HW2.5 capabilities

[JeffK]

We all are hungering for information as owners/fans - but if ran Tesla I would also keep my mouth shut and fire anybody who didn't. Dollars drive everything Jeff.

• Nissan and other automakers which do not actually have a product on the market to purchase now: They aren't any more forthcoming than Tesla is when it does not serve their interests. Auto makers which do not have a product to sell in a world in which Tesla is threatening the market share of every single other automaker - have a strong incentive to show off videos of their R&D to the public. Why is this? It's to reassure their shareholders that the C-Suite set is not sitting on their collective hineys at HQ while the NPV of their business drops through the toilet. I'm convinced this is why Audi is slinging a ludicrously crippled "Level 3" A8 this year - it's just a marketing stunt to also show the world that Audi isn't being left in the dust.
• SpaceX: SpaceX has a product to sell - space launches - and it has 16 successful stuck landings in a row right now. Why not share a great video of the blood sweat, tears and failure that went into developing this amazing success - now that the failures are in the past. If SpaceX was taking $$$DEPOSITS$$$ for rocket launches while the rockets were still in development would @JeffK's marketing consulting firm advise the CEO to start posting up videos of his rockets exploding on the launch pad? I don't think so.

Tesla is doing the right thing, business-wise, by keeping us in the dark. They've made their promise - they're selling hardware - they're updating software - that's enough. The last thing Tesla's owners need is videos of testing fails going viral and candid interviews with engineers going "gosh golly - this here is a hard problem. We're workin' on it..."

There's still a huge difference between FSD and the current autopilot implementation. By keeping silent, they are just causing doubt and potentially losing out on what I will call "FSD deposits." How many people, according to the polls, are not going to purchase the FSD upgrade right away... there's a reason for that.

If Tesla owners were to see videos of progress then they might feel more reassured instead of more in the dark.

 

[Bladerskb]

We all are hungering for information as owners/fans - but if ran Tesla I would also keep my mouth shut and fire anybody who didn't. Dollars drive everything Jeff.

• Nissan and other automakers which do not actually have a product on the market to purchase now: They aren't any more forthcoming than Tesla is when it does not serve their interests. Auto makers which do not have a product to sell in a world in which Tesla is threatening the market share of every single other automaker - have a strong incentive to show off videos of their R&D to the public. Why is this? It's to reassure their shareholders that the C-Suite set is not sitting on their collective hineys at HQ while the NPV of their business drops through the toilet. I'm convinced this is why Audi is slinging a ludicrously crippled "Level 3" A8 this year - it's just a marketing stunt to also show the world that Audi isn't being left in the dust.
• SpaceX: SpaceX has a product to sell - space launches - and it has 16 successful stuck landings in a row right now. Why not share a great video of the blood sweat, tears and failure that went into developing this amazing success - now that the failures are in the past. If SpaceX was taking $$$DEPOSITS$$$ for rocket launches while the rockets were still in development would @JeffK's marketing consulting firm advise the CEO to start posting up videos of his rockets exploding on the launch pad? I don't think so.

Tesla is doing the right thing, business-wise, by keeping us in the dark. They've made their claims that the current cars are headed toward "full autonomy pending regulatory approval" - they're selling hardware - they're updating software - that's enough. The last thing Tesla's stock owners need is videos of testing fails going viral and candid interviews with engineers going "gosh golly - this here is a hard problem. We're workin' on it..." while the Model 3 is ramping up and the entire world is focused on whether Musk will fail or succeed to bring "full self driving" to fruition.

I'm lost so what product does Tesla actually have on the market that Nissan and other automakers doesn't have?

@JeffK - furthermore, Tesla's approach to developing full autonomy may in fact be a real world case of confidence producing future reality. If the massive fleet learning, video-uploading approach turns out to be the only way to build a reliable FSD fleet - then Tesla has a huge head start and the more cars they get on the road (ie the more they SELL) the higher the odds of the R&D effort succeeding. So - keeping up public confidence in the success of the project will help keep sales flowing - which may help bring the claim to reality. If you believe that is true - then again, Tesla is better off staying silent.

