FSD video completely fake?

[McRat]

The box visualization seems to be a waste CPU cycles.

OK, you are looking for objects by defining outlines, characteristics, and from multiple sensors.

Why wouldn't you just XOR the pixels on the display that triggered your response. This makes the object outline be exactly the opposite shade of the vehicle. So a black car on a black road gives it a white outline. A white car on a concrete road yields a black outline. A white road stripe on a concrete road becomes black outlined stripes around a white center.

What it skips is envelope sizing calculations, and a subroutine for selecting the color, size, and placement of the box and move and size it while still determining borders and features.

I guess it depends on how fast you want your code to run though. The trend today is write bloatware and skip the profiling. Assembler routines and calls? Meh, I gotz ghz, why do I need speed?

 

[3Victoria]

It is possible that the system does produce the marked up visualizations for the engineers testing the system. Whether that would continue to production is questionable since it is not meant for the end user.

The demo was a *demo*. I think that they hacked the planner code onto their DNN, and the planner was a limited version. Obviously if they had a finished general solution they would be releasing. As I wrote, the demo is just a *demonstration*, not the finished product.

 

[BinaryField]

1. I never put too much stock in reddit comments. Everyone there is an expert and an insider.
2. I remember seeing the voluntary reports on autonomous vehicle disengagements for the time period this video was made. There was a massive spike in disengagements at the time, as if they were trying to brute force a successful run. Can anyone pull up the reports?

 

[McRat]

It is possible that the system does produce the marked up visualizations for the engineers testing the system. Whether that would continue to production is questionable since it is not meant for the end user.

The demo was a *demo*. I think that they hacked the planner code onto their DNN, and the planner was a limited version. Obviously if they had a finished general solution they would be releasing. As I wrote, the demo is just a *demonstration*, not the finished product.

I think the problem with the strategy was deceptive techniques used. "Only legal requirements force us to have a driver in the car" or something like that. Not exactly creating a correct picture of it's progress. One would assume it will go anywhere in AV mode if you wanted to get in trouble with the law.

1. I never put too much stock in reddit comments. Everyone there is an expert and an insider.
2. I remember seeing the voluntary reports on autonomous vehicle disengagements for the time period this video was made. There was a massive spike in disengagements at the time, as if they were trying to brute force a successful run. Can anyone pull up the reports?

https://www.dmv.ca.gov/portal/wcm/c...a/Tesla_disengage_report_2016.pdf?MOD=AJPERES

Look at the range of dates for 2016 testing, then when the finished video was released.

 

[AnxietyRanger]

I remember seeing the voluntary reports on autonomous vehicle disengagements for the time period this video was made. There was a massive spike in disengagements at the time, as if they were trying to brute force a successful run. Can anyone pull up the reports?

True. None of that is contradictory to what the Reddit poster said. In fact, it could well support it.

 

[electracity]

As long there is no human driving it in REAL TIME, sitting inside the car or even remotely it is not fake.

I would call it fake if they showing Nvidia software for the demo that they are not using now, which I believe is what has happened.

Tesla would not have need to use the mobileye chip to train AP2 if they were as advanced as the demo video. Obviously tagging objects and the driving path are fairly advanced in the video.

 

[lunitiks]

As I wrote, the demo is just a *demonstration*

What are they 'demonstrating' though... Some other company's software (and possible machineware)? At the risk of this thread becoming the 200.000'th "FSD is a scam"-thread, I do think ppl - like myself - viewed the video as a demonstration of how far they were along. A demonstration of their development - of the state of their own software. "Look, this is what FSD can do today, we just have to generalize, tweak some more and deal with the corner cases."

If what the redditor is saying is correct, I'd much prefer the video being called a "concept" video. I don't think one should take the word 'demonstration' so lightly in this context, that's just my opinion. I bet you disagree

 

[KyleDay]

The Emperor has no clothes. I do love my Model S, but let's all be honest-

Things to remember:

1) There are no indications in current EAP code base that hint any FSD functionally exists today. Some have speculated that FSD must be a different code base, but we have nothing to confirm or deny that. The lack of evidence is not encouraging.

2) There have been no additional FSD testing that anyone has seen or mentioned in over 12 months. (I.e. No CA DMV filings reporting disengagements). How is this not an indicator of lack of progress? If FSD were coming, people would SEE it in testing. For example, I live in Phoenix and I see Waymo and Uber FSD all over the place daily and know they're readying a product very soon. Surely Tesla must be testing FSD somewhere?

