The WAN Show incorrect about too many things Tesla

[Jet_ski]

I heard some incorrect info about the new Tesla Model S yesterday on the WAN Show that I wanted to point out. I know it’s not a car channel and it probably won’t matter. But I’m sure the writers don’t want to go on incorrect info.

 

The new Tesla Model S is a refreshed version of the old one and there are two versions: Long Range starting at $73,990 and the Plaid starting at $123,990. Both versions come with the infotainment system that’s based on the new AMD chips which runs video games. It’s not just the Plaid (more expensive model) that comes with this feature.

 

Another thing to point out is that the AMD based infotainment system is a completely separate system from the self driving system. In all of their cars Tesla uses their in-house designed chips for the driving functions. They call it FSD Hardware and it’s based on ARM [I think], currently at version 3. The FSD chip is a specialized chip that’s built to only do neural network computations.

Also Tesla doesn’t make a gaming controller, at least not yet. The images circulating are of a prototype and you have to buy your own Xbox or PS controllers. At the event they were using Xbox controllers.

 

On a side note I want to say that comparing to the Porsche Tycan that’s was showed in a ShortCircuit video, the refreshed Model S is the absolute winner. The Tycan starts at $79,000 with a range of up to 227 miles, the more expensive ones can cost over $185,000 without any of the options!!!! The Model S long range costs $73,990 and gets 405 miles. And according to some rumors they are waiting for the EPA to update this range higher.

 

Linus is right that it’s an expensive car. Even the one that costs $73,990 is lol-years out of my budget.

[Jet_ski]

Ok so I watched the tonight’s WAN show and again Linus said some srsly misleading stuff about Tesla. Yes, please pile on the the fanboi jokes but I’m here only to point out the facts nothing else.

1. Tesla upgrades the self-driving hardware of the older cars for those who purchased the Full Self Driving (FSD) free of charge. So they aren’t maintaining advanced software for different hardware.
2. When Full Self Driving (FSD) was introduced in 2018 it costed $3,000 which is a lot less than the current price of $10,000. Tesla has always said the prices were lower back then because the feature wasn’t as built out as it is today.
3. Radar and video can conflict mainly because [all] radars can fail to detect objects even huge ones. Radar uses the Doppler effect to detect and measure distance to objects. Under certain conditions the Doppler effect is not detected. Here’s a full explanation:
   1. “When the source and the observer are moving mutually in perpendicular directions then the Doppler Effect is not observed. When the medium only is moving then the Doppler Effect is not observed. When the distance between the source and the observer is constant then the Doppler Effect is not observed.”
   2. This means for example, if you are driving perpendicular to an 18 wheeler, radar would not detect it! While a purely vision system would see the truck, radar won't.
4. Tesla isn't the only car maker who's collision systems are not using radar. Subaru has been using a pure vision system (camera) for collision detection for years now! And they've had a very good track record.
5. Tesla Model S and X are not the same cars, they look the same but the internals have all changed. I'm not sure where to begin with this but ignoring the design, it's a completely different car. Completely new motors, new batteries, new chassis, new HVAC etc. It's like saying the iPhone 12 is "pretty much the same" phone as iPhone 6 just because they look similar without knowing what's inside them. There are plenty of sources about the new Model S/X you can find.

[TempestCatto]

I'll just stick to good ol combustion engines. I like the noise, the feel, and the late nights/early mornings with 90% of the car torn apart only to put it back together in a matter of hours right before the track opens only to completely fail and ruin the whole day.

 

But seriously, I'll stick to what I like best

[18358414]

13 hours ago, Jet_ski said:

Radar and video can conflict mainly because [all] radars can fail to detect objects even huge ones. Radar uses the Doppler effect to detect and measure distance to objects. Under certain conditions the Doppler effect is not detected

A robust implementation can deal with conflicting sensor data. If the radar has known issues, then increase the reliance on visual sensors when information is available.  The issue here is that in low-visibility weather, having something that can potentially (even if not 100% of the time) alert you before you pile your face into the back of a truck is better than having nothing.  It doesn't have to self-drive, but it can let you know before you wreck yourself.  If it conflicts with visual information then just rely more on visual data when it is available.

