Apple Car Autonomous Driving Scaled Down, Launch Planned for 2028

[Alameda]

It can be both. I own a 1972 Datsun 240z that I restored from the ground up. I drive it for enjoyment, not transportation. For me it’s meant to drive. I have another vehicle for transportation.

Sure… cars are a lot of things, aren’t they? Fashion, status symbol, work vehicle, luxury, camping, recreation, escape… But the primary purpose is transportation. Get to work and pick up groceries.

There are those who think a car needs to have a gasoline engine, or whatever. That’s their wish. But the need, the fundamental purpose, is getting from A to B. If it does not do that, it is not a car. Most everything else is our personal whim.

 

[PsykX]

You just effectively explained why fully autonomous cars will never be deployed.

Oh they will. Because... competition.

The first company who will be able to deploy that will automatically climb the stock market all the way to become the most valued car manufacturer, and will leave all others in the dust.

 

[fasterthanusual]

if true, I would love for this product to be, beyond successful, thus, it cannot miss anything from what is currently offered in say the latest model3-futuremodelY, combined with superb customer support, ease of parts replacement and availability, authentically fun to drive, solid build, highest safety ratings, and insanely reliable … it doesn’t need to be the fastest, more importantly fun to drive, comfortable and a somewhat timeless interior/exterior design that ages well

 

[techwhiz]

I find it completely amazing that people here assume, due to marketing, that Tesla is remotely close to full self driving.

There is a lawsuit with CA DMV for deceptive marketing because Autopilot and FSD are neither.
Tesla is a L2 system that require hands on the wheel because there is no redundancy in the vision system.

Tesla isn't the leader.
BMW and MBZ offer L3 hands off, no attention required on multiple cars.
Tesla is actually ranked 6th or 7th in ADAS with BMW, MBZ, GM and Ford well ahead of Tesla.

Read the review -> https://techcrunch.com/2023/12/28/how-tesla-bmw-ford-gm-and-mercedes-driver-assist-systems-compare/

Tesla's system is flawed at its core. Vision only with no redundancy. Tesla doesn't even have 360 degree cameras. BMW has 360 degree cameras, 5 radar sensors along with 12 ultrasonic.

Have you ever seen a Tesla try to park with vision only? Compare that to a BMW with DAPP and PAP. BMW gets it done if the car will fit in the space.

Tesla is adding high resolution radar to the 3 and others. Why because the time to interpolate and get distance and speed from vision pales in comparison to a radar module that can directly read speed and distance.

Also we won't even get into driver detection where Tesla isn't close to what BMW, MBZ, GM and Ford offer in terms of camera and capacitive steering wheels.

I'm a car guy and have worked in automotive safety. Tesla... Blah, blah, blah.

 

[ChrisA]

My interest level in a self-driving car: ZERO. Cars are meant to be driven. If I don't want to drive I can call an Uber, Lyft or taxi or get someone else to drive. If I have to choose between having a Porsche and owning a self-driving vehicle my choice is already done.

It might be many years from now. But eventually, they will outlaw humans from driving on public roads. People will be horrified at the thought that they could be killed if some other person makes some tiny mistake. Today we accept this and think "It is OK if only a few tens of thousands of people are killed." In Los Angeles the pedestrian death rate is 2.9 people per 100,000 population. It is odd that we should accept this a normal.

I can imagine there being tracks and parks where people can drive cars manually with steering wheels. Just like there are now places you can ride a horse.

It is the same with tailpipe emissions. Today we accept that thousands of people will die from breathing the poison gases that gas cars emit, but at some point, the idea of driving a gas car will be like smoking cigarettes in a kindergarten classroom, totally unacceptable.

 

[Mac Fly (film)]

It might be many years from now. But eventually, they will outlaw humans from driving on public roads.

Wouldn't surprise me. Then again *they* are actually us. If *we* decide against that then it won't happen. *They* would outlaw happiness if they could.

 

[Surf Monkey]

Oh they will. Because... competition.