3) Since FSD announcement, the AP dev team has turned over significantly, more than once. Lots of reasons why, but after 13 months EAP is still behind AP in several important areas.

4) What I consider the strongest indicator of the expected FSD timeline and actual deliverable features is simply the price being charged for FSD. If EAP is $5/6 grand, and delivers essentially lane-keeping, how in the world does it make financial sense to only charge $3/4 grand for "level 4/5 autonomy"? It just makes no business sense to charge less for far more in usable features. I'll be honest, I think they're charging less because they'll deliver less. Many will disagree and call the FSD price as a "reduced price" or "kickstarter/early adopter" pricing that inevitably will go up once Tesla releases FSD. Perhaps that's true. Regardless, I feel terrible for anyone that leased a Tesla and bought FSD since 10/16 and is getting nothing in return. I almost bought FSD on my car when they released their second video in 11/16 and Elon confirmed that video represented "only 30 days of learning progress". What a sham.


I hope I'm 100% wrong on everything above. I'm bothered that I think I'm right.

 

[KyleDay]

What are they 'demonstrating' though... Some other company's software (and possible machineware)? At the risk of this thread becoming the 200.000'th "FSD is a scam"-thread, I do think ppl - like myself - viewed the video as a demonstration of how far they were along. A demonstration of their development - of the state of their own software. "Look, this is what FSD can do today, we just have to generalize, tweak some more and deal with the corner cases."

If what the redditor is saying is correct, I'd much prefer the video being called a "concept" video. I don't think one should take the word 'demonstration' so lightly in this context, that's just my opinion. I bet you disagree

Agree

 

[electracity]

I do think ppl - like myself - viewed the video as a demonstration of how far they were along.

That was the intention on Tesla's part, of course. Tesla did an equity raise a few months later, didn't they?

 

[myusername]

Obviously it is a preplanned route with a lot of hard coding. Afterall it is a future capability demo.

The visualizations are a representation of what the system infers. Very similar to the color images we see in astronomy of nebulae.

As long there is no human driving it in REAL TIME, sitting inside the car or even remotely it is not fake.

This is wrong. The story was implied to demonstrate the capability of the existing autopilot and it was clearly inferred this way... it was NOT inferred that it was a mock what it could be like. Plugging GPS coordinates into high res mapping is NOT autopilot.

 

[myusername]

Here's what was inferred last year:

Tesla to roll out Autopilot self-driving capabilities 'very soon'
Tesla’s latest Autopilot demo shows us what a Model S ‘sees’ in real time
Tesla Enhanced Autopilot to be released in ‘about 3 weeks’ with incremental monthly releases
This is what a Tesla Model S "sees" in Autopilot mode

Does this sound like what really happened?

"Autonomous driving is nearly upon us, whether we like it or not. The company leading the charge is Tesla, which, despite a few setbacks with the technology, is pressing forward at clip that's almost hard to believe. The general concensus now is that the Model S can go full autonomous with a software update; it's just waiting for infrastructure and laws to catch up. As evidence of this, the company uploaded a video to its Vimeo account showing what the car "sees" through its multiple cameras."

 

[lunitiks]

Here we go

 

[BinaryField]

Here's what was inferred last year:

Tesla to roll out Autopilot self-driving capabilities 'very soon'
Tesla’s latest Autopilot demo shows us what a Model S ‘sees’ in real time
Tesla Enhanced Autopilot to be released in ‘about 3 weeks’ with incremental monthly releases
This is what a Tesla Model S "sees" in Autopilot mode

Does this sound like what really happened?

"Autonomous driving is nearly upon us, whether we like it or not. The company leading the charge is Tesla, which, despite a few setbacks with the technology, is pressing forward at clip that's almost hard to believe. The general concensus now is that the Model S can go full autonomous with a software update; it's just waiting for infrastructure and laws to catch up. As evidence of this, the company uploaded a video to its Vimeo account showing what the car "sees" through its multiple cameras."

Yes, autonomous driving is nearly upon us...in three months maybe, six months definitely.

