AI experts: true full self-driving cars could be decades away because AI is not good enough yet

[diplomat33]

Wall Street Journal has an interesting article today about self-driving cars:

Self-Driving Cars Could Be Decades Away, No Matter What Elon Musk Said

Experts aren’t sure when, if ever, we’ll have truly autonomous vehicles that can drive anywhere without help. First, AI will need to get a lot smarter.

www.wsj.com

Basically, some AI experts are arguing that AI is not good enough yet for true full self-driving cars. They point out that our best self-driving cars still need some help, like with HD maps and remote operators. So they think we will see limited self-driving cars, like we are seeing now with Waymo and others, but true full self-driving cars, that can drive anywhere with no human assistance, are still decades away. They explain that current AI is good at seeing patterns but is not good at extrapolation:

"Problems with driverless cars really materialize at that third level. Today’s deep-learning algorithms, the elite of the machine-learning variety, aren’t able to achieve knowledge-based representation of the world, says Dr. Cummings. And human engineers’ attempts to make up for this shortcoming—such as creating ultra-detailed maps to fill in blanks in sensor data—tend not to be updated frequently enough to guide a vehicle in every possible situation, such as encountering an unmapped construction site.

Machine-learning systems, which are excellent at pattern-matching, are terrible at extrapolation—transferring what they have learned from one domain into another. For example, they can identify a snowman on the side of the road as a potential pedestrian, but can’t tell that it’s actually an inanimate object that’s highly unlikely to cross the road."

Other experts, including at Waymo and Aurora, argue that you don't need to "solve AI" in order to have true full self-driving:

A growing number of experts suggest that the path to full autonomy isn’t primarily AI-based after all. Engineers have solved countless other complicated problems—including landing spacecraft on Mars—by dividing the problem into small chunks, so that clever humans can craft systems to handle each part. Raj Rajkumar, a professor of engineering at Carnegie Mellon University with a long history of working on self-driving cars, is optimistic about this path. “It’s not going to happen overnight, but I can see the light at the end of the tunnel,” he says.

This is the primary strategy Waymo has pursued to get its autonomous shuttles on the road, and as a result, “we don’t think that you need full AI to solve the driving problem,” says Mr. Fairfield.

-----------

My take: I think the article is right about the current challenges with AI and that we probably won't see true L5 in the short term. However, I think the article is probably wrong about it taking decades to "solve FSD". We tend to underestimate the speed of technological progress. Just look at the how quickly computers have evolved. I think it is very possible that we could see some big AI breakthroughs in say 5 years that help us achieve better FSD sooner than "decades". We might also find clever engineering ways to "solve FSD" without solving AI, as Rajkumar suggests. After all, we've solved a lot of tough engineering problems already without "solving AI". So I think self-driving tech will get better and better and we will see more self-driving cars on the roads in the years to come. I am optimistic that it won't take decades to "solve FSD".

I would also argue that limited L4 self-driving may be good enough for now, at least for the short term. Sure, true generalized L5 self-driving cars, with human-like intelligence, would be the holy grail, but I don't think it is necessary. After all, the goal is to achieve self-driving cars that are safe and reliable and serve a useful application like ride-hailing. Does it really matter how we achieve that goal, as long as we achieve it? If it takes some geofencing, HD maps etc to achieve the goal, so what?

 

[sperkin]

Tesla has not shown any improvements to FSD yet. Yes, the car can drive itself, but it still reacts to it's environment and can't fill in the blanks.

For example, one of my neighbor will back out of her driveway pretty fast while she also has a car parked by the curb. As a human, I can see the garage door is open so I will slow down to 10 mph and getting ready to stop. FSD will wait until her car comes into view and then slams on the brake.

Also, on a sharp curved road to left and I'm in the left lane, I can take cues from the cars on the right lane. They can see down sight slicing the angle. They slow to a stopped red light. FSD doesn't know and makes the turn until it sees stopped cars and slams on the brake. Try 55 south to Newport Beach. Not even sharp and FSD has problems there.

Braking can only get so far, but can't beat physics. The car is 5,000 lbs and takes 120+ feet to stop from 60 and 170+ feet from 70mph. It better start getting smarter and anticipate things well ahead of time.

Many other situations I can list. These are not edge cases. It's common everyday problems. People don't follow the law. Al expects all drivers are perfect.

 

[JHCCAZ]

...
Braking can only get so far, but can't beat physics. The car is 5,000 lbs and takes 120+ feet to stop from 60 and 170+ feet from 70mph. It better start getting smarter and anticipate things well ahead of time.
...