Competition isn’t the problem. Limitations of the tech and problems with things like liability for fatal crashes is.

The first company who will be able to deploy that will automatically climb the stock market all the way to become the most valued car manufacturer, and will leave all others in the dust.

It’s fun to dream.

 

[Surf Monkey]

Sure… cars are a lot of things, aren’t they? Fashion, status symbol, work vehicle, luxury, camping, recreation, escape… But the primary purpose is transportation. Get to work and pick up groceries.

There are those who think a car needs to have a gasoline engine, or whatever. That’s their wish. But the need, the fundamental purpose, is getting from A to B. If it does not do that, it is not a car. Most everything else is our personal whim.

Wrist watches are for telling the time. Sure, watches are a lot of things. Fashion, status symbol, luxury… but the primary purpose is telling the time. Look at it and know when in the day to pick up groceries.

There are those who think that a watch needs a mechanical mechanism or whatever. That’s their wish. But the need, the fundamental purpose, is telling the time. If it does not do that it is not a watch. Most everything else is our personal whim.

 

[Surf Monkey]

Wouldn't surprise me. Then again *they* are actually us. If *we* decide against that then it won't happen. *They* would outlaw happiness if they could.

Outlaw happiness = regulate potentially lethal vehicles?

I dunno about that one…

 

[Surf Monkey]

It might be many years from now. But eventually, they will outlaw humans from driving on public roads. People will be horrified at the thought that they could be killed if some other person makes some tiny mistake. Today we accept this and think "It is OK if only a few tens of thousands of people are killed." In Los Angeles the pedestrian death rate is 2.9 people per 100,000 population. It is odd that we should accept this a normal.

I can imagine there being tracks and parks where people can drive cars manually with steering wheels. Just like there are now places you can ride a horse.

It is the same with tailpipe emissions. Today we accept that thousands of people will die from breathing the poison gases that gas cars emit, but at some point, the idea of driving a gas car will be like smoking cigarettes in a kindergarten classroom, totally unacceptable.

It’s fun to dream.

 

[jlc1978]

Of course I do take responsibility for all these things today, but I would be much more productive doing something else.

And AFAIK if a vehicle has a certain Level of autonomy, can't recall if it's Level 3 or 4, but no, you won't be held responsible for anything at this point because it'll be able to drive by itself without any human attention. The algorithm, thus the company who wrote it (i.e. Tesla), will be held responsible.

I doubt that will be the case. First and foremost, a driver has the ability to recognize danger and take over control, failure to do tah would put them potentially at fault and liable. In addition, no car manufacturer, unless they are lead by an idiot, would willing assume the potential liability from such a vehicle as it would potentially bankrupt a company.

They have no control over how a driver uses or misuses their vehicles, and thus the risk is too large. I suspect they would get laws passed to absolve them of liability before they deliver such vehicles.

It can be both. I own a 1972 Datsun 240z that I restored from the ground up. I drive it for enjoyment, not transportation. For me it’s meant to drive. I have another vehicle for transportation.

A classic. I have an 85 CSI for the same reasons that I am fixing up. Not a ground up, just a reliable driver.

Oh they will. Because... competition.

The first company who will be able to deploy that will automatically climb the stock market all the way to become the most valued car manufacturer, and will leave all others in the dust.

That would only happen if they were absolved of all liablity; and even then a self driving car isn't enough, IMHO, to make it a must have.

 

[truthsteve]

This is not a competition between lidars and cameras, as they are different technologies with different strengths.

The person asked: "A system that is interrupted by rain?". Current self driving systems that exist today rely on lidar which fail in heavy rain.

Almost all useful detection technologies (apart from very-short range scanning) use electromagnetic radiation. Frequencies or wavelengths range from tens of gigahertz (radars) to hundreds of nanometers (UV). In practice, there is a gap in between radars and IR, but terahertz technologies are emerging to fill that gap. Detection can be either passive sensing (e.g., cameras) relying on external radiation sources or active with own radiation source (radar, lidar).

