Hacker News Reader

Comment by semi-extrinsic | Hacker News Reader

I'm not sure they've done things anyone considered impossible. People have said these things will be wildly unprofitable, sure, and so far those people are mostly right.

But nobody has said it would be impossible to build a big expensive electric sedan with high range. Nor did anyone say it would be impossible to build a wide coverage fastcharge network. Surely they are impressive feats, but not "thought to be impossible".

In fact, other companies have managed to build similar things. Today the Jaguar and Audi e-SUVs are competing with the Model X, in a few months the Porsche Taycan is competing with Model S, and the ChaDeMo charger network is the same size as the Supercharger one (both have 12,000 units deployed globally).

View on HN

First US car company in 60 years, electric at that. Best crash test ratings. Highest owner satisfaction. First US-owned factory in China. Built in about 6 months. Ability and willingness to use OTA updates. Willingness and ability to put auto-pilot on the road for consumers. Ridiculous performance. Better chip than Nvidia. Top 2 auto battery supplier. Best selling lux car in US (ICE or EV). Dramatically better battery efficiency. No dealerships. Online ordering. Plausible FSD story. Camera-based autonomy. $35k price point.

> Plausible FSD story.

No.

> Camera-based autonomy.

That's among the reasons the former isn't true.

> $35k price point

That PR point has largely been abandoned. I mean, nominally they’ll still sell it to you if you jump through hoops, but they’ve deliberately made it less accessible because, as much as it was an important PR point, it was never really viable.

I've been on the fence about Tesla's approach (no Lidar) being the right way. But did you see the recent presentation at all?

At this point I think it's clear that they have a high chance of succeeding.

And I think the core argument is correct: If you don't have a neural net that can estimate depth from cameras as well as a human, you don't have a viable FSD system. The classifications that the system has to do to actually implement all the requirements of true FSD, is way harder than what's needed to make Lidar redundant with cameras and radar. Lidar is a shortcut to get a proof-of-concept running when your vision system is too weak to actually do full FSD.

> as much as it was an important PR point, it was never really viable.

Of course it's viable. It's just a question of time. It's pretty silly to think that sommehow, right this year, batteries will stop falling in price and Tesla will find no other ways to continue to reduce costs.

> And I think the core argument is correct: If you don't have a neural net that can estimate depth from cameras as well as a human, you don't have a viable FSD system.

I agree that this statement is probably correct. But they're assuming the positive outcome here, and we have no proof that's correct.

In fact, it may well turn out that the current approaches to autonomous driving using neural networks won't ever yield a viable FSD system. In fact I'd take that side of the bet, to be honest.

Continuing to deny something that exists is actually beyond calling it impossible!

If you've got some spare time, go read some third party research on how cameras might actually be better. Lidar superiority is a myth. https://arxiv.org/pdf/1812.07179.pdf

Dude, if you think a CS conference paper is going to be the secret sauce that extracts Tesla from the $5b hole it’s dig itself into, you a) know nothing about business and b) know nothing about CS.

A little decorum, please.

Just an external, corroborating data point. That’s all.

That paper says nothing of the sort. Did you even read it ?

The numbers actually show pseudo-LiDAR being worse in general and significantly worse at longer distances.

I read it. Did you?

"it is plausible that image-based 3D object detection for autonomous vehicle will become a reality in the near future."

You do know that we have 3D object detection today. In fact we've had it for a few years now.

The issue isn't whether it works or not. The issue is whether you need LiDAR in the mix as well. And your paper says unequivocally that the answer is "yes we do".

This comment is the very definition of head in the sand lol.

Apart from the FSD bit, which is still a far way from realized and might never be, none of those things are in the category "people said this was impossible".

Read the list again.

wow.

here we go: > Best crash test ratings.

partly false. there are 30 cars in the same category. kias, hyundays you name it.

> Highest owner satisfaction.

partly false. only the model 3 got it. model s was dropped due to how crappy it is.

> First US-owned factory in China. Built in about 6 months.

false. it’s not built. and there are so many other corps out there that own china factories it’s not even funny.

> Willingness and ability to put auto-pilot on the road for consumers.

how many people died because of this?

> Better chip than Nvidia.

proven to be false.

> Top 2 auto battery supplier.

false. that would be Panasonic.

> Dramatically better battery efficiency.

false. same efficiency as a kia.

> No dealerships.

this is very bad.

> Plausible FSD story.

false. debunked.

> $35k price point.

false. they don’t sell $35k cars.

I could go down your list and write false as well. To take the first one as an example of how you are wrong.

https://www.regulations.gov/document?D=NHTSA-2017-0037-0037

Model 3 not only the very best results, but by a wider margin.

And the last. Tesla absolutely sells a $35k Model 3: https://www.tesla.com/blog/update-our-vehicle-lineup

> Better chip than Nvidia.

That's arguably false. Nvidia has a more powerful FSD chip. But performance/watt might be better for Teslas chip.

>I'm not sure they've done things anyone considered impossible.

Is this also a joke? They've been doing impossible stuff for literally 10 years, there's plenty examples of people claiming so.

Consider this, there isn't a better car than the one they produced in 2012, and GM's best effort at an EV (the Bolt) lost somewhere between 4-8k per unit.

Reply on news.ycombinator.com