[Dedayog]

On 6/12/2021 at 3:24 PM, Jet_ski said:

I heard some incorrect info about the new Tesla Model S yesterday on the WAN Show that I wanted to point out. I know it’s not a car channel and it probably won’t matter. But I’m sure the writers don’t want to go on incorrect info.

 

The new Tesla Model S is a refreshed version of the old one and there are two versions: Long Range starting at $73,990 and the Plaid starting at $123,990. Both versions come with the infotainment system that’s based on the new AMD chips which runs video games. It’s not just the Plaid (more expensive model) that comes with this feature.

 

Another thing to point out is that the AMD based infotainment system is a completely separate system from the self driving system. In all of their cars Tesla uses their in-house designed chips for the driving functions. They call it FSD Hardware and it’s based on ARM [I think], currently at version 3. The FSD chip is a specialized chip that’s built to only do neural network computations.

Also Tesla doesn’t make a gaming controller, at least not yet. The images circulating are of a prototype and you have to buy your own Xbox or PS controllers. At the event they were using Xbox controllers.

 

On a side note I want to say that comparing to the Porsche Tycan that’s was showed in a ShortCircuit video, the refreshed Model S is the absolute winner. The Tycan starts at $79,000 with a range of up to 227 miles, the more expensive ones can cost over $185,000 without any of the options!!!! The Model S long range costs $73,990 and gets 405 miles. And according to some rumors they are waiting for the EPA to update this range higher.

 

Linus is right that it’s an expensive car. Even the one that costs $73,990 is lol-years out of my budget.

Can you come up with other comparisons than just this?

 

On a side note I want to say that comparing to the Porsche Tycan that’s was showed in a ShortCircuit video, the refreshed Model S is the absolute winner. The Tycan starts at $79,000 with a range of up to 227 miles, the more expensive ones can cost over $185,000 without any of the options!!!! The Model S long range costs $73,990 and gets 405 miles. And according to some rumors they are waiting for the EPA to update this range higher."

 

Range is only 1 very small aspect of why most people buy cars.  Look at gasoline cars to prove this point.  So many cars get terrible gas mileage and range from their tanks.

 

If Range is what drives EV buyers then they're wearing blinders, or at the very least the early adopters of this EV tech.  Those of use that appreciate EV aspects but enjoy driving tend to take it with a very large grain of salt.

 

Case in point, my EV only has a 23 mile range   

[TempestCatto]

21 minutes ago, Dedayog said:

Range is only 1 very small aspect of why most people buy cars.  Look at gasoline cars to prove this point.  So many cars get terrible gas mileage and range from their tanks.

Difference being I can fill the tank in a matter of a minute or two, where as you have quite a longer wait with an EV. Not to mention I don't have to worry about where a charger will be; at least, we're not there yet. A gas station is very hard not to find in any town, or even on the side of the highway. That's why people feel as though gas cars are more freeing than EVs.

 

But if you asked Mercedes about fuel economy on any of their cars, they'd say "if you have to ask, you can't afford it." I don't think they'll tell you that anymore but I still find it funny. But these days, while people do care about fuel economy in a car, they mostly care about reliability, safety, and deprecation. Nobody really thinks of range when they buy a gas car.

[Jet_ski]

5 hours ago, Dedayog said:

Can you come up with other comparisons than just this?

 

Range is only 1 very small aspect of why most people buy cars.  Look at gasoline cars to prove this point.  So many cars get terrible gas mileage and range from their tanks.

 

If Range is what drives EV buyers then they're wearing blinders, or at the very least the early adopters of this EV tech.  Those of use that appreciate EV aspects but enjoy driving tend to take it with a very large grain of salt.

 

Case in point, my EV only has a 23 mile range   

I made that specific comparison because the Tycan was the other car mentioned on the WAN show.

 

Range is mainly a topic because most people have this perception that EVs aren’t good enough for longer trips.

4 hours ago, TempestCatto said:

Difference being I can fill the tank in a matter of a minute or two, where as you have quite a longer wait with an EV. Not to mention I don't have to worry about where a charger will be; at least, we're not there yet. A gas station is very hard not to find in any town, or even on the side of the highway. That's why people feel as though gas cars are more freeing than EVs.