Secondly you do realize the reason Tesla is beaming up videos is mainly to work on sensing (object recognition..etc) which Mobileye had already solved in 2015 and no longer include sensing on their presentation or gives very minimal time to it. This is why Tesla AP2 still can't recognize signs. Its not a sign of being ahead its a sign of being behind...

Tesla literally told you why they are collect the short clips.

What really counts, hd map building, drive-able paths and driving policies. Mobileye is already doing with a much larger fleet.

 

[calisnow]

I'm lost

We know.

 

[calisnow]

There's still a huge difference between FSD and the current autopilot implementation.

I agree.

By keeping silent, they are just causing doubt

Maybe in you, me and the rest of the nut-brigade here at TMC - an anxiety-ridden group of internet keyboard jockeys with an unhealthy obsession with Tesla.

and potentially losing out on what I will call "FSD deposits."

Perhaps, perhaps not. No data to guide this either way.

How many people, according to the polls, are not going to purchase the FSD upgrade right away... there's a reason for that.

There are lots of reasons for many things.

If Tesla owners were to see videos of progress then they might feel more reassured instead of more in the dark.

Again - you're describing Tesla owners at large as though they need psychiatric care for an anxiety disorder. You may be correctly describing the incredibly weird group of people we are here in our little nerd-closet but we are not the norm. Furthermore Tesla has no reason to "reassure" us - we already bought the damn car lol.

I don't think the TMC polls are relevant nor do I think you or I are relevant. I have 6 other friends with Model 3 deposits who waited in line like I did. Not a single one of them has heard of Tesla Motors Club or cares. They do not spend their time thinking about which automaker is ahead, how close Tesla is, etc. They want a Model 3. They saw the FSD video last October. They lust after Tesla. It ends there. They do many things with their spare time - typing away on the internet about Tesla's PR strategy is not one of them.

 

[buttershrimp]

I keep tweeting to Elon, Andrej and previously Sterling Anderson before he left. I'd love if they gave us some measure of progress. With other companies we are seeing either data, more demos, researchers are giving interviews showing real data, etc. Right now all we have from Tesla is a set of two sped up videos from last year, which were trained using hundreds of disengagements reported to the DMV in likely the same path. In one of the videos we even see the car stop after making a turn, for no reason, which could have been pretty dangerous as it could have been rear ended.

I love Elon, and I'm a big Tesla fan. I just wish they were a little more forthcoming with where they are, what the challenges are, etc. I don't need to necessarily hear from someone who is optimistic. I just want the brutal truth as to where we are with the state of the technology. The same applies to everyone else working on this too, not just Tesla.

Elon posted a video of some awesome SpaceX rocket landing failures. Granted, SpaceX isn't a publically traded company, still, I like that they owned up those failures and those failures served as a learning experience. Show me some FSD fails. Show me progress.

Nissan has had a self driving car in the works for the last four or five years now and they often show it off. Over the years I've seen videos when it was downright scary and how their tech has improved over time.

I feel very similarly on wishing we'd get more transparency, but I think @calisnow is right, it wouldn't make sense to keep talking if they are on to something. I think (and hope) we will see a candid update on the status of autopilot with the semi launch. I believe Tesla is going to make safety and self-driving the corner stone of the semi truck... it's what everyone wants to see.... I can't imagine them not talking about the particulars with respect to scaling up the autopilot system for the semi.

 

[buttershrimp]

Maybe in you, me and the rest of the nut-brigade here at TMC - an anxiety-ridden group of internet keyboard jockeys with an unhealthy obsession with Tesla.

Oh man, this is so true. Tesla has become my modern day equivalent of the the sore thumbs I used to get after a day of playing tetris 30 years ago. Instead of the tetris music... I have damn autopilot engagement noises in my head at night.

 

[calisnow]

@JeffK - the other thing is - what is there to show? Tesla already showed a video of the end-goal - the FSD vids last October. Anything they show now on "progress" will look like a step backward to the average Joe and the media.