It's not only limited by how fast your car can brake, it's limited by how quickly and safely other drivers can react to your car's braking.

Lack of anticipation causes late+abrupt response, and late+abrupt response causes accidents.

 

[linux-works]

fsd or level5 will "be there" only when ALL OTHER CARS are also computer-controlled and not human-controlled.

the mix is the mess.

also, I strongly still believe v2x is the real answer to getting this usable, faster. yes, it needs infra. so, build it and we'll have another dimension, making a mesh network of cooperating cars instead of the hell we have today.

cars thinking on their own: not going to get to level5. quote me on that

 

[El joe]

Braking can only get so far, but can't beat physics. The car is 5,000 lbs and takes 120+ feet to stop from 60 and 170+ feet from 70mph. It better start getting smarter and anticipate things well ahead of time.

Which is the reason why we see Teslas using FSD hitting stationary objects like emergency vehicles stopped in the freeway. A human can/should see and prepare whereas a computer sees as far as its adjusted and tries to stop in that given distance.

 

[Dutchie]

Tesla has not shown any improvements to FSD yet.

FSD has shown many improvements. Until last year my car did not react to STOP signs or traffic lights. Now it reads and reacts to speed signs and I see great video’s on YouTube with FSD beta.

 

[orion2001]

As I've noted in other discussion threads here, I work in a related field (computer-vision, object-detection, recognition) and I couldn't agree more with Dr. Mitchell and Dr. Cummings. A good related read that is linked to in the original article is this one: How do you teach a car that a snowman won’t walk across the road? – Melanie Mitchell | Aeon Ideas

There is a HUGE gap between where we are and where we need to get for truly autonomous self-driving and as alluded to in this article, that requires huge, fundamental leaps in the field of AI / ML and there currently isn't anything even remotely on the horizon that shows any promise of solving these problems. Currently, ML systems using CNNs, RNNs, or Vision Transformers or whatever the latest and greatest tech is, can all be fundamentally reduced to being dumb building blocks that do a basic job with no "smarts" to them. Need to detect STOP signs?... train on thousands of images of STOP signs and then your network will do okay. Need to still detect it when a person is holding one up, or when it is partially occluded by a tree? Well, collect and label a whole bunch of unique instances of those scenarios, and after all that effort you have a slightly more robust STOP sign detector. Oops, now your fantastic STOP sign detector is detecting a STOP sign painted on the back of a van. What do you do now?

The answer, is not train your network not to detect stop-signs on vans. If you do that, you will also train your network not to ping on STOP signs that look very similar to the painted STOP sign in the examples you trained it on as non-stop signs. What you really want, is a "smarter" algorithm that can reason about the world. That understands the scene, that knows that the STOP sign is painted on the back of a van and isn't a "real" stop-sign. It is just another example of what Dr. Mitchell talks about in the article I linked above. Now sure, you can come up with heuristics and hacky logic to try and account for this one edge-case, but the fact of the matter is that there will always be an infinitely large long-tail of scenarios where you need something "smarter", with common-sense about how the world works to actually reason about the perception of the scene and make reasonable decisions. Currently, there isn't even a hint of the field of AI being remotely close to achieving this. Hence the DARPA program to try and train up algorithms to achieve a level of cognitive capabilities to match an 18-month old!

Waymo and others are trying to hack it by controlling things as best they can, relying on as many crutches as they can to avoid having to solve these harder problems, and I think they will and have gotten quite far with this approach, but I just don't see these ever working in a general sense for the very same reasons described above and in the articles. I stay away from the main Autonomous driving thread because some people are extremely opinionated and mostly seem to opine from a place of wishful thinking, rather than any basis in the facts of where the current state of ML and AI really is. Personally, I don't think L5 FSD will happen without huge leaps in the field of AI and ML. I don't even see any reason for optimism that these leaps could possibly happen in the next 5 years. As a practitioner in this field, I honestly have seen almost nothing interesting and of real value come out of research in the past few years. There is a lot of hype and a lot of crank turning and tweaking and performing incrementally better at some basic task on some basic dataset like COCO/Imagenet, but there has been nothing that gets us any closer to moving past training dumb as doorknobs detectors and classifiers.

Now personally, I just want a really good AP experience as an aid to me as a driver, and Tesla already has something pretty good, and I think Tesla and others can make fantastic AP aids using current sensor tech and ML/Computer-vision algorithms and capabilities. I do feel terrible for everyone who has been hoodwinked into spending money on FSD. It is just outrageous that this happened at all and while I want Tesla to succeed in the long-run, I absolutely want to see them raked over the coals for their FSD snake oil.