 

But if you asked Mercedes about fuel economy on any of their cars, they'd say "if you have to ask, you can't afford it." I don't think they'll tell you that anymore but I still find it funny. But these days, while people do care about fuel economy in a car, they mostly care about reliability, safety, and deprecation. Nobody really thinks of range when they buy a gas car.

Charging rates have improved a lot overall. With Tesla cars you can get up to 200 miles of range with 15 mins of charging. Also you can charge your EV at home; you don’t need charging stations unless you’re on longer trips. Tesla alone has 1000 supercharger locations just in the US.

 

I personally care about the efficiency of the car and so should everyone. I wouldn’t drive a Prius but I wouldn’t get something that’s too inefficient relative to other cars. Mercedes can be as snobbish as they want; it’s expected of a company that makes moving bricks like the G-Wagon.

 

6 hours ago, Yuri Fury said:

A robust implementation can deal with conflicting sensor data. If the radar has known issues, then increase the reliance on visual sensors when information is available.  The issue here is that in low-visibility weather, having something that can potentially (even if not 100% of the time) alert you before you pile your face into the back of a truck is better than having nothing.  It doesn't have to self-drive, but it can let you know before you wreck yourself.  If it conflicts with visual information then just rely more on visual data when it is available.

It doesn’t work that way. If you read through the scenarios in which radar fails, it undermines the reasons for having them on a car. There are a lot scenarios where incorrectly stopping somewhere is dangerous and it’s impossible to reliably choose between different types of sensors. If the weather is so bad that 8 cameras on your car don’t have sufficient visibility, self driving shouldn’t be active and you probably shouldn’t be on the road.

[Kronoton]

3 hours ago, Jet_ski said:

t doesn’t work that way. If you read through the scenarios in which radar fails, it undermines the reasons for having them on a car.

 

I really hate it when all that fake "self driving"/"autopilot" nonsense gets mixed in with EV part of an EV which than gets mixed in with how good a car it is.

 

Last things 1st, watch the Taycan video and it becomes clear that as "a car" it is vastly superior to anything Tesla right on par with non EV Porsches in that price range.

Teslas of all kind still fail in that regard when compared similar (priced) EV and non EV.

 

Back to radar and such, it is an attempt to fix a symptom in part caused by fixing that symptom as it helps with the notion that being distracted while driving ain't that bad resulting in more distracted drivers being "saved" by that tech.

 

Keep your hand on the wheel, eyes on the road, don't text, don't scroll through menus adjusting the AC or wipers and everyone will be fine. Can't do that for how ever long the trip will take? Call an Uber.

That will be true until "steering wheel" becomes an item at the bottom of the options list (something we are at least another decade away from).

[WolframaticAlpha]

Too many uninformed people here. Everyone knows that the best ev is the Geoff:

 

So beautiful.

[Tieox]

Don't care still would own a Tesla in a nano second.   

 

Bring on the Model 2!

[tkitch]

9 hours ago, Kronoton said:

Keep your hand on the wheel, eyes on the road, don't text,

Yes and yes, both of those are pretty much true, and should be a thing.

 

Quote

don't scroll through menus adjusting the AC or wipers and everyone will be fine. Can't do that for how ever long the trip will take? Call an Uber.

 

Now you're just being unreasonable.  Not changing the wipers?  Not adjusting the AC?  Are you kidding?

I can do all of those things in my PHEV (plug in Hybrid) basically without looking away from the road.  the criticism that you can't do it without staring at a computer screen?  COMPLETELY VALID.  

 

No, you don't get to say "suck it up" that's bullshit in this case.

 

[Kronoton]

14 minutes ago, tkitch said:

the criticism that you can't do it without staring at a computer screen?  COMPLETELY VALID.  

 

Ermm, thats what I was saying.

 

As "suck it up", if can't be bothered to drive the car, don't drive the car. That should be a nobrainer.

[tkitch]

1 minute ago, Kronoton said:

 

Ermm, thats what I was saying.

 

As "suck it up", if can't be bothered to drive the car, don't drive the car. That should be a nobrainer.

that's not what you said, and not what you said here, either.