Put yourself in the shoes of the CEO. What is your goal? Money. How do you make money? Sell cars profitably. When a customer is faced with a choice of what car to buy there are decision points. Tesla has already painted the FSD vision in the brains of consumers. Tesla has a sterling reputation to the general public (not the nerds on this forum) and generates awe and lust. Also - they sell a lot of cars already and don't seem to have a serious problem finding more customers right now - their biggest problem is building enough Model 3's to satisfy demand at the moment. They have no problem attracting capital when they need it.

Where is the upside to launching a "FSD progress" blog/campaign/video series? It accomplishes nothing, uses up time and money, is of interest mainly to nerds (not the larger customer base) and has potential down-side risk of losing control of the story.

I am a business owner. Without going into detail on what I do, right now I have a supply problem at a business (a great problem to have). In other words I have too many people wanting to buy what I'm selling and I have existing customers to service. Am I working on improving my product quality, customer service quality and increasing my available supply of product to sell? Of course! Always! I'm no different from any other rational business owner. Do I have a very small percentage of clients who like @JeffK - are interested in my business itself - not just in consuming what I sell them? YES. I do. Do you know what I think of these folks? I'd be happy if they went and jumped in the nearest lake and I never heard about them again. Why? They consume more time but do not pay more money - and for each one of them I have 10 wannabe customers to take their place - and the vast majority of my clients I never hear from. In other words - the "loyal fans" type clients are the least profitable and biggest pains in my ass. And they are entirely disposable - I don't need their viral organic support.

I think, unfortunately, that is what we are to Tesla.

 

[S4WRXTTCS]

The problem I have with this thread is the sheer amount of noise to signal.

It's supposed to be about the capabilities of HW 2.5 strictly from a hardware/software standpoint, but there is very little of that. It's so filled with arguments surrounding L5 that it's impossible to see how 2.5 differentiates unless you limit a search to only verygreen posts.

For a lot of us including myself I'm not concerned about L4/FSD due to how far off it is. Sure it's important to give credit to Waymo for a taxi type car in Arizona, or Audi for a very limited consumer L3 system,

But, the truth is that none of us know what it's going to take for self-driving to be a reality. A reality where millions of people interact with them in a wide variety of situations, and road types. Where the autocars have to deal with bullies on the road, and all the crap that happens on a daily basis. Where you can go from how humans interact in seattle (passive aggressively) to how they interact in Florida (like they took some bath salts).

One side feels as if the sensors are THE issue, but I would argue there are deeper issues. Issues that simply adding radar/lidar/etc won't solve. The other side feels as if its just some magical software that AP 2.5 can run.

I tend to be someone who is looking more at the short term capabilities of HW 2.5 where FSD is a distraction for us owners, and a massive liability for Tesla.

When I look at the automotive market there isn't a whole lot of amazing L2 or limited L3 stuff. So that makes EAP pretty darn attractive if it can do what's promised.

I think if he we could make the thread less about FSD, and more about EAP it would be a bit more productive. It would also be nice if people could make it more technology centric and less brand centric. So I wouldn't automatically know where everyone stood simply by their Tesla Loyalty Score.

 

[calisnow]

Oh man, this is so true. Tesla has become my modern day equivalent of the the sore thumbs I used to get after a day of playing tetris 30 years ago. Instead of the tetris music... I have damn autopilot engagement noises in my head at night.

I can fix that for ya:

 

[calisnow]

The problem I have with this thread is the sheer amount of noise to signal. It's supposed to be about the capabilities of...

The other side feels as if its just some magical software that AP 2.5 can run.

Don't lump me in that category. Yes it is a software problem - but a really really hard one. The people who believe more sensors are needed, or V2V communications, or embedded signals in surrounding infrastructure - they also would acknowledge that it's a software problem. But their claim is that FSD using cameras alone (like humans do) is such a difficult software problem that we will be waiting a very very long time if we do not get busy and use additional sensors which will, they believe, reduce the amount of "intelligence" needed to solve the problem.

So while I hope this can be done with cameras and software I acknowledge it's perhaps the hardest problem ever attempted.

I think if he we could make the thread less about FSD, and more about EAP it would be a bit more productive.

