Former Tesla employees say 2016 ‘Full Self-Driving’ video was staged | Hacker News Reader

Former Tesla employees say 2016 ‘Full Self-Driving’ video was staged

thedrive.com

105 pointsincomplete44 comments

I recommend reading Alberto Savoia's book, "The Right It: Why So Many Ideas Fail and How to Make Sure Yours Succeed"

One of the examples that Savoia provides in his book is especially helpful to recall here:

Back in the 1960s, IBM was trying to figure out if they should make a massive investment in building a speech recognition product. Rather than betting the company on an expensive idea without being certain of the market for it, they simulated the speech recognizer by putting a secretary in another room with a mic, leaving the test subject with the impression that their commands were going directly into the computer in the room.

By doing this, IBM figured out that even if they made a perfect speech recognizer, their customers wouldn't use it! They were concerned about privacy, found typing to be faster and more precise, etc. Rather than investing huge sums of cash on a dud, they were able to put it towards the System 360 line, which turned out rather well :)

Right, this is really what's meant by the phrase "fake it until you make it" -- not that you should make claims that aren't true (even though they may be someday) but that you should simulate automating processes in a similar fashion to your anecdote before actually spending the time and money to automate them, even though up front you're likely to lose money on having humans do what you eventually want computers (or robots, in the case of Amazon warehouses) to do.

Sure, but the difference between savvy product testing and charlatanry is doing this as part of a quiet internal study vs. on stage at a massive public demonstration.

I’ve done that before and referred to it as “wizard of oz testing”

This is called a prototype

Is it? I would think a prototype must bear some resemblance to the final product (as the proto- comes from the Greek root prōtos meaning "first" or "primitive").

I think this is just clever market validation testing.

It completely bears resemblance to the proposed final product, no?

Market validation or tech validation, both different things to prototype for.

> It completely bears resemblance to the proposed final product, no?

Since the final product is computer software: no, I don't think a secretary bears much resemblance.

The key part being emphasized is that the audience believes its more functional than it is. The word prototype is ambiguous on that part

I've heard it referred to as a "pretotype" (god I hate words like that).

Pretotype. Interesting word; preto means dark so it fits really well.

Oh a 'prototype' like that 787 initial rollout?

https://www.aljazeera.com/economy/2014/9/10/the-fake-boeing-...

a demo?

https://en.wikipedia.org/wiki/The_Mother_of_All_Demos

Here is a good timeline (complete with Elon's tweets) for Tesla's self driving capabilities.

He has been lying about it for 5 years now and has been stringing on Tesla customers with statements like, "It's almost there!" and "Next version will blow your mind!" etc

https://www.reddit.com/r/SelfDrivingCars/comments/n6nsmt/elo...

It's stunning how he has gotten away with lies for so long.

Personally, I love being in command and control of 1000+ pounds of steel and liquid, seemingly propelled based on the angle of my foot. I wouldn't have it any other way!

I think autonomous travel would be a great experiment for locomotives on isolated tracks.

You'd be a lot less enamored if you were on or near roadways on foot, bike, or motorcycle. Or drove a truck or vehicle with a trailer.

There are several self-driving vehicle projects that have logged far better safety records than human drivers.

They have radar and lidar units.

Musk has insisted on solely cameras, which are far cheaper, but suffer from a slew of entirely predictable vision problems. Just hit up youtube and search for "tesla phantom braking."

To be fair, Teslas were using Radars until suddenly a chip shortage made all the Radars hard to find.

All of a sudden: Tesla scrubs its blog (which used to talk about all the wonderful benefits of RADAR), removes all RADAR from future vehicles, and talks about how AI has advanced to make Radar unnecessary.

Its pretty funny watching this in realtime. Surely, Elon Musk has heard of archive.org? You can't just scrub your blog and hope that people forget... in fact, scrubbing your blog makes these things more easily seen.

https://web.archive.org/web/20210418114757/https://www.tesla...

What was the famous line in 1984? We were always at war with Eastasia.

> Its pretty funny watching this in realtime

Yes, Streisanding your disregard for public safety is hilarious. Elon should do prime-time comedy.

Hey wait a minute...

While humans driving is ludicrous and it's insane how little we train drivers for it, one thing it does make easy is assigning accountability to the liable party in the case of an accident or some other catastrophic failure.

Tesla's "self-driving" burns that line into thin air, and also misleads drivers into a false sense of security. Musk needs to own up to that, because in the absence of a clear way to see who's at fault, the onus is on him to clear the air and he hasn't.

humans understand corner cases, which are basically infinite when it comes to momentum chnages and obstacle detection.

Have you ever observed humans drive?

Humans have a minimum 150ms response time, which in many corner causes is too high

> I think autonomous travel would be a great experiment for locomotives on isolated tracks.

I think that's a little past an experiment; the first one went in in 1981. At this point, I would think most new fully segregated metro lines are automated. It's ~a completely different problem to self-driving cars, and a far more tractable one.

