When comments about Mobileye vs. Tesla are made in threads that are supposed to be about something else, you can quote and reply here to avoid derailing other discussions people are trying to have.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Mobileye vs. Tesla megathread
- Thread starter strangecosmos
- Start date
@strangecosmos I'm interested in you showing me how Tesla "1 billion mile data per year" is exponentially impacting their sensing capability over competitors. Especially with the video that verygreen provided. Which I won't get into details. Nothing shown in that video is anything that the competitor doesn't have. Would love to see this 1 billion miles data quantified, why can't we see the exponential gap in perception area of SDC. If not perception what other areas can it be seen and quantified? You mentioned planning, but again nothing Tesla is doing puts it ahead in planning.
If miles data is this important vital thing, why is 0 companies out of hundreds doing it and yet have equal/better perception system than Tesla?
TLDR - Verygreen video shows us that tesla vision and the diversity of NN it employs is as basic as most companies.
If miles data is this important vital thing, why is 0 companies out of hundreds doing it and yet have equal/better perception system than Tesla?
TLDR - Verygreen video shows us that tesla vision and the diversity of NN it employs is as basic as most companies.
Uncle Paul
Well-Known Member
Some of the other companies use very expensive and ugly equipment to gather their data. Tesla does it mostly with sensors and video.
Having lots of data can smooth out edge occurances and give the system some memory and predictability.
Tesla seems to have the best system right now that can be used most anyplace on the globe where they sell their vehicles.
Having lots of data can smooth out edge occurances and give the system some memory and predictability.
Tesla seems to have the best system right now that can be used most anyplace on the globe where they sell their vehicles.
I was gonna post the below but then i was like the hell with it. Then i convinced myself to just post the incomplete version i have already. The current automakers sucks! Tesla would easily trump every automaker other than (Nissan & gm) in self driving because they are woefully incompetent. If Mobileye didn't exist, it would be even more disastrous It must be frustrating for the likes of mobileye to work with them. Its insufferable to have a LV5 capable chip and have these automakers misuse it. The abysmal UI, its like they're not even trying. Hiring a couple of guys to implement ping pong lane keeping so you can get your 5 star rating is an abomination. Thank goodness for GM and to a lesser extent Nissan.
Obviously there is the silicon valley people ofcourse.
Here are all (some, gave up half way) 2019 cars using eyeq4 and trifocal camera.
2019 BMW 3 Series that will launch in March 2019 has it both and surprisingly works in construction areas too.
https://i.ytimg.com/vi/P_7XGxtwM7g/maxresdefault.jpg
The All-New BMW 3 Series Sedan.
Another thing that the 2019 BMW 3 series has is driver monitoring which i mentioned a-couple months ago.
2:35
The 2019 X5 SUV also has it but i'm not sure if this model has eye4 or trifocal cam. It only uses its drive monitoring for a hands free traffic jam assist under 37 mph. A L2 version of Audi A8's L3 traffic jam pilot.
The system is called Extended Traffic Jam Assistant System (ETJAS) coming out Dec 2018
BMW's Extended Traffic Jam Assistant System wants to stare into your eyes - Roadshow
It looks like BMW x5 SUV that is launching in Nov 10th this year might have trifocal cameras (hence eyeq4) but i have yet to see a definite close enough photo with all three cameras visible. But i'm 80% sure.
https://cdn.motor1.com/images/mgl/mK1ve/s1/2019-bmw-x5.jpg
There looks to be three camera slot there. But i don't know if the other two cams are actually there or if its just a placeholder. IF they are then it begs the question why the product description has less features compared to the 3 series. Is the 3 series hands free like the Supercruise using the driver monitor or is it this gimped ETJA.. Did they actually take care to create a good system? or did they pull acouple guys and sandwich this up in acouple months? Time will tell..
Obviously there is the silicon valley people ofcourse.
Here are all (some, gave up half way) 2019 cars using eyeq4 and trifocal camera.
2019 BMW 3 Series that will launch in March 2019 has it both and surprisingly works in construction areas too.
https://i.ytimg.com/vi/P_7XGxtwM7g/maxresdefault.jpg
The All-New BMW 3 Series Sedan.
Another thing that the 2019 BMW 3 series has is driver monitoring which i mentioned a-couple months ago.