In general, smaller wavelength (optical) gives better resolution than longer wavelength (radar) but worse penetration and shorter range. This is dictated by physics (diffraction). Active technologies are useful in distance measurements and independent of external illumination. Passive technologies are simpler (and hence less expensive), and they can be used to detect radiation sources, which is especially useful in thermal radiation (IR).

So far these sensors rely on camera fusion for confirmation. You cannot rely on radar/lidar alone. All they add is noise and require a fallback (which is generally camera since it has the most information) when sensors disagree.

There is no silver bullet. Camera-only approach sacrifices a lot of depth information. Accurate and reliable depth detection from two or more camera images is possible in most cases but difficult and prone to pattern-related alignment errors. That is one of the purposes of camouflage patterns.

My eyes don't tell me exactly how far an object is and I drive safely.
Camera-only is the way to go.

It is true that if the ambient illumination is at a good level, passive optical detection is more robust than active in heavy rain or fog. This is due to the physics of photons reflecting back from rain drops between the transmitter and the target. That is why fog lights exist, the problem is exactly the same. On the other hand, there is no difference in the dark, as the only existing light sources are own headlights, and then we talk about active detection (without the timing benefits).

Cameras on Tesla cars have 12-bit sensors. What is pitch black to us, is bright enough for the neural nets to work on. There is plenty of examples on X showing that even without headlights, the cameras can see better than a human. As long as this is true, camera-only approach can absolutely work.

Another example would be a situation with dark concrete, moderate rain, opposite direction traffic with bright headlights, poor road lighting, and a pedestrian in black clothes. The light from opposite traffic headlights scatters from rain and droplets on the lens cover deteriorating the contrast and essentially hiding the dark object. In this case, a NIR-lidar would be able to see the pedestrian (due to being an active technology, and due to the different wavelength where the black clothes would not be that black anymore), and also a passive IR imager would see the heat from their face. A terahertz radar might also be very useful.

Camera only solution has no problem in this scenario.

Very few people actually follow the DMV handbook.

I didn't really ask if they do. A safe self driving car will stop driving if it cannot see.

Trying to make it work with optical cameras only is a rational decision from the commercial point of view; cameras are dirt cheap. Discarding the capabilities offered by other technologies, however, reduces the amount of information significantly.

There is no reason to think lidars would be hideously expensive in the future. ToF (time-of-flights) cameras are essentially lidars and cost relatively little. Lidars can be made inexpensive once the volumes scale up. There are also some interesting opportunities if the headlights are modulated and used as an active source for ToF imaging, but this requires some R&D before it happens.

(Disclosure: I am involved in remote sensing industry but not in automotive lidars.)

Camera-only is not for the sake of business. It's for the sake of safety. More sensors = more noise and latency into the entire system.

It's not a question of cost. Even Teslas says they wouldn't use lidar if you paid them per sensor.

 

[duffman9000]

Of course I do take responsibility for all these things today, but I would be much more productive doing something else.

And AFAIK if a vehicle has a certain Level of autonomy, can't recall if it's Level 3 or 4, but no, you won't be held responsible for anything at this point because it'll be able to drive by itself without any human attention. The algorithm, thus the company who wrote it (i.e. Tesla), will be held responsible.

Tesla will or as wrapped so much legalese around its crappy self driving that Tesla will not be held liable. The driver, like the captain of a ship, is ultimately responsible.

 

[truthsteve]

Here we go... because humans drive with vision so can the car? That's the stupid thoughts from an egotistical Elmo. We use forms of radar to determine distance because vision is consistently poor with distance.

Clearly you've never seen what radar spits out.

https://twitter.com/x/status/1223798507234111488

Look at how noisy the data is. Look at the dots under the bridge. Those dots are directly from radar and they bundle up under the bridge.

Radar detects overpasses as objects because it has poor vertical resolution. Many have phantom brakes because of radar.

And as for shadows, Vision is piss poor at determining distance in shadows. I have seen at least one form of phantom braking because the car is too stupid to ignore shadows cast by an overpass.