 

[electracity]

Actually FSD should be here within seven weeks. It's "easy"

Elon Musk Says Tesla Vehicles Will Drive Themselves in Two Years

 

[BinaryField]

2) There have been no additional FSD testing that anyone has seen or mentioned in over 12 months. (I.e. No CA DMV filings reporting disengagements). How is this not an indicator of lack of progress? If FSD were coming, people would SEE it in testing. For example, I live in Phoenix and I see Waymo and Uber FSD all over the place daily and know they're readying a product very soon. Surely Tesla must be testing FSD somewhere?

Just poking around the DMV website a little bit, it looks like disengagement reports and submitted on an annual basis, though I could be wrong. But if that's the case, you'd probably have to judge in January 2018. In any case, I'd assume that if Tesla was testing FSD, you wouldn't be able to tell just by seeing the car pass by, since all of the sensors are embedded into the frame of the car. It's not like a Waymo vehicle that has a huge dome sticking out of its roof.

I wonder how all of this stacks up with Elon's claim that there will be a coast to coast FSD demo around the end of this year / beginning of next year. (Not that anybody here really puts a whole lot of stock in his time estimates for deploying future technology.)

 

[harry]

1. I never put too much stock in reddit comments. Everyone there is an expert and an insider...

Does anyone else find it amusing that an entire thread was triggered by a reddit comment from the undoubtedly knowledgeable and sophisticated "PM_YOUR_NIPS..."? Really?

 

[McRat]

There is 2017 data, but be aware Google (and others?) moved much of their public road testing to a suburb of Phoenix. It's a fairly nice place for entry level testing. Modern tract homes with good roads. But AV has to work in major cities centers. This is where the initial profits will be. Taxi service in suburbs isn't common. City centers and airports are where taxis get most their money from.

Something isn't right about 2016/2015 though. GM/Cruise was operating both years, first with 5 Leafs, then with 20 Bolts. There should have been more incidents. Cruise started testing on public roads in June 2015, and used different 25 cars prior to Jan 1 2017, and were testing each month ever since. Perhaps incidents were not required to be filed?

But 2017 shows where Google/Waymo reduced participation in California, about April. While there was a single Waymo event in August, it was a situation the AV car was at fault. Debris in the road caused a disengagement as the driver tried to avoid a collision unsuccessfully. Granted there is not always a collision free option with road debris.

In 2016 there were 11 company filings for AV testing in Calif.
In 2017 it jumped to a whopping 43 filings for companies wishing to test in Calif as of Oct 2017.

For more complete info, start here: Testing of Autonomous Vehicles and try to find the right links for various data.

It makes me wonder if companies want to avoid California's very public data releases. This seems unethical if California wants to become a hub of AV testing. By listing AV test performance, it can exaggerate the problems with AV systems. Nearly all collisions have been AV-Not-At-Fault, but they are listed the same. It does not discern between difficult testing and simple testing. Going through the city center of SF is worth the same as a 4 lane deserted road.

But the most significant point of this whole essay is that a huge amount of competitors are working on public road AV testing now, at least in California. Cruise is also operating in Arizona and Michigan and recently started in New York City (which has perhaps the most draconian restrictions, including mandatory police escort).

Who will win this contest? That will not be known for years. It is not rare that the first company to develop a technology ends up losing in the marketplace. However, I think several companies will see various degrees of financial success. Tesla is not going to make ICE cars, and Google/Waymo currently are far behind is automotive production.

My bets on the 3 biggest winners? Google, GM, Tesla. Google will save money on their Google Earth costs, and be able to license their system to other makers, as well as run taxi service. GM has a wide range of vehicles suitable for many tasks up to delivery trucks and route buses and down to subcompacts that could be fitted with Cruise. Tesla will be able to sell their technology to Tesla buyers as well as run Tesla taxis.

 

[KyleDay]

Does anyone else find it amusing that an entire thread was triggered by a reddit comment from the undoubtedly knowledgeable and sophisticated "PM_YOUR_NIPS..."? Really?

This guy actually works for Tesla on their AP team, or so he claims. Read his other reddit posts. Their pretty insightful; such as the effectiveness of front-facing sonic sensors having trouble with wind noise, etc. He's got several really informative posts.

 

[AnxietyRanger]

Does anyone else find it amusing that an entire thread was triggered by a reddit comment from the undoubtedly knowledgeable and sophisticated "PM_YOUR_NIPS..."? Really?

You do realize that "NIPS papers" refers to scientific papers at a well-known machine learning conference?