[TempestCatto]

15 hours ago, Jet_ski said:

Charging rates have improved a lot overall. With Tesla cars you can get up to 200 miles of range with 15 mins of charging. Also you can charge your EV at home; you don’t need charging stations unless you’re on longer trips. Tesla alone has 1000 supercharger locations just in the US.

On a road trip or business trip 15 minutes is a lot to wait. My point here is you have to worry about where the fast charging stations are located, because you have to adjust your route as such. Where as with gas, you pass them along the highway. I do wish they'd put at least a few fast chargers at bigger gas stations. Would be easier for those road tripping their EVs.

 

15 hours ago, Jet_ski said:

I personally care about the efficiency of the car and so should everyone. I wouldn’t drive a Prius but I wouldn’t get something that’s too inefficient relative to other cars.

It's important for sure. But like I said, most people are concerned about safety and reliability. Pretty much any car you buy today makes good enough fuel economy that most people don't even bother looking.

 

15 hours ago, Jet_ski said:

Mercedes can be as snobbish as they want; it’s expected of a company that makes moving bricks like the G-Wagon.

Hey leave the sexy G-Wagon out of this.

[Jet_ski]

45 minutes ago, TempestCatto said:

Hey leave the sexy G-Wagon out of this.

That thing is an abomination in this day and age. But I understand why people like it.

[Montana One-Six]

18 hours ago, Jet_ski said:

Range is mainly a topic because most people have this perception that EVs aren’t good enough for longer trips.

Really depends on where you are driving. If you are driving through remote areas a lot or you like doing road trips any EV is out of the question. They only make sense in areas with very good infrastructure and that's pretty much it.

 

19 hours ago, Jet_ski said:

I personally care about the efficiency of the car and so should everyone. I wouldn’t drive a Prius but I wouldn’t get something that’s too inefficient relative to other cars. Mercedes can be as snobbish as they want; it’s expected of a company that makes moving bricks like the G-Wagon.

I also care about efficiency but that is not the most important point. I care a lot more about reliability, relative ease of maintenance (so I can do stuff myself) and looks.

Talking about looks the G-Wagon looks better than any EV. Which is probably one of the biggest complains they just don't look good at least not as good as their non EV counterparts.

 

As someone who drives a modded Wrangler things I like about cars like mine is the freedom of movement. I can drive to a camping ground somewhere in a National Park or in a remote area which is only accessible through old rugged dirt roads without worrying if I'm going to make it. I can drive my car onto the beach at night to meet with friends, camp there and get off the beach in the morning without worrying about getting stuck or things like that.

Since there aren't any EVs yet that are even remotely capable of that why even bother.

[Mihle]

On paper electric cars should be more reliable and require less maintenance than non electric as they have fewer parts.

 

The ones with most range like Tesla's or the new Ford one have so long range that when you have to charge you would probably want to eat too.

[wanderingfool2]

The part that bothered me is always the talk regarding how they treat their workers during that pandemic...without any thought to what would happen if they hadn't tried restarting.

 

For Linus, I would ask, what his response would be if he was told he had to shutdown the LTT channel, and floatplane during the pandemic (If lets say for some reason they had to if for some weird reason editors working from home wasn't an option).  It's easy to forget that Tesla is still a smaller company, in that they can't afford to really halt production for a better part of a year.  Yes, it's not great, but it's not like what it's painted out to be by the media (as only one side of the story is being told, in the perspective of the employees).  I'm not trying to defend Tesla, I don't agree with all that was done, but I do understand and accept the reasoning (speaking as someone who was laid off during covid initially because the company couldn't afford to keep me).  The pandemic has wreaked havoc on some sectors, and when the choice can mean make and break for a company.

 

The other part is where Linus implied that he wouldn't be surprised if Tesla was playing fast and loose with the safety (by changing the structural components).  Yes, he did add the disclaimer, but I mean come on...Tesla's are rated top in safety, and the body changes are making it safer and get tested as per regulations (and unlike other unnamed companies they haven't had a safety scandal like the one where the execs knew about a missing 25 cent part issue involving the key that lead to the death and accidents and still decided to cover it up).

