Elon: FSD Beta tweets

[Knightshade]

Can we please stop this nonsense about how lidar and lots of sensors will make cars too cost prohibitive. This myth needs to die

Sure, just as soon as it's proven a myth somewhere other than a "coming in the future" press release.

Show me a general works-everywhere robotaxi I can buy for less than 6 figures and you'll have something.

Currently this does not exist from anyone- including Tesla- though it's a stated future goal just like everyone else.

Yes, usually things are visible first and put in memory, with an expected trajectory. Something FSD will need.

Something it already has.

This is why the car will often slow down or stop when it sees a pedestian or a bike nearing a crosswalk-- because it expects it to enter that crosswalk in the path of the car.

in fact the 10.13 release notes specifically call out making an improvement to this already existing ability to reduce false positives (ie the current system is TOO cautious and overly likely to think the pedestrian or bike will enter your path)

Reduced false slowdowns around crosswalks by better classification of pedestrians and bicyclists as not intending to interact with ego

The situation here was a toddler you can not see stepping out from behind a parked truck, such that the car still moving at speed only sees it when it's a couple of feet away.

The car hits the kid in that situation, no matter how much you spent on sensors, or if a human is manually driving....because physics. Even with a reaction time of 0, hitting the brakes full on once the kid is visible still won't stop the car in time.

If the kid IS visible, at all, far enough away/early enough for the car to not hit him, then FSD will beat the human at avoiding an accident, because it has a quicker reaction time AND no chance it might be "looking another direction" at the moment the kid enters visibility range.



Remember- this entire tangent was from a remark about the "blind spots" of the car.... which only exist in a TINY area within a foot or two of the vehicle, and only below the belt line, and only in the very front of the car.... if something first appears that close, while you are moving at speed, you will hit it- no matter who or what is driving- because physics. If it appears earlier, then the blind spots don't matter- the car can see it with existing cameras and can react to its expected path- something it already does today

 

[daktari]

“I remain confident that we will have the basic functionality for level five autonomy complete this year,” he said. “I think there are no fundamental challenges remaining for level five autonomy.” (c)ElonMusk

And these same dummies KEEP falling for it. Every year:
"he SAID this year, but he didnt really mean THIS year. Isnt that obvious?"

But you are supposed to KNOW Musk, because only then it is obvious that he is lying or making it up. He is entitled to that, and shame on you for not knowing his ways. ;-)

 

[Bladerskb]

Can we please stop this nonsense about how lidar and lots of sensors make cars too cost prohibitive. This myth needs to die. Baidu just announced their next gen robotaxi with 38 sensors, including 8 lidar, that will only cost $37k. So it is absolutely possible to put lidar and lots of sensors on a car for autonomous driving that is affordable. Heck, even the Waymo I-Pace with its 50+ sensors costs half of $250k. Putting lidar and lots of sensors do not make the car cost $250k.

Baidu reveals the next-gen robotaxi that will be deployed throughout China | TechCrunch

Baidu, the Chinese search engine giant that has plowed money into AI and autonomous vehicle technology, unveiled Wednesday a new all-electric robotaxi

techcrunch.com

He won't stop. He has ZERO original thought than to regurgitate whatever Elon says. Whenever someone respond with fact he just deflects. Notice how i exposed him on the $250k statements and he goes crickets. Then the other statement that "humans don't use lidar and the road system is built for vision" Yet he cheers for superman who uses x-ray vision in the world built for normal vision... huh? Again like I said its a exercise in time wasting.

He came at me in the TSLA investor thread and I exposed him. He quoted old posts of mine and took every statement out of context. Turns out in actuality everything i said was right.