LMAO that's not because of shadows. Weird how I already explained why even before reading this part.

This would be because of radar. Andrej literally explained why this happens in overpasses

(25 minutes)

with ultrasonic sensors don't have a problem ignoring my average driveway.

My Model 3 with ultrasonics literally this problem. WHAT???? LOL. I can record a video of this if you want.

Sorry, but you're completely wrong.

 

[truthsteve]

You just effectively explained why fully autonomous cars will never be deployed.

Nope.

Tesla sells auto insurance. They also have the data of FSD. If their FSD data shows it lowers crashes on average, they can still profit from auto insurance even taking liability. Simple math.

 

[DrV]

I doubt that will be the case. First and foremost, a driver has the ability to recognize danger and take over control, failure to do tah would put them potentially at fault and liable. In addition, no car manufacturer, unless they are lead by an idiot, would willing assume the potential liability from such a vehicle as it would potentially bankrupt a company.

Let's get back to the definitions. SAE J3016 2021-04 defines:

5.4 Level or Category 3 - Conditional Driving Automation​

​

The sustained and ODD-specific performance by an ADS of the entire DDT under routine/normal operation (see 3.27) with the expectation that the DDT fallback-ready user is receptive to ADS-issued requests to intervene, as well as to DDT performance-relevant system failures in other vehicle systems, and will respond appropriately.​

...​

NOTE 2: The DDT fallback-ready user need not supervise a Level 3 ADS while it is engaged but is expected to be prepared to either resume DDT performance when the ADS issues a request to intervene or to perform the fallback and achieve a minimal risk condition if the failure condition precludes continued vehicle operation.​

...​

So, even at Level 3 the fallback-ready user does not need to supervise the ADS, i.e., does not need to pay any attention to what is happening outside of the vehicle. The user needs to intervene within a reasonable time frame if the ADS decides to give up. The reasonable time frame is defined as "at least several seconds". Often 10 seconds is quoted as the required response time for the driver.

With Level 4 the terminology changes. The driver/dispatcher determines the conditions are suitable for autonomous driving and engages the ADS. The passenger/dispatcher does not need to perform anything during ADS Level 4 operations. (The roles may switch; if the ADS is disengaged due to, e.g., exiting the area for Level 4, the passenger sitting behind the wheel becomes a driver.)

The vehicle manufacturer may want to push the liability to the driver, but if the owner's manual allows not touching the wheel, it might be a very difficult case from the consumer law perspective. Of course, the obvious solution is to use some extra print in the manual to tell the driver to pay attention to everything and be ready to intervene if things go sideways. This, however, is called Level 2.

That would only happen if they were absolved of all liablity; and even then a self driving car isn't enough, IMHO, to make it a must have.

Again, there are two sides to liability, risk and size. The risk is quite controlled, a car with ADAS may crash once in a while due to unforeseen circumstances. The more interesting factor, however, is the size of damages. If there is a risk of punitive damages, the cost of bearing the risk increases by orders of magnitude.

Punitive damages are rare or non-existent in Europe (except for an island state drifting away into the Atlantic, and relatively rare there, as well). The same applies to Japan. The US and China apply them rather frequently, and these markets will be difficult for any autonomy above Level 2 unless specific regulation is created to limit the risk to compensatory damages (unless there has been gross negligence or criminal acts).

However, even compensatory damages may be difficult to define. What is the right compensation for someone who gets a brain injury? There are huge differences between jurisdictions.

 

[Surf Monkey]

Nope.

Tesla sells auto insurance. They also have the data of FSD. If their FSD data shows it lowers crashes on average, they can still profit from auto insurance even taking liability. Simple math.

LOL

Sure. Tesla will save us all. No one will ever be liable again.

It’s fun to dream.

 

[Morky]

Apple has hired hundreds of not thousands of car people, including the lead developer of Taycan.

Exactly. The talent can be assembled.

 

[Morky]

I'm very sceptical about Self-driving cars appearing anytime soon beyond say, some major US cities.

