.

Well, Tesla has removed safety drivers from some of their robotaxis, meaning your prediction that their technology is “never going to be reliable enough” was falsified within a few months.

1. https://news.ycombinator.com/item?id=45661368

[UltraSane]

It is true because millions of people own Tesla's with FSD. Moreover if FSD really worked this well Tesla would be happily publishing this data at tesla.com/FSDstats . But instead Tesla is very secretive with their FSD data because it overall isn't very good.

"Tesla has removed safety drivers from some of their robotaxis"

Why doesn't Tesla has as many empty Robotaxis as Waymo has cars? Because Robotaxis aren't good enough. The entire Robotaxi rollout is a very carefully choreographed smoke and mirrors show to inflate Tesla share price. Tesla doesn't even have the permits to operate Robotaxis in California and probably never will because they are actually stricter vs Texas.

[ggreer]

> Tesla doesn't even have the permits to operate Robotaxis in California and probably never will because they are actually stricter vs Texas.

I'd like to note that in just a few months, the goalposts have moved from "Tesla will never get rid of their safety drivers in their Robotaxis" to "Tesla will never operate Robotaxis in California."

As before, would you like to bet on your prediction? I'm willing to wager you any amount up to $1,000 that before the end of 2028, Tesla will have Robotaxis in California, available to the public, without a safety driver. Note that this may not require a permit for deployment, just driverless testing, as that is how Waymo currently operates.[1]

December 31st, 2028 may seem like a long ways off, but it's much sooner than never.

1. https://www.dmv.ca.gov/portal/vehicle-industry-services/auto...

[UltraSane]

"Tesla will never get rid of their safety drivers in their Robotaxis"

They didn't though, they just moved them to a chase car. Because Elon Musk is an incredibly successful con artist. Why isn't every robotaxi driverless and open to the public like Waymo is? Tesla is so far behind Waymo it is laughable.

https://www.reuters.com/business/autos-transportation/musk-t...

Tesla touts California robotaxis but does nothing to get permits

Documenting driverless testing miles is critical for a series of permits

Tesla logged zero miles last year in California for the sixth straight year

Waymo documented 13 million miles over a decade before securing driverless ride-hailing permit

[ImPostingOnHN]

> Very few people take the time to drive thousands of miles in a short period of time

Luckily that's not a requirement here. If your 1 example, your 1 datapoint isn't a fluke, then someone driving 100 miles should have 1/10th the interruptions as someone driving 1000 miles.

If you have to intervene 0 times in 1000 miles, and that isn't an outler, then we should expect to see 0 intervention required in all 100 mile drives, all 10 mile drives, and all 1 mile drives. Unless it's a fluke.

Statistics > 1 anecdotal post about a car on the car company CEO's website.

[bdangubic]

Tesla is on fire, took them 12 years to have few cars drive from point A to point B (not always safe but progress nonetheless)… suprised their P/E is not in the many thousands with this amazing feat they accomplished, well ahead of all of their competitors

[ImPostingOnHN]

Happily, I don't care about Waymo. I do care about statistical competency among the population. Everyone should know that a single data point is irrelevant, as it could be noise just as easily as it could be signal. It is statistically insignificant.

You can actually do the math to find out the sample size you would need to derive a statistically significant conclusion. No need to be incredulous at my answer to your question: it's basic statistics.

As an aside: in general, if you find yourself going to a single person for data and expertise, you might be falling victim to a personality cult.

[ggreer]

The reason I'm using a single example is because we don't have comparable statistics across manufacturers, and because it's enough to demonstrate my point about Tesla's autonomous capabilities versus other brands. I am saying the same thing you are: Examples exist of FSD being used for thousands of consecutive miles without intervention. But you cannot find such examples for any other consumer car brand today. If Blue Cruise, Super Cruise, Mercedes-Benz Driver Assistance, or any other technology went similar distances without any human intervention, then it would be accurate to claim that, "Other brands have had self driving features for years now. Some even operate at a higher level of automation." But we don't have such examples, so it's not accurate to make such claims.

Nowhere in this thread am I claiming that FSD is safe enough to be used without human oversight. I'm also not claiming that Tesla has delivered on their promises (they obviously haven't). I am comparing the capabilities of autonomous systems using evidence that is publicly available. Since we don't have comparable statistics across manufacturers or some sort of standardized road test, I think "miles between human interventions" is a useful measure of autonomous capabilities. That's what has been used to demonstrate safety of other autonomous systems, and I see no good reason why such a metric should be ignored in this case.

[ImPostingOnHN]

> Since we don't have comparable statistics across manufacturers or some sort of standardized road test

We don't have data on the intervention rate of Teslas, either. So far all you've presented is a single, non-statistically-significant anecdote.

>I think "miles between human interventions" is a useful measure of autonomous capabilities.

Exactly. What is Tesla's average miles between human intervention? The anecdote you presented is maybe 1/1000th of the minimum data you would need (of randomly selected participants) to answer that question. 1,000 is a bold claim, it requires statistically significant evidence