The post he made calling me out:

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

How do you kill a competitor with a production of 500 per year? How do you know GM doesn't have a secret weapon? I have a suspicion they are getting ready to blow away the automotive press and win back the hearts and minds of the motoring public with the stunning announcement that they plan to...

teslamotorsclub.com

I respond and expose him with indisputable facts:

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

Rolled 0715-c$700 to 0729-$700 for a $30 credit in a Beneficiary IRA. Kind of risky, but not expecting much the next 2 weeks beyond an unimpressive 2Q ER. Although given recent history, we could have a plodding ER and the stock could still take off (low correlation of SP to fundamentals)...

teslamotorsclub.com

He responds and completely deflects:

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

AAPL as market mover had a bullish brake out. QQQ broke the downtrend an consolidates. Every switch from bear to bull market started with tech leading the way. Tech earnings start as soon as next week. Could be the catalyst the tech market waited for - especially when guidance is good. That will...

teslamotorsclub.com

That is his routine:
He makes ridiculously blatantly incorrect statements in a post.
Gets exposed with indisputable facts and then he just responds with deflection that makes ZERO sense.

I bring this up because sensor coverage is what we are talking about.
To break it down for others...

1) I posted in 2016 that HW2 with half variant of Nvidia Drive PX2 (10 Tops) will have difficulty getting regulatory approval for L5. I was RIGHT (Its not even possible anymore as cars are not being made with HW2 nor is Tesla trying to).
2) I posted in 2016 that EAP Navigate on Autopilot will require 8 cameras, more than the 4 Tesla initially allotted. I was RIGHT.
3) I posted in 2016 that EAP Smart Summon will require 8 cameras, more than the 4 Tesla initially allotted. I was RIGHT.
4) I posted in 2016 that EAP will require maps (Tesla ended up using their ADAS 2.0 Map Tiles, OpenStreetMaps and geofenced out areas that had problems.). I was RIGHT.
5) I posted in 2016 that EAP NAP was similar in function to L4 highway system that others were trying to deploy but the difference is that L4 will not require human supervision and hence be better. (I even told someone in that same thread that "I hate to break it to you but hands on ready to take over will still exist and actually more than ever. You will need to be even more vigilant. Why? Because EAP is actually still L2 automation") I was RIGHT.
6) I posted in 2017 that Cruise was "one of the front-runners in self driving tech." (They ended up being the second to launch a robo-taxi in the US) I was RIGHT.

Reminder that Everything I posted back then I had MAJOR pushback from, by almost everyone.
Now its common to find Tesla fans today expulsing these same line of thinking.

 

[daktari]

Why are y’all focusing on anything lower than level 5? Perhaps many of you have not seen this, starting around the statements at 1:19 and again, at 2:30. His goal is clear and his timeline is clear

A dire need for historical revisionism.

 

[2101Guy]

Naah- you just conflated different statements. Because you're incapable of honest debate.

Dude. It’s not EVEN a debate. As you defined “facts”, it’s “facts” that Elon Musk has REPEATEDLY stated that Tesla FSD WILL be capable of L5 autonomous driving before 2023. He’s stated it in writing and in videotaped words. He has not wavered. His statements this year are not contradicting his prior statements in the video (with exception of specific date) but in fact, SUPPORTING his prior FSD L5 commitment. His statements all clearly define that with a deadline of prior to 2023, FSD WILL reach L5 autonomy.

Did you not comprehend his messages clearly voiced in his video?

Is he paying YOU the $12k for that vaporware?

 

[AlanSubie4Life]

Something it already has.

Was talking about toddler obscured by car that humans can see and identify which car cannot see. Fairly sure FSD does not have this capability to analyze 95% obscured objects and quickly and accurately identify them.

I understand that originally you made it very clear it was a situation where the toddler was not visible at all until 1-2 feet before (a different, less common situation).

 

[jabloomf1230]

Tesla has opted for simplicity. LIDAR and radar are active systems. They transmit a beam and then a sensor measures the return signal. This is more complicated than a vision-based system which passively senses the condition of the environment. Tesla's philosophy even extends to its auto wipers. The most common system in other vehicles transmits an IR beam outward though the windshield and a sensor measures scattering of that beam. Tesla has opted to use the windshield camera to compare wiped vs. non-wiped frames.

