Hacker News Reader

> This is pretty normal in marketing demos. For example, in the original iPhone presentation: The engineers identified a “golden path,” a specific set of demo actions that Jobs could perform in a specific order that afforded them the best chance of the phone making it through the presentation without a glitch. For example, Jobs could send an email and then surf the web, but if he reversed the order, the phone tended to crash. https://www.internethistorypodcast.com/2017/01/the-history-o...

And by the launch of the phone that was fixed... I fail to see how that's the same thing.

> And by the launch of the phone that was fixed... I fail to see how that's the same thing.

I mean... they haven't launched Full Self Driving. You can (for money) get a beta that requires you to remain in control of the car and often makes mistakes. But the beta does most of the stuff shown in the video.

> But the beta does most of the stuff shown in the video.

Today, after ~7 more years of development? Can it do them consistently or "80% of the time it works every time"? Because that's the problem, the bar for an autonomous driving system to be useful is high.

Below that bar you either have just a set of driver assists that have no claim at real autonomy, and have existed for (relatively) cheap for years, or you have the the "Goldilocks zone" of dangerous tech, smart enough to coax you into handing over the control but too dumb to be actually capable of taking over.

None of this is relevant to my comment.

He is saying that "FSD" is unlikely to launch any time soon so either you spin it in your favour or you don't.

It's hard for some to see this point of view unfortunately, no matter how you try it. Elon has been intentionally skirting and blurring the line between lying and marketing for a while now.

They sold FSD as something coming soon. I considered buying a car with it in 2019. Had I bought that car, it would likely have hit the end of its 10 year lifespan before FSD ships. Assuming that Tesla ever ships FSD for a ten year old vehicle.

Not just coming soon. In 2019 Musk claimed it would be financially insane not to buy a Tesla because by 2020 it would become a robo taxi earning you tens of thousands a year.

They have launched Full Self Driving - it's being sold to general public, it's being advertised to general public, and it's being used by general public as a product; the legalese fine-print disclaimer about 'beta' doesn't change that.

But not in the sense of the video’s claim, that you do not have to interact with the vehicle. The beta is a driver assist system you must carefully monitor and intervene, the final product is autonomous.

FYI, Tesla does not sell anything called “Full Self Driving.” That people think they do is a product of their [probably intentionally misleading] marketing. Tesla sells “Full Self Driving Capability.”

Just like I have the capability of running a marathon. I just haven’t done it. And I probably won’t…

That is what most people would call "a distinction without a difference".

I think lawyers would disagree, and that the extra word is there to limit legal exposure.

Do you actually have capability to run marathon? You know you would finish before time runs out and you know your body would not be able to complete it? Untrained people can't and get injury, get sick or just can't continue.

Full Self-Driving Capability is not capable of full self-driving.

(An honest name might be “Full Self-Driving Aspirational Future Upgradeability with Limited Self-Driving Capability”, but that probably wouldn't sell as well since it would set realistic expectations.)

I have FSD and it is a joke. It is so bad I don’t know where to begin. This software has no place on a public road. All it’s good for is tech demos. They are going to kill people who were fooled into thinking FSD actually works.

An exit from beta has been promised for years. I think this is what most people take issue with.

https://youtube.com/watch?v=zhr6fHmCJ6k&feature=shares

If a phone crashes what’s the worst that can happen?

However if a car crashes you can put multiple life’s in danger, isn’t it?

I think the opening is what separates this from "not technically lying". By leading the video with "The person in the driver's seat is only there for legal reasons" when they know for a fact that the vehicle had previously crashed on that exact route with that exact software they have in my mind elevated it to an outright lie.

I interpret the opening claim as "if it were legal for us to film this without a driver we would have", and I think that's probably true. But while that's reflective of a pretty worrying safety attitude for a car company it's not lying.

> I interpret the opening claim as "if it were legal for us to film this without a driver we would have"..it's not lying.

I interpret lying as reclining your body in a horizontal position.