I think with the current tech on hand, and the current state of incremental advances, we would either need all cars to be on some sort of mesh-network for truly L5 autonomous self-driving (highly unlikely that ever happens), or we will need to make our peace with having less capable "self-driving", whether that be being restricted to well-controlled, geofenced areas, or highways, or only regions with up-to-date HD maps, or just always requiring an attentive driver at the wheel. Perhaps we will get to reliable L4 without too many restrictions, but even that seems pretty iffy to me at the moment.

 

[Tam]

The reason is the moron behind the wheel WHO WAS SOLELY RESPONSIBLE FOR OVERSEEING AP/FSD, wasn't.

What part of that don't you understand? I am so freakin' sick of this kind of bovine excrement that blames AP/FSD but ignores the fact the driver was given clear warnings that he/she is RESPONSIBLE. PERIOD.

You are completely correct but this thread is not about holding human drivers responsible but relieving humans (such as a blind man) from the driving task and let the machine take over as in a robotaxi without a human driver.

In 2013, Tesla thought that could be done 3 years away.

Then in 2015, it's 2 years away.

Then in 2016, it's 2 years away.

Then in 2017, it's 2 years away.

Then in 2018, it's by the end of the year.

Then in 2019, it's by the end of the year.

Then in 2020, it's 1 year away.

Then in 2021, "pure vision" comes to the rescue at last!

So, this thread is among many others trying to revise the expectation of Autonomous driving such as for those who paid $10,000: How soon?

I think to achieve the autonomous driving goal, Tesla is still behind:

1) Reliability in collision avoidance because "the moron behind the wheel WHO WAS SOLELY RESPONSIBLE FOR OVERSEEING AP/FSD, wasn't." Waymo doesn't blame that on the moron.

2) Once Tesla can achieve collision avoidance parity with Waymo, then both Tesla AND Waymo still have to figure out how the car would be smart enough, not just about safety but also about how to deal with new scenarios or intelligence. The "intelligence" for the machine to win a chess match is doable because there are finite moves and the machine can learn all new moves until there would be no more "new moves". With driving, it seems the "new moves" are infinite so it might be unreasonable to think that it can be solved soon, even with the advance of AI or "pure vision".

 

[DanCar]

You are completely correct
...

No, he is completely wrong. Such uncivil behavior should not be tolerated on this forum.

Back on topic:
I mostly agree today's A.I. is lacking and will continue to be lacking for the foreseeable future. Elon said that A.I. breakthrough is needed, so Elon agrees. Where I don't agree is that much can be done with remote operators and we don't absolutely need level 5, level 3-4 is great.

Since A.I. isn't good enough the winner will be the one that will tolerate the most risk of accidents. In other words the one that will tolerate accidents has big advantage over those that don't. Companies like Waymo and Cruise will not tolerate risk of accidents. Elon / Tesla has demonstrated ability to tolerate a lot of risk.

So again this points to Tesla having a big advantage 5+ years down the road for level 4. Level 3 is achievable within 3+ years in stop and go traffic on the freeway. Tesla can throw out something that almost works all the time, where as other companies won't. Waymo should have an easier time with its trucking business, if the goal is to just have it work on limited access freeways. I also see Tesla having a short term advantage in that level 2 assist has and will excite the market.

As a counter point about A.I. being dumb, here is a paper I found exciting: [2010.14439] Differentiable Open-Ended Commonsense Reasoning

 

[DrDabbles]

This is going to start a pretty heated debate, I'm sure, but "AI" flat out does not exist. There is no intelligence in so-called AI in a sense that we'd relate to ourselves or any living creature. Machine learning as a technology is hilariously brittle, easily attacked with minimal effort, and unbelievably immature for a technology that is basically from the 1960s.

I'm beyond skeptical that neural nets will be the technology that makes completely autonomous vehicles a possibility. Not only because the technology is, as I said, incredibly brittle, but also because of the insane energy required to run these massive data processing systems. The amount of data they require to begin making sensible probability calculations is also pretty foolish, and the concept of detecting new objects that a network has had no training for essentially does not exist. In fact, the absolute state of the art is to determine if an object in an image belongs to a class of objects for which training has occurred.

This single failure in the design of "intelligence" systems shows how little intelligence there actually is. You can place nearly any "intelligent" animal into a scene with something it has never seen before and that animal will immediately recognize that there's an object present. Humans will deeply understand the object and its nature without having ever seen anything like it before. Computers? Not at all. To quote from Westworld, "It doesn't look like anything to me".