Cost is not the only factor in choosing passive versus active systems. Reliability is reduced with complexity. The wiring harness is also more detailed. Tesla does use an active system, its ultrasonic sensor array.

 

[diplomat33]

Reliability is reduced with complexity.

Having sensors of different types increases reliability because it is less likely that 2-3 sensors will fail at the same thing than just 1 sensor. For example, one camera can fail at detecting a child crossing the street. Camera + lidar is much less likely to fail at detecting the child. And a camera + lidar + radar is even less likely to fail at detecting the child. Reliability is all about reducing the chance of perception failure in different conditions. You cannot afford to accidentally kill a child crossing the street because your simple vision-only system did not see the child.

So yes, adding more sensors will add more complexity but it is worth it to have higher safety. Simplicity through fewer sensors in this case is not better because your system will be less safe.

Tesla does use an active system, its ultrasonic sensor array.

Which is extremely short range, only 8 meters. So it is only useful for parking and some lane changes.

 

[2101Guy]

Tesla has opted for simplicity. LIDAR and radar are active systems. They transmit a beam and then a sensor measures the return signal. This is more complicated than a vision-based system which passively senses the condition of the environment. Tesla's philosophy even extends to its auto wipers.

How are the overall reviews on that feature...

 

[dramsey]

Edit* Found the qoute

“I’m highly confident we will solve full self-driving, and it still seems to be this year. I know people are like, ‘he says that.’ But it does seem to be epic. It does seem as though we are converging on full self-driving this year.” (c)ELON MUSK

I kinda wish Elon would say something like "I think we'll have SAE Level 3™ (yes, it's trademarked) standard J3016 this year" instead of vague things like "solve full self-driving". I mean, since there's a defined standard for this stuff (even if it's not very detailed).

 

[Knightshade]

The post he made calling me out:

Yes, because you've made closing in on 3000 posts here, going back 5 years, and every one is about how FSD sucks.

You have literally nothing else to say or contribute- ever- in 5 years of posting.


Which makes your accusing others of "only ever repeating the same thing" extra funny.

Dude. It’s not EVEN a debate

Right.

You asked if you were reading Elons remarks from last week correctly.

Multiple people explained you weren't.

So you moved the goalposts to try and bring in things he said a year or two ago instead.


That's not a debate- that's you arguing dishonestly- as always.

 

[Knightshade]

Was talking about toddler obscured by car that humans can see and identify which car cannot see. Fairly sure FSD does not have this capability to analyze 95% obscured objects and quickly and accurately identify them.

<citation required>

I understand that originally you made it very clear it was a situation where the toddler was not visible at all until 1-2 feet before (a different, less common situation).

Right.

Because anything further away the cameras can see

Just like a human can.

The human and the car have roughly the same blind spots looking in a single direction.

Except the car can look in all directions at the same time- the human can't.

Plus the car has faster than human reaction time.


Thus the argument the car can't be safer than a human without 360 downward facing cameras--- the actual original claim being discussed- is not correct.

here's the original, incorrect, quote again in case you're unclear how this came up:

HW3 Capability Limitation - The car doesn't have any down facing 360 degree cameras, and is incapable of detecting nearby pedestrians/obstacles. It has to rely on slow speed and anticipating from what it did see.

Humans don't have those cameras either of course. So the spots the "car" can't see (below the hoodline in front) are the same spots the human can't (because the human also can't see "through" the hood- in fact the human is worse at this because it can't look all 8 ways at once and lacks ultrasonics and has slower reaction times to what it CAN see)

The poster mistakenly thought the car computer can't/doesn't anticipate if something it DOES see further away is going to move across its path-- which is not only untrue, the most recent release notes specifically call out an improvement being made to that ability.

 

[2101Guy]

I kinda wish Elon would say something like "I think we'll have SAE Level 3™ (yes, it's trademarked) standard J3016 this year" instead of vague things like "solve full self-driving". I mean, since there's a defined standard for this stuff (even if it's not very detailed).