Agreed, imagine paying thousands of dollars for a software product that is never delivered and continues to be pushed forward to the next year for five years running - it's unbelievable a solid case hasn't been brought against the company. Only a performer like Musk with such an influence on his fans could pull something like that off.

What about Star Citizen?

* https://news.ycombinator.com/item?id=14319340 [5 years ago]

* https://news.ycombinator.com/item?id=17553371 [3 years ago]

* https://news.ycombinator.com/item?id=27628957 [6 months ago]

Wow I didn’t realize it costs thousands of dollars!

You can certainly spend thousands of dollars on Star Citizen if you'd like! https://starcitizen.tools/List_of_pledge_vehicles

Not a lawyer, but don't "excusable" errors done for profit enter the legal category of fraud?

Tesla is not capable of fraud. Nikola on the other hand...

I mean all of these are missed deadlines or estimates, not lies, right?

Every few months over six years - he'd have to be an incredibly slow learner for that to have been just poor estimates all along.

Why?

> The Drive could not reach out to Tesla to verify or contest the claims, as the automaker dissolved its communications and public relations department some time ago. The New York Times says that neither Musk nor a top Tesla lawyer responded to requests for comments over several weeks.

I always thought it funny that Musk saw this as some sort of 4D chess move and not a tacit admission that his company's products/behavior suck so much, he had to effectively nail a "No comment" placard on the front door.

You know what would be awesome? Say negative PR about Tesla comes up, and someone from Tesla calls/emails/tweets a reporter to provide a statement...the reporter ignores them, or tells them to go pound sand, because given Tesla does not have a communications or public relations department, someone from Tesla couldn't possibly be contacting them in an official capacity to provide an authorized statement.

>You know what would be awesome? [...]

It makes for an "awesome" gotcha moment but thats about it. All it would really do is trash your newspaper's reputation while making tesla look like the victim. Asking for comment is a courtesy that professional journalists give to subjects. It's not some sort of reciprocal agreement.

> I always thought it funny that Musk saw this as some sort of 4D chess move and not a tacit admission that his company's products/behavior suck so much, he had to effectively nail a "No comment" placard on the front door.

That's one way of looking at it; another is that he's avoiding the spin and dishonest misleading nature of today's media. He doesn't have to deal with that BS and he probably feels a lot less stress because of it.

> You know what would be awesome? Say negative PR about Tesla comes up, and someone from Tesla calls/emails/tweets a reporter to provide a statement...the reporter ignores them, or tells them to go pound sand, because given Tesla does not have a communications or public relations department, someone from Tesla couldn't possibly be contacting them in an official capacity to provide an authorized statement.

Except Musk has plenty, plenty of reach thanks to his Twitter and other social media accounts, and in most peoples' eyes even though he has been misleading about the whole self-driving thing, hearing a statement directly from the horse's mouth most likely carries more legitimacy in their eyes than hearing it through the media. [0]

[0]: https://news.gallup.com/poll/355526/americans-trust-media-di...

> another is that he's avoiding the spin and dishonest misleading nature of today's media.

Oh please. Elon Musk is one massive BS generator on himself.

> in most peoples' eyes even though he has been misleading about the whole self-driving thing,

Basically, he can lie more effectively this way.

Is "staged" the right description here? I was worried we were looking at a Nikola type event, but that's not what this is.

Just because the route was mapped with lidar, etc. doesn't make this a non-self-driving car. It drove itself under the supplied conditions. It wasn't like they had plotted the route with the stops and starts, etc. From what I understand the system just had prior knowledge of the route. This is in comparison to any other self-driving which MUST have prior understanding of the route, vs Tesla's current product which apparently can make all decisions on the fly.

I think the allegation is that Tesla’s current product cannot actually make all decisions on the fly, but they have been claiming it can—which is exaggeration at the very least.

It doesn't allege anything of the sort – it alleges the 2016 demo used a pre-mapped route and got into a fender bender in the parking lot.

Don’t know if staged is the right word, but it’s a little sketchy at least that the car drove using a method they’ve mocked other companies for using and have stated they have no intention to ever use in their consumer product.

I think they started saying that years later though, after they'd developed it more.

Lidar would increase car cost so much that most people would die of sticker shock. The strongest candidate for proving lidar AV tech is Waymo, which to me means it has a relatively low chance in the consumer space in the 2020s; and unit cost needs to come down dramatically. The cost and complexity starts making more sense in vehicles with very high uptime.

In addition to being prohibitively expensive, lidar units are large, ugly, have moving parts (which should worry any engineer), and increase drag.

Tesla's current Autopilot tech is a bet on the promise of computer vision, and the intuition that if sight is good enough for humans, it's good enough for Autopilot. I think it's a smart bet.

That was true when Tesla wanted to start selling “fsd-ready” cars, but it’s getting cheaper all the time. Luminar has $500 lidar, Huawei says they’ll be selling $100 sensors soon.

I agree. And Tesla isn't alone in this. Waymo was doing the same thing in the early days (no idea if they still are)

Reply on news.ycombinator.com