 

[Carl_Drogo]

@orion2001 You no doubt have more knowledge than me, but even if you are correct that level 5 is much further away than some expect, I wonder if that’s really the right way to think about the issue. Sorry if this has been mentioned before, but is perfection (or near perfection) really necessary for robo-taxis? Isn’t the goal simply to be five or ten times better than a human (or some other arbitrary number)? That seemingly can be achieved without artificial intelligence that can extrapolate, infer, and reason. It becomes a social and political issue at some point, driven in part by media hysteria and people who don’t understand statistics. It may be a tough sell to the public to accept some injuries and deaths — even if far fewer than human drivers. But level 5 perfection and major advances in AI aren’t necessarily required for a successful outcome of self driving technology and robo-taxis.

We know that no matter how well we drive there might be some idiot on the road that will kill us, yet we still drive. We know that planes crash sometimes and yet we still fly. We know the stock market might crash for some unforeseen reason, yet we still invest. Even if AI takes decades to significantly improve self driving there will be a time long before then when self driving technology is good enough. And I think that time isn’t too far away, relatively speaking.

 

[DrDabbles]

there will be a time long before then when self driving technology is good enough

And that date is decades away.

 

[Carl_Drogo]

And that date is decades away.

Is Waymo not already working in Chandler, AZ? Will it take them decades to roll that out for wide release — or is their AZ thing just smoke and mirrors (I haven’t followed closely)?

what do you think self driving tech will look like ten years from now?

 

[JHCCAZ]

Wall Street Journal has an interesting article today about self-driving cars:

Self-Driving Cars Could Be Decades Away, No Matter What Elon Musk Said

Experts aren’t sure when, if ever, we’ll have truly autonomous vehicles that can drive anywhere without help. First, AI will need to get a lot smarter.

www.wsj.com

Basically, some AI experts are arguing that AI is not good enough yet for true full self-driving cars. They point out that our best self-driving cars still need some help, like with HD maps and remote operators. So they think we will see limited self-driving cars, like we are seeing now with Waymo and others, but true full self-driving cars, that can drive anywhere with no human assistance, are still decades away. They explain that current AI is good at seeing patterns but is not good at extrapolation:


Other experts, including at Waymo and Aurora, argue that you don't need to "solve AI" in order to have true full self-driving:



-----------

My take: I think the article is right about the current challenges with AI and that we probably won't see true L5 in the short term. However, I think the article is probably wrong about it taking decades to "solve FSD". We tend to underestimate the speed of technological progress. Just look at the how quickly computers have evolved. I think it is very possible that we could see some big AI breakthroughs in say 5 years that help us achieve better FSD sooner than "decades". We might also find clever engineering ways to "solve FSD" without solving AI, as Rajkumar suggests. After all, we've solved a lot of tough engineering problems already without "solving AI". So I think self-driving tech will get better and better and we will see more self-driving cars on the roads in the years to come. I am optimistic that it won't take decades to "solve FSD".

I would also argue that limited L4 self-driving may be good enough for now, at least for the short term. Sure, true generalized L5 self-driving cars, with human-like intelligence, would be the holy grail, but I don't think it is necessary. After all, the goal is to achieve self-driving cars that are safe and reliable and serve a useful application like ride-hailing. Does it really matter how we achieve that goal, as long as we achieve it? If it takes some geofencing, HD maps etc to achieve the goal, so what?

@linux-works responded earlier while I was starting this V2X comment - with a V2X-centric comment that I agree with in part but not completely, see below.

On the base topic of whether AI is required and realizable as the key enabler for L4/L5, I agree that trying to get AVs to act like experienced human drivers is a very high bar, but also maybe an unreasonable and unnecessary bar.

What is needed to work instead of and along with other humans, is something that does exceed human-driver capability, but realistically not expecting human or super-human reasoning/extrapolation ability. It needs to be solved another way.

Once again I will posit that a major breakthrough would be in the introduction of initially simple V2V/V2X technology. I admit that I've been thinking about this for only a limited time and haven't read up extensively. Still I can envision how it could be rolled out in a tolerably inexpensive and seamless way, without lock-step regulations, lock-step or simultaneous adoption and I think without requiring universal simultaneous deployment (which if really true is a really serious problem):

...ALL OTHER CARS are also computer-controlled and not human-controlled...