I dont disagree with you. But let's be real: If he did as you suggested? The FSD take rate would be even LESS than 7%. Gotta stretch the truth to get the money from the lesser IQ'ed.

 

[Matias]

Plus we don't want new cars to cost $250,000.

Baidu unveils Apollo RT6 Level 4 autonomous robotaxi with detachable steering wheel

During its Baidu World 2022 technology conference today, the Chinese technology company unveiled its sixth-generation fully-autonomous vehicle – the Apollo...

electrek.co

According to the AI company, it was able to get the cost of the Apollo RT6 down to about $37,000 per EV,

 

[Knightshade]

Baidu unveils Apollo RT6 Level 4 autonomous robotaxi with detachable steering wheel

During its Baidu World 2022 technology conference today, the Chinese technology company unveiled its sixth-generation fully-autonomous vehicle – the Apollo...

electrek.co

According to the AI company, it was able to get the cost of the Apollo RT6 down to about $37,000 per EV,

We literally just covered this.

That isn't a current product anyone can buy.

it's one they hope to deploy, themselves, in the future, citing their internal cost, not a public price, and again a thing not even on the road today let alone providing L5 driving everywhere today.


Your own link says they 'hope" put have them in "trial" operation second half of 2023.


Much of this thread is lambasting Tesla for "hopes" that missed their target dates.

Get back to me when there's a $37,000 L5 car anybody can actually buy today. From anybody.

 

[FloridaJohn]

We literally just covered this.

That isn't a current product anyone can buy.

it's one they hope to deploy, themselves, in the future, citing their internal cost, not a public price, and again a thing not even on the road today let alone providing L5 driving everywhere today.


Your own link says they 'hope" put have them in "trial" operation second half of 2023.


Much of this thread is lambasting Tesla for "hopes" that missed their target dates.

Get back to me when there's a $37,000 L5 car anybody can actually buy today. From anybody.

So your argument is that they would sell their car that costs them $37k for $250k?

Pretty good profit margin on that business plan you got there.

 

[AlanSubie4Life]

<citation required>

Sure, it definitely is not something I KNOW to be true. But I doubt it is currently within the capabilities. I haven’t seen anyone test it though. Would be a relatively easy test to set up. I don’t believe that the car can currently identify “feet only” moving in front of a parked car, though clearly it is possible. Just have to filter out blowing leaves and whatnot. As I have made clear, separate topic from your original discussion.

A good test would be rolling balls (tennis balls, etc) across the street from behind an enormous vehicle parked on the side of the street, dispensing with the feet to start with, and see how the car reacts. I assume it would rapidly slow down in anticipation. If it doesn’t, you could add in feet under the large vehicle (visible to the driver), and see how it behaves.

Just try to duplicate a situation that occurs many times a day.

Thus the argument the car can't be safer than a human without 360 downward facing cameras--- the actual original claim being discussed- is not correct.

I agree it could be safer than a human.

 

[Bladerskb]

Much like the people who know the earth isn't flat, that vaccines are effective, and that global warming is actually happening, all have the same talking points.

No they don't. They don't regurgitate the same thing word for word. They make statements based on the actual evidence and facts that they know and can elaborate, expound and build on it. The people who only regurgitate word for word are people who have no idea what they are talking about.

When you reason from first principles, you generally will say similar things to others who do so.
On the other side we have the "Don't understand how anything works and rely on magical thinking contrary to facts" folks like yourself.

This is what your fact based first principles thinking looks like? I hate to break it to you but being distracted and texting ISN'T a sensor suite comparison. Imagine if people thought fact based scientific sensor suite comparison consisted of this line of reasoning: although golden eagles can see more than 8x farther than humans and more colors. It sleeps longer and get distracted by prey more easily so that means human have better eyes. Immaculate fact based logical reasoning right?

They can see in all directions simultaneously- a human can't.--- and without ever being distracted by the radio, their phone, a screaming child, a bad medical diagnosis, etc....
So they've already got a pretty better-than-human sensor suite.

