Hacker News Reader

Comment by jliptzin | Hacker News Reader

Definitely within 10 years. 50 years is insane.

Just the other day my Tesla model 3 drove me home from a restaurant 10 miles away, at night, with plenty of other cars on the road, through side streets and on a highway, without intervention. Did it do weird things? Sure, but nothing that made me nervous enough to intervene. Compared to just a year ago when it was still phantom breaking on the highway and couldn’t do side streets at all. All the people saying it’s still decades away don’t realize how quickly things are progressing.

10 years is a long time and progress is not always linear. Think about what happened to the internet between 1995 and 2005.

View on HN

Progress takes many shapes. Sometimes it goes quicker than expected. Sometimes the “last 5%” ends up being completely intractable. Lots of technologies got “nearly there” (supersonic flight is a good example) but never made it across the finish line.

We haven’t even begun to address emergent behaviors. You say your Tesla does “weird things.” What happens when all the cars on the road are Teslas and they do the same “weird things” at the same time (because they have the same programming)? This is an important aspect of systems engineering that we haven’t even scratched the surface of when it comes to self driving cars.

I'm assuming you mean commercial supersonic flight? I'm pretty sure there are plenty of military supersonic planes.

If you can do 95%, that might be good enough. Obviously if you could automate the entirety of the process, that would be ideal, but consumers may simply change their behavior to accommodate the 5% difference. Uber/Lyft already changed consumer behavior in a marked way: they allowed more people to move closer to the city, not pay for a car, not pay for a parking spot, etc. In Seattle before the pandemic, it was common for young professionals that didn't own a car to take Uber/Lyft most days. They did this so frequently that yuppies were blamed for the increased traffic, even though many of them were served by Seattle's great mass transit options.

The cost, simplicity, and availability of self-driving cars will undoubtedly change the consumer.

> If you can do 95%, that might be good enough.

Self-driving cars have been at 95% since the 90's. They've been at 98% since 2012. Humans are higher than 5 nines.

It turns out that human drivers, even poor ones, have a much better driving record than self-driving cars. With this latest Waymo experiment we may finally see some significant progress.

> This is an important aspect of systems engineering that we haven’t even scratched the surface of when it comes to self driving cars.

There are definitely people in the industry who are thinking about this and working on it, but AFAIK none of them work at Tesla.

I am also worried about the adversarial nature.

1. People "hacking" cars for fun/malice. This could be anywhere from "painting on the road/sign" to wireless comms penetration attempts or nation state attacks.

2. People who OWN the cars replacing firmware/hardware for perceived benefits. IE put back the "rolling stop" feature. Or "improve" behaviour/performance. (Car modders changing headlights in a Tesla might mess with the optical only approach) Or change the default decision in "trolley problem" events.

It's even more interesting to consider the ethical engines built into self driving. If a car has no choice but to hit a man or a woman, which one should it choose? Adult or child? Two adults or one child? Should the car save a child pedestrian but risk the driver's life? This must be built into the driving self driving system or I don't see how it can be legally approved.

Okay, so you've got a private company explicitly writing down the value of human life in a big list, top to bottom. Let's pretend we fast forward the several years to decades for the public to accept this new reality. Now we realise that different locations operate under different value systems. In Saudi Arabia, one man is worth three women. You best believe they will require the car to run down the woman. So each country/location is going to require an ethics update, and it's all going to have to be public. Imagine the heat Tesla is going to take by complying with Saudi rules and explicitly programming their cars to mow down women instead of men.

I think that given that choice the car will choose "which is less likely to hurt my passenger" since the passenger is the (paying) client and not the pedestrian. Pedestrian infrastructure & defense will have to adapt to respond instead :).

I think this is the correct answer. It will be simpler and more straightforward from a legal perspective too, since it can be argued that any driver would do the same - they would try to protect themselves first, if for no other reason than the fact that humans simply don't have the ability to evaluate situations that quickly and play through a trolley problem in their head while the car is heading for a collision - you instinctively do whatever action feels right to protect yourself. So I think self driving systems will be coded with that imperative as well - protect occupants first and foremost, and they won't be calculating the "worth" of whatever or whoever they are about to hit. The extent of damage calculation will end at "there is one occupant in the driver seat, so the damage to them will be lower if I hit the obstacle with the passanger side of the car", but it won't actually know or care what the "obstacle" is.

Modern cars are already hackable - there has been several demonstrations where people connected to the car stereo, and through a chain of exploits could get access to the streering/breaking ICUs, essentially taking the car over.

Yet there has been no report of this exploit used maliciously in the wild.

> 10 years is a long time and progress is not always linear. Think about what happened to the internet between 1995 and 2005.

And then compare it to what happened to the internet between 2005 and 2022...

Progress isn't always linear but it tends towards diminishing returns more often than exponential progress.

