Dude, where's my self-driving car? | Hacker News Reader

Dude, where's my self-driving car?

theverge.com

19 pointsmfiguiere34 comments

I just bought a 2022 Kia Telluride and while taking it on our first long road trip, I discovered that it had full-on adaptive cruise control (embarrassed to admit I didn’t even know this while purchasing it). What this translates to is “make minor steering adjustments to stay in your lane, try to cruise at speed X but if there’s a car in front of you going slower, follow them at distance Y”. Intuitive steering wheel controls to adjust X and Y.

As someone who has never driven a self-driving car, I was amazed. I thought this was something only in Teslas and high-end Mercedes. Now I’m discovering it’s basically becoming standard at a certain level of medium luxury.

I do not need, nor do I want more than this. It enables me to take my eyes off the road for enough time to rest my eyes, stretch my back, feet, change the tunes, grab something from the back seat, then settle back in. This is world changing especially when driving at night - I would bet it close to doubles my tolerance for continuous driving. I’ll bet it’s great in stop and go traffic as well, if you don’t care about being cut off by a ton of people.

Given my (pretty in-depth) level of understanding of current state of sensors, control algorithms and AI, this is an activity I trust a car manufacturer to reliably implement. I also like driving. I’ll check back in ten or so years to see how the whole self driving thing is coming along.

What you have in the Kia is a level 2 autonomous driving [1]. This feature is called a Advanced Driver Assistance Systems (ADAS) is now even available inside non luxury cars and vehicles, and I think one of nicest features is able to perform the ADAS in traffic jam thus the vehicle can perform stop and go function if the vehicles have electronic parking brake rather than the manual [1].

Level 5 will be equivalent to AGI, and most people think that it is most probably not achievable in the near future.

[1]The 6 Levels of Vehicle Autonomy Explained:

https://www.synopsys.com/automotive/autonomous-driving-level...

Same, I was quite happy with the adaptive cruise control on my Tesla Model 3. I decided to try enabling FSD beta, and it's basically a disaster. I've experienced it:

* Get confused about its lane position and swerve around in the lane

* Attempt to get into the left lane despite a right turn coming up in a few seconds

* Try to change lanes into a lane that doesn't exist, off the side of the road

Once a month or two, I decide to give it another fair chance, and almost every time, it seems to want to do something stupid and/or dangerous.

> It enables me to take my eyes off the road for enough time to rest my eyes, stretch my back, feet, change the tunes, grab something from the back seat

Oh for your and the sake of others, please don't do this.

I have a similar system in my car, works great 99% of the time. But that 1%, where it went wrong, could have resulted in crashes had i taken my eyes or my hands or my feet or my mind off the driving aspect.

There is no room for error here. Just stop on the side of the road if its too tiring. Focus more if you're late. Or just take an Uber or have someone else drive. Or better yet, take public transport. If none of these are options, just suck it up and drive.

Please don't put yourself, or others in danger relying on this bullshit. When things go wrong, only one person is at fault and pays the price, it's you. Not Kia, not Tesla, just you, your family, the people in the crash and their family. Not worth it really.

I’m not sure your risk analysis is correct, if the driver is going to do those activities anyway, without the aid of self driving tech.

People already do things like grab food from the backseat, fiddle with the music, read their texts etc.. while driving. Are they less of a risk to those around them if they have some self driving guard rails? I would argue very likely yes.

Your conclusion of “you’re bad if you do this, so don’t do it” is rarely going to change human behavior. Giving people a tool to make the behavior they already want to do much safer is more likely to succeed.

My knee jerk reaction was to point out the flaws in your argument, but I figure thats not constructive, so here goes.

My first point is that taking eyes off the road while driving is risky business whether in a 2024 FSD Tesla or a 1996 Toyota tacoma. It's clear we don't want that.

My second point is the risk is entirely on the driver, and the people involved in the crash, not the producer of the technology who absolves themselves of all liability from the use and effects of said technology.

My personal experience that formed my opinion on this is that the technology is not ready to be fully trusted, not to the extent of taking eyes off the road (which really brought out my instinctive comment).

I will agree that this does have potential to avoid some crashes. But human nature would make us sloppier drivers.