Does anyone who ever actually took a science class understand why putting x-ray vision on a car is a bad idea?

Where in my post did you read that we should put xray vision on the cars?
Did you even read my post? Or did u "read" it the same way you read my 2016 posts? Make it make sense.

Great example here where I previously mentioned NO vehicle could avoid the toddler accident unless it had x-ray vision.

Your example demonstrated perfectly how you regurgitate whatever elon and Teslas says. You don't even do the bare minimum to reason using their PR marketing.

"I've been driving a lot of years and can't recall that ever happening to me.... but yes I suppose if one toddled into the street from being hidden behind the bumper of a big truck just a foot or two before you reach the spot FSD might hit it. A human 100% for sure would too though. So would an FSD car with $250,000 of LIDAR because the car STILL had no line of sight (or light) to the toddler until it was 1-2 feet away and a car at speed can't stop in 1-2 feet.... so not sure how this is relevant to anything about safety- NOTHING could avoid that accident unless the car has x-ray vision."​

You continually spew the same talking points handed to you by Elon, that Tesla with their garbage, old, narrow and outdated forward radar can see the vehicles ahead of the vehicle in-front. And you have claimed it has avoided accidents because of that. Yet you can't seem to deduce logically that a 360 degree imaging radars with 1250x times more resolution (500k vs 400 pps), orders of magnitude more virtual channels (2,304 vs 8), orders of magnitude better dynamic range, orders of magnitude better side lobes, etc, would be able to see pedestrians behind any car/suv/truck. Not just see but also distinguish it as a pedestrian due to their unique signature, track their precise velocity and distance and finally classify its dimensions due to its resolution.

This is exactly what i mean by you regurgitate Elon's talking points word for word without having a clue what you are talking about.
You don't need X-Ray vision. Camera plus Lidar plus Radar provides everything you need. Something you can't seem to grasp until I guess Elon mentions it. Then it automatically becomes factual truth.

The moral of this post is: Don't let Elon think for you. Its not yet illegal!
Oh I'm sorry, that would be considered "magical thinking contrary to facts".

 

[Knightshade]

So your argument is that they would sell their car that costs them $37k for $250k?

Nope.

My argument is there's still no evidence it's a buyable L5 car for $37,000.

All we have a is press release, of a future vehicle they "hope" to have in "limited testing" next year and almost certainly heavily geofenced, and not for sale at any price to a customer.


Because that's what the actual facts are here.


So citing it as an example of someone offering a cheap L5 consumer-buyable car is weird since it's literally none of those things presently and we have no evidence it will be one anytime soon.

Right now, nobody has an L5 car. At any price.

A very few companies have L4 cars, that are very very heavily geofenced and still don't work at scale, and are not actually for sale to the public at all (and would cost a lot more than $37,000 if they were).



Just about every company on earth, in both the EV and autonomy space, continually makes future promises they never meet (or are at least many years beyond when they originally claimed they'd meet them). Tesla is hardly alone there.

 

[FloridaJohn]

Nope.

My argument is there's still no evidence it's a buyable L5 car for $37,000.

All we have a is press release, of a future vehicle they "hope" to have in "limited testing" next year and almost certainly heavily geofenced, and not for sale at any price to a customer.


Because that's what the actual facts are here.


So citing it as an example of someone offering a cheap L5 consumer-buyable car is weird since it's literally none of those things presently and we have no evidence it will be one anytime soon.

Right now, nobody has an L5 car. At any price.

A very few companies have L4 cars, that are very very heavily geofenced and still don't work at scale, and are not actually for sale to the public at all (and would cost a lot more than $37,000 if they were).



Just about every company on earth, in both the EV and autonomy space, continually makes future promises they never meet (or are at least many years beyond when they originally claimed they'd meet them). Tesla is hardly alone there.

Got it. Can you please refrain from making arguments that an L5 car will cost $250k, since as you state yourself, you have no information on what an L5 robotaxi would sell for?