I think we could take major strides well before that, and with major benefits from the start of deployment, advantages that are realizable within just a couple of years.

The key to this would be, initially, deployment of network receivers in AVs, and progressively-improving beacons/sensors/repeaters to feed the mesh network.

Here is a possible and sensible progression, remembering that the whole point is to start small but effective, and achieve better and better results without disrupting the traditional environment, also without mandating sudden new behaviors or expense from the driving population:

1. Construction barriers/markers equipped with solar-recharged beacons, broadcasting simple GPS location and desired keep-to-left / keep-to-right / keep-away guidance. An updated version of flashers common today.
2. Status beacons added to traffic signals -some of this is developing already but I'm not really familiar with the standards.
3. Similarly, instructional beacons for unattended placement on stop signs, school-crossings etc., and operator-attended ones for use by traffic-cops, construction-site traffic directors, school crossing guards etc. Can be made very simple to use, in some cases just like the old handheld stop-sign placards but with embedded position-aware beacons.
4. Encouragement for pedestrians, especially children, to wear simple beacons on their person while outside near traffic. Pets too. Clearly the AV should have a very low expectation of harm to people without this, but simple extra precautions like this would likely reduce the likelihood by two or three orders. And remember it's not just something to "enable AVs", it's consistent with an overall desire to reduce vehicle/pedestrian accidents - it's not like we are now in good shape in this regard, before putting AVs on the road.
5. Stepping up in sophistication, beacons as above but with ability to detect and report vehicular and pedestrian traffic as well as unusual or unidentifiable obstructions/hazards, e.g. camera/lidar/sonar equipped as appropriate. These functions can also be merged into city traffic-management networks, security monitoring networks etc. (No intent to raise new privacy debates here; all these networks exist and are growing today).
6. Of course, ability of newer vehicles, having in-use AV modes or not, to contribute to the local information mesh.

The V2X mesh will naturally grow as infrastructure is upgraded, and older vehicles are naturally superseded by newer ones. The cost, including installation and upkeep, of various kinds of beacons/repeaters will become shockingly low as long as market competition is allowed under the standards umbrella,. Even for the more capable camera/sensor-laden units, costs will be very reasonable in volume.

There's nothing canonical about the order or completeness of the list above; it's basically just brainstorming targeted to add simple, quickly-deployable aids and intended to address the most challenging scenarios that can and do trip up ML.

As I see it, there's no need for disruptive mandates nor highly simultaneous and difficult-to-coordinate roll-outs. This is the key point that enables important success well before we get to universal V2X. L4 and, if you believe in it, L5-capable vehicles can exist at any stage of this progression, but will become cheaper, require less support and perform statistically better as the V2X environment grows in number and capability of vehicles and fixed beacons.

So will non-V2X cars become unwelcome, incompatible or partly illegal because of this - a common concern of skeptics? Over years of time, probably yes but I'd say not in any real oppressive way. Eventually, high-speed roads will have designated lanes for V2X vs. non-V2X traffic (sorry can't really help it if they get tagged as the "dumb lanes"!). The V2X lanes will carry higher traffic flow more safely and will reduce the need for ever more road widening, They'll start as one special lane, maybe along with HOV, and over years will take over more of the highway lanes until yes, major highways don't accommodate non-V2X - just like they don't accommodate horses or golf carts today. However I think the thrust of the planning and integration should be not to disrupt, but to coexist.

 

[diplomat33]

Is Waymo not already working in Chandler, AZ? Will it take them decades to roll that out for wide release — or is their AZ thing just smoke and mirrors (I haven’t followed closely)?

No, Waymo is not smoke and mirrors. They have self-driving cars deployed now for public use in Chandler, AZ. You can watch the videos of Waymo rides with no driver. They are real self-driving cars. The issue is that the self-driving is still limited. It is geofenced and it does require human help from time to time. So, while the cars are self-driving, they are not the kind of full self-driving cars that some people might have thought we would have by now. They are not the full self-driving that is as smart as humans and can drive anywhere, any time, with no human help at all. This is the whole point of the article in the OP I shared: we have limited self-driving now but we don't have full self-driving that works anywhere, anytime, with no human help yet.

what do you think self driving tech will look like ten years from now?

Personally, I think it will be way better than what we have now. Ten years is a long time in computer years. So I think we will see a lot of progress in autonomous driving in the next 10 years. Will we achieve L5 by then? Who knows? But I think we will see a lot of progress.

 

[linux-works]

after all is said, tried and retried, we'll eventually conclude that machines driving in chaos (mixed environ with humans) just wont be safe enough to 'take a nap in' while it drives you someplace.