Thanks for sharing - I'll schedule a vote and send it to committee.

It would also be nice if people could make it more technology centric and less brand centric. So I wouldn't automatically know where everyone stood simply by their Tesla Loyalty Score.

Good luck. But seriously - I attempted this kind of thing in a thread a while back where i asked for expert opinions on whether software and cameras are sufficient given the state of today's AI. There was some good discussion there and some good private opinions sent my way.

 

[calisnow]

@S4WRXTTCS - as for the EAP vs FSD discussion - does it really matter? EAP is just a sub-set of FSD with the disclaimer that you have to pay attention. It is still "self driving" in that the car will supposedly make decisions on lane choice - maneuver through traffic lanes to change freeways, find exits, etc.

 

[AnxietyRanger]

They want a Model 3. They saw the FSD video last October. They lust after Tesla. It ends there.

Ignorance IS a bliss!

 

[buttershrimp]

I can fix that for ya:

Whooooo dang son... that's some straight up-down-up-down-b-a select start skill!

 

[buttershrimp]

Ignorance IS a bliss!

If ignorance is bliss, then Audi fans are busy snorting ecstasy in the bathroom, before going back out to party with Mobileye and GM Supercruise at the 2018 Blissed out Laser Light Show with Pink Floyd.

Meanwhile, the rest of us Tesla owners are driving AP-driving responsibly each day, contributing to fleet data, living the dream... saying nasty things about tesla because we are bored.... and are due for some AP love soon

 

[AnxietyRanger]

If ignorance is bliss, then Audi fans are busy snorting ecstasy in the bathroom, before going back out to party with Mobileye and GM Supercruise at the 2018 Blissed out Laser Light Show with Pink Floyd.

Meanwhile, the rest of us Tesla owners are driving AP-driving responsibly each day, contributing to fleet data, living the dream... saying nasty things about tesla because we are bored.... and are due for some AP love soon

Be that as it may, even you have to admit someone basing their Tesla desire on the FSD video from 2016 still today is ignorant.

 

[AnxietyRanger]

But their claim is that FSD using cameras alone (like humans do) is such a difficult software problem that we will be waiting a very very long time if we do not get busy and use additional sensors which will, they believe, reduce the amount of "intelligence" needed to solve the problem.

So while I hope this can be done with cameras and software I acknowledge it's perhaps the hardest problem ever attempted.

That's IMO only a part of the claim, though.

There are at least two other lines of thought that doubt Tesla's approach:

1) Lack of redundancy and camera cleaning in the camera only approach. Cameras are susceptible to weather conditions that radar and lidar can help with (darkness, fog, radar seeing through obstacles). Without stereo vision estimating distance is also harder, though not impossible of course. Tesla only has redundancy and stereo vision for a very narrow strip up front whereas the competition is aiming at 360 radar and 360 lidar in addition to 360 vision...

2) Questions about Tesla's ability to solve the driving policy questions as fast as competition that has been working on this for much longer. Even if Tesla solves recognition reliably, they do not have Google's or MobilEye's long history on solving the actual driving.

Maybe Tesla has something revolutionary cooking in the labs. As an FSD owner I want to believe, I have vested interest... But it takes more faith than knowledge to believe in that at this stage.

 

[buttershrimp]

Be that as it may, even you have to admit someone basing their Tesla desire on the FSD video from 2016 still today is ignorant.

I don't think the video is a fake.... I think they are having a hard time integrating the systems to work everywhere.... its easier show a general software that gives you an idea on how they intend to do this.... quite another thing to robustly validate.

The video is still on the Tesla site.... and that speaks volumes.... we are getting tired of waiting but some impressive breakthroughs shall be arriving in 3....2....1

 

[buttershrimp]

@AnxietyRanger , I think you are going to happy soon... I can tell you are wanting to be impressed already. I think semi event will really be helpful in knowing where things are at.

 

[AnxietyRanger]

I don't think the video is a fake.... I think they are having a hard time integrating the systems to work everywhere.... its easier show a general software that gives you an idea on how they intend to do this.... quite another thing to robustly validate.