So what, in your view, is the value of adaptive cruise control? Why did you purchase it? If I have to keep my eyes glued to the road, hands on the wheel and foot over the pedals 100% of the time, what’s the point? I’d rather just drive.

To be clear I’m not saying I shut my eyes and nap - “rest” in this case meant “look at something other than the road / car in front of me / taillights”. I’m doing things that, as another commenter said, I’m already going to do anyway, but now it’s safer to take a momentary break to do them without resorting to “knee-steering” or other hacks. I’m also doing these things only when it’s a straight section of open road with little traffic.

Wait until you realize Tesla is behind on self driving compared to the major manufacturers.

I drove a Volvo rental that was a couple of years old that had level 3 self driving that vastly outperformed any Tesla self driving. And that wasn't even the latest tech from Volvo.

Tesla are behind on self driving and they charge extra for it!

Volvo has never shipped level 3 autonomous driving. At most they are level 2.

https://www.sae.org/blog/sae-j3016-update

The current market leader in terms of shipping products that are available for consumers to purchase is probably Mercedes-Benz Drive Pilot. They are legitimately at level 3 and significantly ahead of Tesla in most (but not all) respects.

https://www.caranddriver.com/reviews/a45326503/mercedes-benz...

Wow, since Tesla is so behind I guess my Volvo must be able to drive me through LA? https://youtu.be/0kvK_20mNuo?si=ooQj7S98GyCMAJEI

No? No unprotected lefts, no roundabouts? Not even turning? Sounds very advanced, Tesla must be in a panic they're so far behind...

This is the same type of thinking that prevents most developers from working in safety-critical fields. The presence of a feature, or even its impressive capabilities, is/are not what constitutes "performance." It's also the same type of thinking that gets Tesla drivers decapitated or slammed into an abutment.

That would have an acceptable point to make if FSD was not capable of safely doing what all other ADAS systems can do, but it obviously can. It's just that Tesla has no interest in restricting its features and get L3 approval for such a small subset of driving.

If today you claim FSD V12 is "behind" the lane keeping assist of other manufacturers, that is just delusion, I don't know what to tell you.

I'm truthfully not understanding whether you've missed the point or if you're so interested in defending Tesla that you choose to look past it.

I'm not making a claim about relative feature capabilities (yes, having replied in this specific thread makes that confusing, but that's where your comment was). I'm making a claim that the vast majority of developers do not know how to judge a system's safety.

As an example of this from you, and for a more critical view of Tesla from me, I'll say that when I look at Tesla's capabilities, I could not make a claim that it does anything safely. Safety implies that the system knows its limits and refuses to operate outside of them. This is definitely not Tesla's way. Else it would be impossible, for example, to have sun blind spots make it not "see" a semi.

https://www.youtube.com/shorts/dhxK4OKHhOw

My point stands: just because Tesla chooses to make FSD an L2 system at the moment doesn't mean that it's unable to act safely in the same scenarios as for instance Mercedes L3. So concluding that FSD is behind Mercedes is delusional. If you actually look at evidence available, (the thousands of videos available for FSD, and the ... tens available for other brands), and still claim FSD is behind, that is delusional.

I'm defending Tesla because these comments are just insanity, and it points to an extreme anti-Tesla or Elon bias, which is very common these days. If you can't impartially evaluate a piece of technology that is clearly more advanced in every possible way, you really need to reflect.

I don't quite understand how anyone could actually compare systems from different vendors. The failure modes aren't something the vendors have been able to identify in advance of a cataclysmic event.

So I agree with your "delusional" statement, and also consider your remarks of the same caliber. You cannot determine a system's safety by watching videos.

Think how many videos there would be by now of the 737max taking off, flying, and landing if airplanes were as hip and new as robotic drivers.

Riding a Waymo is a delightful experience, zero complaints. It might be happening slower than expected, but it's happening.

What happened was that VCs and marketers took over the conversation and the researchers and engineers who were shouting from day 1 about how unrealistic all the "self driving cars by 2020" claims were were simply ignored by the general public. Even today I'd wager that >50% of the people actually working on the tech don't think level 5 autonomy is reachable in our lifetime.

But what concretely do they mean when they think that? I haven't seen many people argue that some law of physics prohibits level 5 autonomy, or that there's some known plan which will take 100 years to execute. It seems to me that, when someone says level 5 autonomy isn't reachable in our lifetime, they're thinking something along the lines of "it's a really hard problem and I don't currently see a way to make progress".