2:35
The 2019 X5 SUV also has it but i'm not sure if this model has eye4 or trifocal cam. It only uses its drive monitoring for a hands free traffic jam assist under 37 mph. A L2 version of Audi A8's L3 traffic jam pilot.
The system is called Extended Traffic Jam Assistant System (ETJAS) coming out Dec 2018
BMW's Extended Traffic Jam Assistant System wants to stare into your eyes - Roadshow
It looks like BMW x5 SUV that is launching in Nov 10th this year might have trifocal cameras (hence eyeq4) but i have yet to see a definite close enough photo with all three cameras visible. But i'm 80% sure.
https://cdn.motor1.com/images/mgl/mK1ve/s1/2019-bmw-x5.jpg
There looks to be three camera slot there. But i don't know if the other two cams are actually there or if its just a placeholder. IF they are then it begs the question why the product description has less features compared to the 3 series. Is the 3 series hands free like the Supercruise using the driver monitor or is it this gimped ETJA.. Did they actually take care to create a good system? or did they pull acouple guys and sandwich this up in acouple months? Time will tell..
Last edited:
My hypothesis is that the millions or billions of miles of data Tesla collects from production cars will provide a significant advantage in self-driving car development. That hypothesis may be true or false. I think we will probably have to wait a while before we get any real empirical evidence on way or the other.
We can’t really quantify the performance of different companies’ perception or path planning neural networks, since key metrics are kept secret. So, we may have to qualitatively assess the rate of progress over time. For example, if Tesla launches superhuman self-driving cars before Waymo, that would be evidence that my hypothesis is true.
As is, we unfortunately can’t make a quantitative or even qualitative comparison between different companies’ capabilities. There is too much of a mish-mash of production systems and non-production systems.
Not a satisfying answer, but a true one I think.
Last edited:
Maybe this is a better spot for this convo than the prestigious neural networks thread
@Joel @Vitold @Bladerskb
Discussion of Mobileye's chips should be redirected out of the neural networks thread (where it is off-topic) to another thread, like this one. Please.
jimmy_d is taking time to answer questions, and off-topic posts make the thread more cluttered and waste time for people checking for on-topic replies.
@pkodali @Bladerskb @chispas
From an academic study I read recently:
"Cynicism refers to a negative appraisal of human nature—a belief that self-interest is the ultimate motive guiding human behavior. We explored laypersons’ beliefs about cynicism and competence and to what extent these beliefs correspond to reality. Four studies showed that laypeople tend to believe in cynical individuals’ cognitive superiority. A further three studies based on the data of about 200,000 individuals from 30 countries debunked these lay beliefs as illusionary by revealing that cynical (vs. less cynical) individuals generally do worse on cognitive ability and academic competency tasks."
@S4WRXTTCS
I want to clarify that I didn't mean that Blader was a cynic, but that he was taking advantage of the tendency for things to fail more often than they go right. Especially when it comes to Tesla because Elon takes an approach that's really pro-failure. He doesn't mind failure, and considers it part of the process.
Everyone who has an AP2 car really has something that is a failure, and that wouldn't have been released by any other car manufacture. A perfect example of this is the new 360 degree visualization. Is it completely ridiculous at times? Yes, but it's still cool. It will get better over time, and we'll probably laugh about how bad it was.
Blader is clearly a smart guy, and I'm positive that is his cognitive ability is above average. It's not his fault he has no boat to ride on.
As to cynicism I think there is disconnect from the kind of cynical personality/diagnoses that you were referring to versus just taking a cynical view of a specific topic that I was talking about.
It doesn't take a full blown cynic to smell something fishy about selling a product that can't do what it's advertised to do.
Last edited:
It depends on the test.
Let's say we polled football fans (the American kind of football) asking them if their team would win the Super Bowl.
The cynic would naturally say no
The optimistic would naturally say yes
The Engineers or scientist like me would run simulations
Out of all of us who is more likely to be right? The Cynic because the odds are in their favor.
Now of course that doesn't carry over to actual cognitive ability, or to academic tasks. The cynic obviously can never truly be successful as they won't give themselves the opportunity. The whole cynic thing is really self defeating in the long run.
Plus a lot of stuff in life is made up, and a cynic can't deal with it. A lot of happiness is giving yourself the ability to enjoy a lie for however long you can.
We're taught to accept lie as an escape from reality.
There is a lot of power in myth. Technology is basically mythology. It's magic really.