The video is still on the Tesla site.... and that speaks volumes.... we are getting tired of waiting but some impressive breakthroughs shall be arriving in 3....2....1

I don't think the video is fake, either. But I do not think it is of much value either. Basing anything just on that is ignorant IMO.

With Google/Waymo, MobilEye and e.g. Audi we at least have a decade or so of their work and progress in the autonomous space in the public. Sure, it is not in consumer products, but in labs or prototypes. However with Tesla, all we have is one FSD video, some vague information about Tesla's work and progress in the hardware space (chips etc.) and what we know of AP1, loss of MobilEye partnership and the status of current EAP. Including some spyshots from 2015-2016, we can see some other hints of what Tesla has been testing earlier on (AP 1.5, the early self-driving prototype hardware etc.), but overall we are looking at a company that has been at it for a relatively short while, whose AP team is has been a mess and has had a few reboots already. It is not exactly instilling confidence.

With others we have actual evidence of their progress over a number of years. There are believable roadmaps, there is history, there is evidence of other things than just recognition etc. being seriously looked at. There is talking to legislators - they are that far.

If Tesla is already doing the same things, then we lack the data that they are. Hence it is more a question of faith at this stage.

@AnxietyRanger , I think you are going to happy soon... I can tell you are wanting to be impressed already. I think semi event will really be helpful in knowing where things are at.

Of course I want to be impressed. That's why I bought AP2. But my happiness is rather irrelevant for a thread like this. In here I am far more interested in finding out actual status of AP2 (or 2.5), than being fulfilled in my car. The quest for knowledge, in the end, makes me more happy than any car.

 

[stopcrazypp]

Lol you're joking right?

So if Tesla releases Level 3/4/5 software complete update in say 20 years. They will be the first SDC car? What kind of warp logic is that?

Not sure what you mean by "complete update", but for example if the first non-Tesla to release a level 4/5 car releases it in 3-4 years, and Tesla releases their first L4 software ("complete" or otherwise) in 4 years, I'm not sure if history will have that other car be counted as first rather than Tesla. Obviously 20 years is different case (in that case most of the cars would be end of life already). But these are all hypothetical, so far Tesla's the only one with hardware (claimed or otherwise).

In that case, audi isnt the first L3 and first sdc car and all the hundreds of news articles l, industry leaders, companies, were wrong in referring to it as that.

Since some guy strapped some cameras and a gpu in a car and called their car l5 capable.

Well level 3 wouldn't qualify for FSD, so isn't that relevant to the conversation. But I don't believe Audi sold any cars yet equipped with the L3 option, so we still have to wait for that to happen first, before they can claim first on that.

You proved my point yet again as i quoted.

"One of the most amazing things about Tesla is how Tesla fans manage to hold thoughts which are so patently false, against all types of evidence."

I respect @JeffK because instead of coming up with some warped excuse. He is looking forward to actual statistics that can show Tesla progress not some fable marri land idea that tesla is ahead.

Again, your point is not proven until a manufacturer actually does release and sell such a car. You post a lot about development, and that is appreciated, but the point is not made until the vehicle is actually sold to a consumer. Last minute delays does happen (for example Supercruise was announced in 2012, was poised to release 2016, but delayed until this year and still not out yet).

Theres no difference. Like i have said there will be companies like GM and Google who will have thousands of cars hailing Thousands of REAL customers around dozens of cities in 2018/2019. Starting with google in phoenix in 2017 and SF and other cities by GM in 2018

Its up to people like you if you want to live in the land of myth or acknowledge reality.

No difference? I have one of those cars on reservation right now. I can't buy a Waymo Pacifica or a Cruise Bolt. Those will be public transit (no different me than a bus or a train, which I can't own either). There have been many vehicles already with tons of expensive sensors that have worked for many years already as FSD. Personally, I think it's significant if the vehicle is affordable vs something that is out of reach for most people to own (not even by stretching).

Anyways, I don't want keep hijacking this thread (which should be about HW2.5), so for any further responses, I'll be responding on a stub thread I just created:
Autonomous Car Progress