And we're not talking simple things like "can billions of people and Moore's law result in something a bit more exciting than Facebook with a newsfeed and ads and YouTube without Flash", we're talking about progress in the form of monotonic improvement in complex intractable software to solve edge cases the dev team haven't even thought of where the failure mode involves corpses, and do it to the satisfaction of regulators across many jurisdictions.

The gap between "drive 10 miles without serious incident this time" and "consistently drives a billion miles between incidents serious enough to be fatal without any possibility of intervention" is enormous.

Do you live in a place with weather? I think we'll get there eventually, but my timeline is closer to 20 years. I just don't see AVs nailing that last 1% for true peace of mind in places like the Eastern USA with torrential downpours in South Florida or snow and icy conditions in New England.

I hope I'm wrong though, I would prefer it if my kids and other road users didn't have to drive.

I live in NY

I think about all the times during winter when my car's traction control straight up gives up as I try to accelerate through an intersection or when I need to start slowing down an extra two blocks away because the ABS goes on every time I try to stop. I have no idea how AVs will nail that stuff, it's so much harder than keeping the car between the lanes and sussing out road signs.

I think about all of the lanes in my city where line paint disappears intermittently, or where old lines/construction-era lines mess with my ability to comprehend where the lanes are, let alone an autonomous vehicle's.

And then I think about all the potholes that I dodge when I drive around town. And how 4-ton electric vehicles are going to make road wear worse, not better.

The control aspects of self driving are much easier to solve.

Here’s an autonomous drifting Delorean https://youtu.be/3x3SqeSdrAE

The physics and control of cars are quite well understood, even in extreme conditions. Further, sensors allow a car to detect changing road conditions much more quickly and adeptly than a typical human.

The hard part of self driving seems to be understanding and predicting the humans in its environment.

That example is still a car on a known good road conditions. Do you have an example where half the parking lot is covered in ice, or trying to break on a road with unpredictable traction so that they don't end up halfway through the intersection?

> Do you have an example where half the parking lot is covered in ice, or trying to break on a road with unpredictable traction so that they don't end up halfway through the intersection?

If the car is "smart," maybe it will simply refuse to travel (because the calculated risk of traveling over an icy parking lot is above some maximum threshold risk).

Humans have no trouble in these conditions (after a brief training period).

What happens the first time a steering wheel free car strands a family in a blizzard due to a few feet of icy pavement, ultimately freezing them to death?

I don't think cars will ever remove steering wheels. That was just Elon's fantasy for a limited use (driving through Boring tunnels). Similarly how you don't have a steering wheel in a driverless metro. Or perhaps it will be retractable when not in use, like a tow hook.

That would make commuting in northern areas very interesting.

You're wrong. It will happen way before 20 years.

And computers are better at reacting to unexpected things like icy road. And cameras are better at seeing through rain and fog. What's barely perceptible to you, computer sees clearly.

>>You're wrong. It will happen way before 20 years

Other than random people saying this on the internet, I have seen zero indication that this will be the case. You cannot will something into existence just by strongly believing in it.

Are you just saying things that you think are true? Have you ever tried to do even rudimentary image processing in foggy / rainy / nighttime conditions? Cameras are absolutely not better at seeing through rain and fog.

Humans aren't that good at seeing in rain and fog either (which is why many people avoid driving or drive much more slowly in extreme rain and fog conditions), so saying robotic vehicles aren't good at driving in inclement weather conditions isn't very interesting. In addition to cameras, Waymo and Cruise vehicles have multiple lidar and radar units. These are all affected differently by inclement weather but in principle combining sensor data for perception gives the vehicle a lot more data to work off of than a human relying solely on vision.

Driving in truly bad weather is still an open research problem, but it's also one that most of these companies haven't dedicated a lot of resources to, since they all have initial rollout plans in places that don't experience much extreme weather. But assuming these vehicles can drive well in good weather, I don't think it's a stretch to assume that progress will be made on driving in more extreme conditions once more resources are dedicated to it.

Apart from weather, a big issue outside the USA is also car infrastructure. The US is extremely car centric with large, straight, easy-to-maneuver roads everywhere. Here in Germany, it's different. We have rather narrow, chaotic roads, unclear signage, "right car has right of way" traffic rules which sometimes get resolved via hand signs, no jaywalking laws, etc.

I am still hoping for fewer cars on the road overall. The car itself is inefficient and hopefully on its way out.

NE cities are more similar to european cities than most of the USA. I have no doubt cars can handle european cities if they can handle NE cities.

Isn't that like saying that motorcycles will happen eventually but not anytime soon because: weather?

Traditional cars with drivers haven't solved the weather problem 100%. But they are here now. Sometimes we ban them in some conditions, but they are a game changer the rest if the time.