Living in England, I have doubts about SDCs making an appearance beyond cities anytime soon.
Our country roads — especially down in the south west, with inclines, narrow lanes are often a nightmare.

These narrow hedged lanes, often with zero visibility for oncoming traffic are going to be an SDC programmers nightmare.
Oh and never mind the livestock you get wandering around.
Will the AI recognise a goat or a sheep? It seems to have issues recognising humans… so… yeah.

I want to see it decide if it wants to back up to let a tractor or combine harvester or a **** in a humungous 4x4 through.

Here's a road. Not unusual down here. They're supposedly 1 3/4 cars width… barely 1 car width mostly.
The slight widening to let someone pass are unmarked, often near invisible, and yet these are used by thousands of people every year as. Nightmare during summer holidays.

There's a reason they're only being trialed in Milton Keynes due to its long, straight roads in a block formation. The most atypical town in the UK.

The same goes for up north in the Lake District — or Scotland.
Oh and how about negotiating a Sicilian hill town? Apple Maps just goes into a tail spin right now (To be fair, so does Google Maps!).

The answer is end-to-end ML trained on videos/telemetry of human beviour (I spelled that wrong just for you) on those same roads.

 

[Morky]

Because Apple have been working on it for years. They've been testing self-driving software for years now and won't be simply launching to market cold in 2028. Their chassis designer designed the chassis of Panamera, and was one of the lead developers of Taycan. Apple's car will be judged not by when it launches, but by what they launch when they launch. They've acquired several car battery companies at this point and this will another interesting facet of the design. As Apple develop their car the very technology they will be using to built it will be concurrently maturing, including the charging network they will need to avail of.

It will be interesting to see their approach. Comma.ai and Tesla have gone end-to-end ML and that seems to be the way to solve it. I fear apple is going the Waymo sensor overload and HD maps route, based on the precision mapping we see in Apple maps in places like NYC.

 

[Harry Haller]

Sure… cars are a lot of things, aren’t they? Fashion, status symbol, work vehicle, luxury, camping, recreation, escape… But the primary purpose is transportation. Get to work and pick up groceries.

There are those who think a car needs to have a gasoline engine, or whatever. That’s their wish. But the need, the fundamental purpose, is getting from A to B. If it does not do that, it is not a car. Most everything else is our personal whim.

I wonder what the long term impact of remote work and delivery services will be on future auto demand? The folks I know doing remote work love having 2 hours a day of their lives back not having to deal with commutes and road raged drivers. Some are reducing their car counts, too.

 

[omenatarhuri]

They’d be much better off buying someone like Fisker or Lucid so they don’t have to deal with all the legacy vehicle support and industrial relations.

Those companies don’t really know about automotive production at a large scale and globally. Your points are valid though as well. Apple would be smart to keep them operating as their own brand as well and do the legacy support etc and then build on top a separate line of Apple product.

 

[j26]

How thin will it be?

Will it have all-day battery?

 

[QuantumEraser]

1. Manufacturing innovation
2. Vertical integration
3. Software
4. Safety
5. AI
6. Dry process battery manufacturing (currently scaling).

Battery breakthrough reports are to be taken with a huge grain of salt. Improvements are incremental year in year but add up.

Thanks for the response. Like I said, I'm not sure I see how Tesla have a huge lead in any of these areas aside from maybe AI and battery manufacturing. Where Apple can/will gain an edge also seems hazy. Aside from "reinventing" transportation or ownership, it's hard to see how Apple make a niche for themselves in this area. Interesting to think about though....

 

[Harmonious Zen]

Thanks for the response. Like I said, I'm not sure I see how Tesla have a huge lead in any of these areas aside from maybe AI and battery manufacturing. Where Apple can/will gain an edge also seems hazy. Aside from "reinventing" transportation or ownership, it's hard to see how Apple make a niche for themselves in this area. Interesting to think about though....

You don't see how they have a huge lead in manufacturing innovation?