They would not tried that without a driver. I think even Elon has that level of restraint and sense

I wonder if we're disagreeing on what they thought the odds of an embarrassing incident were or whether those odds would have been acceptable to them?

The video is 3:44, and I think doesn't have any bits cut out. I'm guessing that after running the route many times for practice and giving it extra data they could have gotten down to about odds of somewhere around 1:10k. And I think they would have accepted those odds.

(I would definitely not accept those odds, and would not want to ride in or near a car made by people who would. I'm glad that the law prevented them from doing runs without a safety driver at that stage. And I don't think they're anywhere near ready to do that today either.)

> I'm guessing that after running the route many times for practice

Why practice? They did the run until it worked well for the video. It wasn't a demo with a live non-Tesla audience. In this case not working 9/10 times is good enough rather than 1/10000

They did practice, though?

The odds I'm talking about are not "it doesn't work as well as the video" but instead "things go horribly wrong in a public way".

There is no reason to think it was better than today's FSD even with those "crutches" (as I believe Elon himself refers to them when talking about other SDC companies), and the chance unsupervised FSD does something horribly wrong in a public way on real streets even in 3.4 miles is far higher than 10000:1 today. After all the car did hit a barrier during the making of this video. I doubt they spent much effort improving the odds when they had infinite takes.

Looking around I see something like 6mi per disengagement [1] which, yes, is unacceptably high. The video is probably about 2mi, which naively would give 1:3. But disengagements have a large amount of precaution, so this isn't an accident rate. And some accidents wouldn't be ones they'd need to be public about, like the one in testing which only came out years later.

[1] https://electrek.co/2022/12/14/tesla-full-self-driving-data-...

Not just crashed, but also:

    Q. And isn't it true that during the various
       attempts to generate the video, there were instances
       in which the person sitting in the seat had to intervene?
    A. Yes.

The charitable interpretation is "The person in the driver's seat is not interacting with the car in any way in this video".

I don't think that's a reasonable interpretation considering the next line is: "He is not doing anything. The car is driving itself"

Why say the exact same thing twice? I think its clear Tesla wanted to make a stronger claim than "he happened to not do anything this particular time"

> The engineers identified a “golden path,” a specific set of demo actions that Jobs could perform in a specific order that afforded them the best chance of the phone making it through the presentation without a glitch

A glitching IoS is mildly annoying, a glitching autopilot kills you.

> a glitching autopilot kills you.

Or worse - others.

You at least had control over the decision to use FSD (bad already, given human driver psychology) and then not pay attention to be able to intervene in a split second (which is ambitious to begin with especially over periods longer than a few seconds).

Below [0] is a link to a comment I recently made about a very dangerous incident I myself experienced with only a much more mundane "lane assist", on a German autobahn, while fully engaged and in control. Imagine I had had something like FSD and had required even a split second longer to get control of the car back... it would probably have been too late by then. I don't think it's a good idea to think well, I can just disable assistants, or FSD, unless conditions are perfect. Some day you will forget.

[0] https://news.ycombinator.com/item?id=33986939

Milton, ex-CEO of Nikola, also technically didn't lie about their truck demo (it was rolled down the hill, and said it was "in motion").

He was found guilty of fraud, and will be sentences soon, facing up to 25 years in jail.

If you read the actual fraud charges [1] it does list the demo video, but along with very clear lies like saying it "fully functions and works" at an unveiling for the Nikola One or "Nikola is producing [hydrogen] well below $4 a kilogram" when the charge says In fact, Nikola has never produced any hydrogen at any price, nor at the time could it have produced hydrogen for below $4 per kilogram. To the contrary, Nikola has never obtained a permit to produce hydrogen or installed the equipment necessary to produce hydrogen. At the time that MILTON was claiming that Nikola was producing hydrogen for less than $4 per kilogram, it was in fact purchasing hydrogen from a supplier for $16 per kilogram.

(Something can also be fraud without being a lie.)

[1] https://www.justice.gov/usao-sdny/pr/former-nikola-corporati...