One thing that is going to be awkward to navigate in the transition to self-driving cars is how they handle driving in bad weather. Obviously that's a technical problem but I'm not sure it's actually solvable because it's not actually solvable for human drivers. We just kind of do it anyway. But when a self-driving car has some quantified sense that "driving in these conditions is unsafe" will we get to the point where we just say f it and let them keep going? Or, maybe more likely, we remain in the gray zone where the car is fully self-driving but the human driver has to be ready to take control at any time which I think limits some of the larger impacts.

I mean if you live anywhere with weather your motorcycle lives in the garage for 3/4 of the year. Not useless by any stretch but also not “taking over the world.”

It’s not that traditional cars have solved for weather, it’s that self driving cars haven’t solved the problem of “being as good as human drivers in weather commonly encountered.”

Motorcycles are only on road for maybe 5 or so months out of the year where I live. If that is the best we can ask for out of autonomous driving, it'll never be more than a luxury toy.

My Tesla still phantom brakes. And Tesla has been trying to solve it for, what 10 years ? My rain sensor also barely works. I have seen enough FSD videos and all the scary things it does. There is no way it will be ready to solve the last 1% in the next 10 years

I rode with my uncle in his new Tesla car in rural Pennsylvania this winter. Clear weather, daylight, only a few other cars on the road. In a 10 mile trip the automatic emergency breaking threatened to kick in with blaring alarms 5-6 times: whenever an oncoming car approached us in the other lane, and over hills where it apparently thought we were driving off a cliff. I thought it would do better than that, given how much it's touted as so good.

Tesla fans remind me of Linux fans. "It can definitely 100% do this thing... if you don't mind x, y, and z. Also you need to do a. And b. Some sometimes c if a does j. Also it doesn't work on Tuesday. But that doesn't matter because you shouldn't need to use your car on Tuesday anyway."

I dig the commitment and passion, but as a consumer, I want the thing to work perfectly first time, every time. If it doesn't, I don't want it.

Did it do weird things? Sure, but nothing that made me nervous enough to intervene.

Doesn't sound like much of a data trail. No one disputes that it at least sort works (while also doing "weird things", as you put it) in some conditions. As long as, you know, nothing weird happens out on the road. Like for example rain.

It's the fat tail of "weird things" (and how these systems currently handle them) that leave the question of whether FSD will ever arrive -- very much up on the air.

Wide scale adoption is going to bring a ton of emergent consequences that you might never see in a million man-hours driving a single car.

Totally agree. People become SOOO embrace full self driving so quickly.

I know 5 people with the Tesla FSD Beta, and EVERY SINGLE ONE of them has admitted to letting Tesla drive them home when they drank too much and went past the legal limit. And not a single one of those 5 people has ever had a DUI. So FSD rapidly changed there drinking behavior.

Also I get that I have an atypical and shitty social circle. I don't think this represents all Tesla drivers.

Yeah consider me unconvinced that someone willing to do tesla fsd drunk wasn't also willing to drive drunk themselves. Lack of a DUI is hardly proof otherwise.

I know it's off-topic but your username reminded me of Bonzi Buddy. Is that the inspiration for your name?

Does anyone else remember that spyware?

Yup :)

Yes, and somehow this immediately triggered my memory of the CueCat from the same era.

https://en.wikipedia.org/wiki/CueCat

I'd say closer to 50 years myself. But my standard means that we not only have near-perfect self-driving, but we're confident enough in its quality that we're willing to manufacture cars without conventional physical driver controls, and people are willing to ride in them. And then to actually go through the process of replacing the entire national inventory of conventional cars with these new cars without controls.

No way. At least 50 years to get somewhere. But we need computing power 1000 times better and optics at least 16 times better. Before that it never can fix all scenarios

Here is a video of a customer vehicle driving on snow covered roads, now. It isn't perfect, but it is leagues better than a year ago. https://www.youtube.com/watch?v=oDK3dRHxOzo He has tons of other drives in various weather/road conditions.

Improved software, and maybe one hardware refresh should be sufficient to get us all the way there for 99.99% of drives.

It appears that if it doesn't know where the lanes are, it drives in the middle of the road. If someone turned onto the street from 0:35 to 0:42 there would have been a problem.

Aside: some really old drive assist tech for snow plows on I-80 Donner pass - https://path.berkeley.edu/sites/default/files/advanced_snowp... /// https://highways.dot.gov/public-roads/janfeb-2001/safe-plowi...

I'm still waiting to see how an AI would handle Figure 5-1 from that. And no, the snow plow's solution of burying magnets in the road isn't practical (Cost of infrastructure installation for the test sites is approximately $11,000 per kilometer ($17,000 per mile), including surveying, installation, and magnets).

99.99% ? How does that statement work? I feel it needs more clarification.

99.99% means 1 in 10,000 "drives" has an accident of some sort? So 100 accidents per million people/drives?

Not sure how to compare if that is better or worse than what we have now.

Reply on news.ycombinator.com