I don't understand how you write this paragraph

> Sure, there are some self-driving cars on the roads today. They’re in San Francisco, Phoenix, Los Angeles, Hamburg, and Beijing, among other cities. They’re operated by some of the biggest, most well-capitalized companies in the world. You can even ride in some of them.

without immediately reconsidering the thesis.

The cars work ONLY in areas which have been specifically mapped out, to the smallest detail. In that way, they're very much like a vehicle following a programmed track.

A car that can drive in an unfamiliar situation like a human isn't possible yet.

It seems to me that going from "self-driving cars, what a wild idea" to "cars routinely drive themselves through city centers" in under a decade is a massive success story for the technology, even if they still have limitations that early boosters were hoping wouldn't be an issue. I'm not being coy here - I honestly do not understand how the author can look at the facts he's reported and believe that "the imminent arrival of safe, reliable self-driving technology" didn't happen.

It's because it was promised, a long time ago, and still has not arrived.

If we hadn't been given the expectation of fully self driven cars, we wouldn't have come to expect it.

My pet theory is that people simply don't really want self-driving cars.

Maybe we just like driving ourselves, even if by some definition it is "irrational"?

I think it’s irrational but from a different perspective: urban and suburban driving sucks - you’re paying a lot of money for a metal box you use to sit in traffic unable to do anything else, and do that for a fair fraction of your free time every day. Due to the 20th century decision to subsidize car travel at the expense of everything else, most people in at least North America don’t have a great alternative.

That means there’s a HUGE market for the dream of having your own personal chauffeur because that means getting a non-trivial chunk of their waking hours back to play games, watch video, read, catch up on work, use social media, etc. I remember one of the researchers at Google mentioning testing a prototype on the trip out to Tahoe because not being awake meant you could do a ski trip over a normal two day weekend.

That also means that a lot of people really want to be selling this product: millions of customers and they’re already used to paying thousands of dollars a year and taking out loans to get cool toys. That’s really going to skew judgement towards saying you’ll have it soon, even if you aren’t willing to cross the line into full Musk-level fraud.

There are already thousands of people who do ski trips from the SF Bay Area to Lake Tahoe over a normal two day weekend. It's kind of a grind due to heavy traffic and mountain road conditions but generally not too bad. And if you don't want to drive yourself then there are already scheduled charter buses that run every weekend.

The “kind of a grind” was the point: he valued being able to have a nice dinner, leave in the evening, and sleeping in the car to having to stressfully watch traffic. I think there are a LOT of people who’d pay for that.

> have a nice dinner, leave in the evening, and sleeping

Ah, like a train!

SF Bay Area residents can take an Amtrak train from Emeryville to Truckee. But the times aren't convenient for weekend ski trips, and it's a hassle to get from the train station to any ski resort. Only a real masochist would attempt that with children plus luggage.

The ski buses are going to be a better option for most people who don't want to drive because they go direct to the resorts.

https://shop.sportsbasement.com/collections/sports-basement-...

Yes. A lot of this dream is wanting that convenience without having to share space with strangers.

I don't particularly like driving myself and I really want a level 4+ autonomous car but like most consumers my wants are tempered by my limited budget. If I could get real self-driving as an option for about $3000 per year (amortized purchase price plus annual subscription) then I would take it, but if it's $10000 per year then forget it. For most people, driving just isn't a huge burden and they're already financially stretched just paying for regular cars that don't have a bunch of expensive sensors.

Sure. I don't think it's irrational, either.

But from time to time I really wish I could just take a nap and not have to pay attention on long trips.

Yes, I think from time to time that would be nice.

I don't have to drive long distances much, and if I do, it's usually for pleasure, not work, so I'm usually happy to be driving.

For short distances I think I'd find it frustrating to not be the one driving. I get bored and fidgety as a passenger.

(I like riding trains, though. I think being able to walk around makes me feel more comfortable about not being in control, as at least I can stand up when I want.)

I would argue it's like home office before the pandemic: they just don't know they want it, but once it will be there — oh they'll know.

Dude, where's my efficient public transport?

Reply on news.ycombinator.com