I firmly believe that the train or trolly concept (following a trackless but 'reserved' path that is free of most of the randomness you get today on actual roads) is what is going to have to happen.

actual *thinking* does not happen in computers and I dont think we will even see it in our lifetimes. we just don't know enough about how *human* thoughts work, let alone how to achieve even 1/100 of that in computers. I can always think of situations that would happen in the real world that coding just can't deal with, safely.

the automation that I hear about in industry is using reserved tracks and factory floors to move things around. this is not the chaos that we see in public roads.

I just dont think its doable on the random roads that we have in the world. even in just one town, there are too many spurrious things that can happen and it takes decades of human understanding to be an effective and safe driver.

so, I'm happy with the L2 assist I get from my m3. and if it stays there and does not progress more than an 'assist tech' I'm good with that. I think that's as good as it will get until we breakthru computer 'thinking'. and thinking is not millions or billions of machine-learned rules. that bulk of info does not scale and I dont think its the answer. its one way to chip away at some problems, but I dont think AI is ever going to be *thought*. I just dont see it happening in our lifetimes, if ever.

 

[DanCar]

... Will it take them decades to roll that out for wide release ...

It will take longer than people suspect. 3+ years is very optimistic. I do expect some good news soon in SF though. There are some smoke and mirrors in that the deployment is smaller than people think. Keep in mind that it costs billions each year to fund Waymo.

 

[jebinc]

Wall Street Journal has an interesting article today about self-driving cars:

Self-Driving Cars Could Be Decades Away, No Matter What Elon Musk Said

Experts aren’t sure when, if ever, we’ll have truly autonomous vehicles that can drive anywhere without help. First, AI will need to get a lot smarter.

www.wsj.com

Basically, some AI experts are arguing that AI is not good enough yet for true full self-driving cars. They point out that our best self-driving cars still need some help, like with HD maps and remote operators. So they think we will see limited self-driving cars, like we are seeing now with Waymo and others, but true full self-driving cars, that can drive anywhere with no human assistance, are still decades away. They explain that current AI is good at seeing patterns but is not good at extrapolation:



Other experts, including at Waymo and Aurora, argue that you don't need to "solve AI" in order to have true full self-driving:



-----------

My take: I think the article is right about the current challenges with AI and that we probably won't see true L5 in the short term. However, I think the article is probably wrong about it taking decades to "solve FSD". We tend to underestimate the speed of technological progress. Just look at the how quickly computers have evolved. I think it is very possible that we could see some big AI breakthroughs in say 5 years that help us achieve better FSD sooner than "decades". We might also find clever engineering ways to "solve FSD" without solving AI, as Rajkumar suggests. After all, we've solved a lot of tough engineering problems already without "solving AI". So I think self-driving tech will get better and better and we will see more self-driving cars on the roads in the years to come. I am optimistic that it won't take decades to "solve FSD".

I would also argue that limited L4 self-driving may be good enough for now, at least for the short term. Sure, true generalized L5 self-driving cars, with human-like intelligence, would be the holy grail, but I don't think it is necessary. After all, the goal is to achieve self-driving cars that are safe and reliable and serve a useful application like ride-hailing. Does it really matter how we achieve that goal, as long as we achieve it? If it takes some geofencing, HD maps etc to achieve the goal, so what?

Thanks for the summary!

 

[aronth5]

I've posted this before but this level of FSD while not L5 or fitting any of the existing definitions would still make Tesla a boatload of money and be very helpful to many owners. I like many others don't care about robotaxi.

Lets assume for a minute that I have to:
1) pay close attention for the first minute after I set the destination. Ensures FSD hardware is working and my entered destination is good. FSD requires I acknowledge this before fully taking over.
2) the last minute when I'm arriving at my destination so I can take over if needed to deal with complex parking as needed
3) during the drive when FSD runs into an edge case like a closed road or police redirecting traffic. In none of these cases would I need to take over immediately just an alert to take over within a minute rather.

And what do I get for this?
Read, text message, browse or watch a video. The value would be enormous and waiting for Level 5 without a driver doesn't have to be the end all be all so many owners are fixated on. And I dare say they are minority. Tesla's revenue upside in this scenario is very significant and would be a major advantage at least in the near term over other car makers.

And this seems doable as well. I don't really care about L5 although I realize many people do. Like most things that are worth waiting for, good things often come incrementally.

 

[bobby g]

When Skynet is online L5 will be available but only for a short time.