 

I do think that auto-pilot was oversold in what their intended capabilities were...but to be clear, there is auto-pilot and FSD (where FSD was suppose to be the one that drives you from point a to b...which seems as though it might be happening soon given the current beta program).  The software though is going to a vision only, and will run on the same hardware, so they don't really have to have to worry about as many differences as one may think.  With that said, I do think they are overzealous in how soon things will arrive.  (Also, on older vehicles they do have limited feature sets in terms of the auto-pilot)  [They've scaled the price though based on how advanced FSD has become though]

 

Now onto what a few people have said

On 6/18/2021 at 11:22 PM, Jet_ski said:

Tesla isn't the only car maker who's collision systems are not using radar. Subaru has been using a pure vision system (camera) for collision detection for years now! And they've had a very good track record.

Yes and no.  Removing radar, after the next update I don't think will be an issue...with that said, they should have delayed car delivery until the update was ready to push to the customers (The May vehicles without radar shouldn't have been delivered, or at least given people more of a heads up).  I don't agree with how Tesla is rolling out that.

 

On 6/19/2021 at 12:48 PM, Yuri Fury said:

A robust implementation can deal with conflicting sensor data. If the radar has known issues, then increase the reliance on visual sensors when information is available.  The issue here is that in low-visibility weather, having something that can potentially (even if not 100% of the time) alert you before you pile your face into the back of a truck is better than having nothing.  It doesn't have to self-drive, but it can let you know before you wreck yourself.  If it conflicts with visual information then just rely more on visual data when it is available.

Making it decide between radar vs vision creates a whole set of issues.  Radar can lie (about objects about to be hit), which on Tesla's can cause phantom braking.  When comparing data, you now have to spend valuable resources on figuring out whether vision is telling the truth or radar is telling the truth (which in itself is a hard problem that we can't really solve).  I think one of the interviews with George Hotz (security hacker), who created comma.ai mentions that they are moving away from radar as well because of things like this.

 

While yes, radar has saved lives, it has also caused injuries as well (which no one seems to talk about).  https://www.cbsnews.com/news/automatic-emergency-braking-in-cars-will-be-standard-by-2022-but-false-activations-still-a-concern/

That is 14 crashes and 5 injuries that have been reported and linked to the AEB system failing (most likely due to radar giving a false positive)

 

As well, if you have low-visibility to where radar would save you, you are driving too fast (and it's safety features like this that put people at ease for driving too quickly in poor conditions).

[Jet_ski]

2 hours ago, wanderingfool2 said:

The other part is where Linus implied that he wouldn't be surprised if Tesla was playing fast and loose with the safety (by changing the structural components).  Yes, he did add the disclaimer, but I mean come on...Tesla's are rated top in safety, and the body changes are making it safer and get tested as per regulations (and unlike other unnamed companies they haven't had a safety scandal like the one where the execs knew about a missing 25 cent part issue involving the key that lead to the death and accidents and still decided to cover it up).

Completely agree. Another thing to add is Tesla’s vehicle safety ratings and crash history, their cars have all earned the highest safety rating. It’s all public data; what about that says they are playing it fast and loose?! They are literally rated safer than Volvos!! It’s insane that companies like GM, VW, Mercedes, and Toyota can have massive 80,000 car recalls for break failure or fire hazard and they end up with essentially no news coverage whatsoever or blow back.

 

There are about 200,000 car fires in the US from combustion engine cars. I’ve never seen a media frenzy about it. I can’t remember the last time any incident was covered in the news. But any Tesla fire or incident gets so sensationalized like it’s the apocalypse.

 

2 hours ago, wanderingfool2 said:

Now onto what a few people have said

Yes and no.  Removing radar, after the next update I don't think will be an issue...with that said, they should have delayed car delivery until the update was ready to push to the customers (The May vehicles without radar shouldn't have been delivered, or at least given people more of a heads up).  I don't agree with how Tesla is rolling out that.

I don’t know if they could've handled that any better. Not everyone buys Autopilot or FSD, so it doesn’t impact that many people who are about to get their order delivered. Ppl can cancel or return their order though. One would hope they tried to minimize the number of people this effected.