There is always going to be the story weaver who is an optimistic. In this book it's Elon selling FSD.
Then there is the one telling everyone it's a myth. This is Blader. Okay, so there is a lot more than Blader, but he's the loudest.
Then there is the rest of us saying "Yes, we know. But let us indulge ourselves in our fantasy for a bit longer would yah?"
MobileEye is a character in this novel who had a falling out with our hero, and is doing their own thing. Maybe they'll come back to win, or maybe they'll just be overlooked and forgotten about.
In the first article you quoted from, I just said that Tesla’s plans were to release full self-driving in 2019. I didn’t say that would necessarily happen on time. I was comparing Tesla’s stated timeline to the stated timelines of other companies.
In the second article you quoted from, I wrote:
“Tesla plans to launch self-driving in 2019 or 2020...“
I believe Elon’s timeline at the time was “about two years”, which I interpreted in that article as 2-3 years. In the same article, I also discussed what might happen in “the event of unexpected delays to Tesla’s development timeline”. So, I stated there was a possibility that Tesla might not develop full self-driving until after 2020. This is quite different from saying definitively that full self-driving will happen in 2019.
Also, “in 2019” doesn’t mean by January 1, 2019. If I say, “The next Canadian federal election is in 2019” that doesn’t mean the election is on January 1. If someone says, “I’m getting married in 2019” that doesn’t mean their wedding is on January 1. “In 2019” means “sometime in 2019”. It can be any time from January 1 to December 31.
So, “in 2019” doesn’t mean “in a little over a month”, and I have never said definitively that full self-driving would happen in 2019.
I’m leaning less toward “bald-faced lie” and more toward “really struggle to understand” because it seems to me you misunderstand the term “in 2019”. I am willing to give you the benefit of the doubt and assume you are fundamentally sincere and well-meaning, but have some trouble communicating and being civil.
The purpose of modelling a scenario like this is to see what could happen given certain assumptions. An input to a model ≠ a prediction. You typically want your assumptions to be in the bounds of reason/plausibility, but you don’t need to commit to the belief that all your assumptions will turn out to be true in order to engage in the exercise of making a model. In my model, the year could have just as easily been 2021, 2022, or 2023 rather than 2020. It’s just an exercise in sizing out the financial opportunity of autonomous ride-hailing.
In my model, I only imagined that full self-driving would happen by the beginning of 2020. I don’t say anything about “after months of seeing their FSD adopted in 2019”; that’s your embellishment. “2019” is not mentioned in the article once.
In a recent article, I look at both a 2019 scenario and a 2025 scenario. In the 2025 scenario, I assign a 50% chance to the outcome that full autonomy won’t happen for Tesla by 2025. Where do I get that probabilities? Nowhere in particular; 50% is just an easy probability to think about. Modelling, to me, is just about imagining different scenarios — stipulating different inputs — and seeing what the output/result is. I’m not trying to predict the future. I don’t know what will happen, and I’ve never claimed to know. I’ve just imagined what could happen.
See this quote from another article written around the same time as the articles you quoted:
“Since self-driving software is still under development, ultimately we don’t know for sure what kind of sensor and computing hardware is necessary. There is irreducible uncertainty. The best that any company can do is deploy hardware that they are confident is sufficient, even if they aren’t absolutely certain. Absolute certainty is impossible.
This is an underlying risk with Tesla’s self-driving development program and with any self-driving development program.”
There is irreducible uncertainty in thinking about future technology. That’s why models are just models and should be taken as such. They are just imagined scenarios. I never made any predictions because my guess is no better than anyone else’s.
What I wrote was an accurate reflection of automakers’ plans at the time of writing. This is what Ford said in January 2017:
“The 2021 vehicle that we will launch in to production will have no steering wheel and no pedals. That’s a different ball game: there will be no opportunity for a human to drive that car.
It’s a ride hailing, ride sharing service that will be owned by a fleet operator. The fleet operator will have them everyday to clean them, fuel them, update software: it’s going to be a very controlled environment, but it will be available for you guys, or me, or whoever to summon it with an app and take me to wherever.”
Ford only started backpedaling on this in August 2017, a few months after a new CEO was appointed.
This is an example of how you jump too quickly conclusions and put someone else down for being wrong, when you yourself are mistaken.
Last edited:
Similar threads
- Article