> but along with very clear lies like saying it "fully functions and works"

"The driver is only in the seat for legal purposes. The car is driving itself."

Not in any way that any other Tesla on the road could drive itself, with HD maps and custom tuned firmware for that route.

A person seeing it would reasonably conclude that the claim wasn't specific to this exact situation, which makes this misleading, but not technically lying.

(I still think it's bad, and am not a Tesla fan)

It is a lie by omission, some facts are so important to a statement that eliding them is lying. You can't just say that the driver wasn't necessary when he had to intervene in your other tests, it is like saying that the safety net for your line dancer isn't needed even though he would have fallen and died many times over without it. Such lies kills people.

I didn't watch that trial closely because the fraud seemed so obvious and the company was still a fledgling (?)

The Musk / Tesla fraud also seems extremely obvious, but even tho Musk has haters, he's made a bunch more people tons of money, I would expect him to have much better lawyers, etc.

Problem is, now it's years later, the product has been in the hands of consumers for a while now, and yet it still doesn't work. People are pissed off, people are getting hurt, and yet he continues to insist on calling it "Full Self Driving" when it is very far from that. If the government can't do anything to protect consumers for whatever reason, then the consumers must take it to the courts and to the media relentlessly until the knowledge of what FSD really is becomes a part of the zeitgeist. Much value and many lives will be destroyed needlessly because of one snake oil salesman holding a whole bunch of money.

> This is pretty normal in marketing demos.

Except Tesla kept on with the lie for years. Tesla has already copped to false advertising around "full self-driving":

https://electrek.co/2022/12/12/tesla-ordered-upgrade-self-dr...

Now Tesla doesn't want to be done for fraud so their current spin is that they have failed at full self-driving but "failure is not fraud":

https://electrek.co/2022/12/07/tesla-self-driving-claims-fai...

It's not helping Tesla's case that they took money from customers for an "aspiration". Tesla should do the right thing and refund everyone's money.

at this point the fsd refund is basically pocket change for the company, not sure why they wouldn't do it or at least offer the option

These are not similar cases at all. Any competent engineer can work through the iPhone bugs and fix them. That's well-trodden ground, the sort of thing thousands of engineers do each day. Fixing a self-driving neural network is a research topic. Even years later it's not a fully solved problem, and that should have been clear in 2016.

And of course they fixed them before the public release of the product (or shortly after).

Teslas FSD is released for years (which they call beta for marketing / libility) and is still missing a lot of features (including the ability to work without human oversight) and being quite buggy on others.

> The engineers identified a “golden path,” a specific set of demo actions that Jobs could perform in a specific order that afforded them the best chance of the phone making it through the presentation without a glitch.

In a past life, when I sold my soul to the devil, I did a lot of demos on behalf of Microsoft that were precisely scripted. And, even then, NT4 BSOD'd on me in front of a packed auditorium once.

That's why one of the key skills of a corporate public face in IT is stand-up comedy. They didn't see the flawless ISS crash containment they were promised, but at least they had some great laughs.

Funny now that the devil isn't that much of a baddie anymore.

Did you read the actual deposition? The linked article is not clear on the exact details, I don't think it's possible to come to any conclusions from the article alone.

The iPhone example is an interesting one. I would say that was still a 100% legit demo, because Jobs was demonstrating the actual working features (rather than say bring up screenshots of a web browser and claim to be surfing the web.)

I can't tell for sure whether the Telsa video is closer to the iPhone demo or the Nikola semi demo (which was obviously way worse.)

In the context of using a neural network model this is very much lying.

A phone is not a car. You can't die by believing in a feature of a phone but you potentially can or be hurt when it comes to a car.

They also crashed the car trying to park it at the end.... I wonder how many takes it took to get the demo working.

People don’t die if an iPhone doesn’t work right. Autopilot/“fsd” on the other hand…

Two wrongs don't make a right.

> This is pretty normal in marketing demos.

This was my thought, isn't this stand operating procedure for pretty much all companies? How is this news?

If you put "Musk" in the headline you get more clix.