[MultiGamerClub]

Maybe in the future if tesla isnt so expensive, it could be cool to get an RWD one to do some stuff

Edited June 21, 2021 by MultiGamerClub
forgot adding cool.

[wanderingfool2]

6 hours ago, Jet_ski said:

I don’t know if they could've handled that any better. Not everyone buys Autopilot or FSD, so it doesn’t impact that many people who are about to get their order delivered. Ppl can cancel or return their order though. One would hope they tried to minimize the number of people this effected.

If could potentially affect more than just FSD and autopilot people.  See the quote from Tesla below

 

Quote

 Emergency Lane Departure Avoidance may be disabled at delivery

This is vehicles delivered in May 2021 and beyond.  I think the key word is may, in that it might not necessarily be disabled, but either way at this stage it's unacceptable to deliver vehicles with that.  (Their pure vision update would be adding back in this feature, but from what I know it's still not out yet, although it apparently is the next update...which even on Tesla's timeline would be in less than a few months).

 

*This isn't to say that the safety feature might not already be in there, but maybe not allowed to be called that...either way it's unacceptable*  Given that FSD and autopilot still work with this, it'd be my guess that the feature still exists but just not legally allowed to advertise it yet (although since it does affect "safety" they are getting the users to sign saying they acknowledge the missing feature when the vehicle is delivered).  The rumor is at least that it's tested to be just as safe, but just needs to regulatory approval to be able to call it that.

[Jet_ski]

5 hours ago, wanderingfool2 said:

*This isn't to say that the safety feature might not already be in there, but maybe not allowed to be called that...either way it's unacceptable*  Given that FSD and autopilot still work with this, it'd be my guess that the feature still exists but just not legally allowed to advertise it yet (although since it does affect "safety" they are getting the users to sign saying they acknowledge the missing feature when the vehicle is delivered).  The rumor is at least that it's tested to be just as safe, but just needs to regulatory approval to be able to call it that.

When you put it like that I’m inclined to agree. From the official page it seems like the following are impacted:

 

• Autosteer will be limited to a maximum speed of 75 mph and a longer minimum following distance.
• Smart Summon (if equipped) and Emergency Lane Departure Avoidance may be disabled at delivery.

https://www.tesla.com/support/transitioning-tesla-vision

 

If I were to buy a car I would postpone purchasing those features unless they’d give me a discount for not having those features right away.

[wanderingfool2]

19 hours ago, Jet_ski said:

If I were to buy a car I would postpone purchasing those features unless they’d give me a discount for not having those features right away.

See, this is where I do think Tesla utterly failed.  They pretty much showed up and said the feature was missing and had it as a take it or leave it option.  While I am still guessing it's likely not as big of a deal as mention (like I suspect the ELDA is likely still active), it still is something that shouldn't have been done until the update was out.

[18358414]

On 6/19/2021 at 10:22 PM, Jet_ski said:

It doesn’t work that way. If you read through the scenarios in which radar fails, it undermines the reasons for having them on a car. There are a lot scenarios where incorrectly stopping somewhere is dangerous and it’s impossible to reliably choose between different types of sensors. If the weather is so bad that 8 cameras on your car don’t have sufficient visibility, self driving shouldn’t be active and you probably shouldn’t be on the road.

I never said it should have the authority to STOP the vehicle.  I said that giving drivers a warning in low visibility can be beneficial (in a roundabout non-committal way).  Have you ever looked into the wonderful world of space weather prediction?  There's plenty of noise coming from sensors all around the world and some of that data is no good; so you apply a model that can deal with noisy information.  If you can't design a robust system around input noise, then don't use the input (in this case RADAR) for self-driving and rely on the visual data when you have it.  Good luck telling people not to drive in bad weather, some people couldn't even wear masks in a pandemic. 

 

On 6/21/2021 at 2:31 AM, wanderingfool2 said:

Radar can lie (about objects about to be hit), which on Tesla's can cause phantom braking.  When comparing data, you now have to spend valuable resources on figuring out whether vision is telling the truth or radar is telling the truth (which in itself is a hard problem that we can't really solve).  I think one of the interviews with George Hotz (security hacker), who created comma.ai mentions that they are moving away from radar as well because of things like this.

 

While yes, radar has saved lives, it has also caused injuries as well (which no one seems to talk about).  https://www.cbsnews.com/news/automatic-emergency-braking-in-cars-will-be-standard-by-2022-but-false-activations-still-a-concern/

That is 14 crashes and 5 injuries that have been reported and linked to the AEB system failing (most likely due to radar giving a false positive)

 

As well, if you have low-visibility to where radar would save you, you are driving too fast (and it's safety features like this that put people at ease for driving too quickly in poor conditions).

Well, if we're removing radar because it's unreliable, then that tells us a lot about which system should be trusted more (thanks for the link BTW).  If you're going to argue that it's an issue that we can't solve, then you're approaching it with a defeatist attitude.  In fact, enhancing RADAR sensor fidelity is the basis for a lot of research work right now (https://arxiv.org/pdf/1504.01716.pdf, https://www.nature.com/articles/s42256-020-00288-6, https://arxiv.org/pdf/2007.11102.pdf, https://openaccess.thecvf.com/content_CVPR_2020/papers/Guan_Through_Fog_High-Resolution_Imaging_Using_Millimeter_Wave_Radar_CVPR_2020_paper.pdf to name a few).  Saying that we'll never solve it and giving up is just a bad attitude.  We know the issues with the technology.  We know it needs to be fixed.  What did I say?  I said leave it as a notification for situations where you otherwise wouldn't be able to tell if your moving towards an accident.  Rely more on visual data (translation: if you can't rely on radar, rely on camera; perhaps my language isn't clear?  I apologize if I can't make the intended meaning more obvious.  It means, if you've decided RADAR is unreliable, then don't use it).  If the RADAR has known issues it doesn't mean it can never be used for any application.  Even if the driver can't brake in time, the vehicle can notify the driver so they can slow down, and the manufacturer can tune the confidence interval so you don't have it beeping and pinging every 10 minutes.  It's not impossible to implement in a way that keeps focus on the road.  If people don't want it, they can turn it off.  Just like the seatbelt warning ding.  Antilock brakes don't work 100% of the time and may even lock up.  I still drive around with it in the winter. 

 

The only thing you've really said that pertains to my response is that "if RADAR would save you, you're driving too fast" and "Safety features like this [...] put people at ease for driving too quickly in poor conditions".  Well, tell that to these fine people: https://www.nhtsa.gov/risky-driving/speeding (speeders gonna speed).  We already know what "risk compensation" is and how it affects drivers (https://www.ishn.com/articles/83639-does-feeling-safe-make-us-more-reckless, https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6237.2007.00510.x, https://injuryprevention.bmj.com/content/6/2/82.full).  If that's the argument, then it's time to rip out seat belts (no one talks about seat belt injuries either), antilock brakes, windshield wipers, "check engine" lights, and retro-reflective road markings (none of which offer the "perfect" solution to their respective problems, but are better than having nothing). 

 

That's about as much effort as I care to put in to this topic. I don't intend to own a self-driving vehicle, so I really don't care beyond clarifying an apparent misunderstanding (I never say things in a committal way and people seem confused by my language).  If Tesla wants to take RADAR out, they're allowed to do so without the permission or justification of the internet people.  I now conclude my dyslexic comment, which probably be read by none.

[Jet_ski]

1 hour ago, Yuri Fury said:

I never said it should have the authority to STOP the vehicle.  I said that giving drivers a warning in low visibility can be beneficial (in a roundabout non-committal way).  Have you ever looked into the wonderful world of space weather prediction?  There's plenty of noise coming from sensors all around the world and some of that data is no good; so you apply a model that can deal with noisy information.  If you can't design a robust system around input noise, then don't use the input (in this case RADAR) for self-driving and rely on the visual data when you have it.  Good luck telling people not to drive in bad weather, some people couldn't even wear masks in a pandemic.

I wasn't trying to imply you said that. I was only giving an example of sensor conflicts that can't be resolved in a safe way imo. These are still open problems and most researchers can't agree on which approach is best. I have a personal opinion but I'm no authority on the subject

 

As luck would have it, Tesla's head of AI just gave a presentation a couple of days ago and he explained why they've dropped radar. Worth a watch if you find the topic interesting.