Cars Are Death Machines. Self-Driving Tech Won’t Change That | Hacker News Reader

Cars Are Death Machines. Self-Driving Tech Won’t Change That

nytimes.com

39 pointshelloandyhihi70 comments

The title of this article is unsupported by its contents. The only mention of self-driving cars in the article outside of the title is in this paragraph:

> There are many who say that autonomous or smart cars will solve this. So far, I’m unpersuaded. Since 2014, over $80 billion dollars has been spent on “smart” or connected cars (more on these in a minute) and autonomous vehicles, arguably to make cars safer. But investing in the car of the future is investing in the wrong problem.

So the only evidence the author presents that "Self-Driving Tech Won’t Change That" is the amount of money that's been spent on the problem so far? I don't see how that proves anything.

The argument is simpler, they're saying they don't care how much progress has been made because they want cars to go away, not get safer.

Then perhaps the headline should have been "Cars Are Death Machines. We Should Get Rid Of Them."

That would have communicated the authors intent much more clearly, without making unsubstantiated claims about the effectiveness of future technological advancements.

A vacuous flaunting of luxury opinions diametrically opposed to personal freedom, that serve to elevate their proponent who, of course, would never consider following their own advice? In the NYT? Surely not.

It seems somewhat silly to only talk about the negative aspects of cars. What about all the benefits?

If you stop and think about it, cars are truly revolutionary. Even many low income people can buy a machine that will travel 50 miles (at 70 mph) on a gallon of liquid that costs $3.

It not only allows individuals the freedom to travel, vehicles power a massive transportation network that brings food to our tables and a ton of consumer goods to our local stores.

I would argue the automobile is one of the greatest technological inventions of the past 200 years.

It appeared that our man-made environment expanded to fit the requirement and capabilities of the car rather than the other way around.

Our ability to travel allowed us to do that though. Instead of living in crowded slums the poor can afford to get out a little bit, it might not be much but I see poor people with a garden in their own yard - a yard that before the car the almost rich would have instead.

That's not a hard argument to make, at all. The world pre and post car is a completely different place.

That said, the step towards Self Driving Cars will not be even close to the same level of revolutionary, and so discussing the negative aspects is important here so we don't have an 'ends justify the means' situation on our hands.

I feel like this goes without saying. I doubt there are more than a few people would argue with anything you wrote above.

The trends are basically flat to declining long term for deaths, alongside significant growth in total miles driven - indicating things are consistently getting safer for everyone.

So the author is clearly cherry picking data with anecdotes.

The author doesn't even attempt to have a measured argument here and I think this is an example of how people are thinking "both extremes" need be heard but just assuming the "cars are revolutionary" is the extreme on the other side.

>just assuming the "cars are revolutionary" is the extreme on the other side.

You don't need to assume though because you can look at history. Regardless of whether you think it is good or bad cars changed how we live in a large way. I think they fit the definition of revolutionary.

Yeah, we've celebrated that for the past 50 years or so. It's time to move on.

The bicycle is a far greater invention. Cars are not sustainable.

Yea, but you're not going to get a book shelf to your house using a bike. Or a lot of groceries for the family. If your 8-year breaks his arm, a car of some sort makes that easier to get to the hospital instead of waiting for an ambulance. Oh, just a big pack of toilet paper. That's kind of awkward on a bike. Kind of sucks when it rains or snows. Then you also have to carry around/use more deodorant per month so you don't smell bad at the office.

And food is optimized solely to be easy to transport and look good on a store shelf after a long ride rather than be healthy.

And burning that liquid is one of the top contributors of global warming. For a good while we were adding lead to that liquid in the name of a better burn, despite knowing for much longer than that that lead is terrible for organic creatures like ourselves.

And restructuring cities around cars has ripped out highly functional, happy neighborhoods, and brought us the suburb, with its vast expanses of identical houses miles and miles away from anything even remotely interesting, and it’s immense hostility to someone who doesn’t have a car (and it’s denizens’ immense hostility to anyone spending their tax money on improving public transit).

You've painted with a pretty broad brush here; a lot of this isn't true. I suppose "remotely interesting" is a matter of opinion, but I live in a suburb that has its own "downtwon" and other amenities, to include a performance art theatre, beautiful plaza with ping pong tables, big screen TV, outdoor pool tables, etc. It's really pretty awesome if you have kids. It's also a subrub that isn't hostile to people who don't own cars. It's entirely practical to get around by bicycle.

I guess the bottom line is: the hyperbole doesn't really serve your argument.

What you find interesting (some stuff for kids and some shops in a purpose built suburban plaza / mall) is likely not going to be found interesting by someone who lives in an actual city.

I wouldn't consider what they said to be total hyperbole. It seems that way to you, because you're happy where you are, but that isn't for everyone.

Also, by suburban standards, yours sounds like an outlier on the good side, and I still wouldn't live there. Most people don't have a nice plaza and so on thoughtfully built for their mass housing.

I would argue that the automobile is a brilliantly designed tool for making flexible, medium-long distance trips quickly and cheaply. In America, this tool has been stretched far beyond the use cases where it shines. We've tried to make it "the only mode you'll ever need" for almost all Americans, deploying it where transit, cycling, and walking would be far more appropriate. It is great that many low-income people can technically afford to own a vehicle; it is not great that the layout of our cities _forces_ them to own and maintain one.

Also neglecting long distance transit. Passenger trains are pretty much not a thing in America, right?

Passenger trains suck at long distance transit. Trains work well for trips up to around 500 miles where they are faster than planes after factoring arriving at the airport early, after that trains are more expensive and slower than airplanes, without allowing the flexibility of a car. Note that if you live in Europe you can cover several countries in this distance and thus feel well traveled without ever going far from home on a continental (much less global) scale.

A full car is cost competitive with planes and trains, and gives you flexibility to stop and see things along the way.

Not really, the distances are too great to make them economical. That's why air travel is more common.

For distances where trains make sense, cars are preferable: You can travel at your own schedule, and then you have transportation at your destination. If you take a train, you have to meet the train at a certain place at a certain time, which is stressful, and then you still have to arrange transportation from the train station to your final destination.

Even worse, if you have to change trains during your trip... It's much, much easier to put all your stuff in a car once.

This basically leaves trains for traveling between dense cities, where the passengers either live close to a train station, or have a destination close to a train station.

And, to be honest, after flying twice with young kids this year, I put my foot down and told my wife road trips only for the next few years. I wouldn't want to travel by train with young children for the same reason that travel by air with young children is difficult.

On the other hand, you have to actually drive the car and it has much less space so it's less comfortable.

Maybe if you fly first class. Coach is far less comfortable than a car seat.

Traveling by train with not tiny kids is pleasant.

I think the main problem is that cars can be a great last mile transport but a horribly scaling transport for common long distance travel (commuting to city cores, etc).

Cars are great for long distance travel. They do well for medium distances in some situations and are terrible for last mile.

When my family wants to go to grandma's (or any other vacation destination) our car is the best way to do it. The key here is a family - we fill our car so per passenger we use less fuel (carbon), the train or bus could do better but because the route isn't direct it isn't that great of a win - and they need to be full to compete which they might not be.

For medium distances (less than 50 miles) they do great if and only if there are few other people going the same way. Which is to say rural and non-dense areas. However because they are such low occupancy in denser areas the congestion and storage needs when you get there get you.

For last mile the storage needs at both ends means that even though they seem great they just spread the places you want to get to farther than if you didn't have a car and just walked. In suburban areas we ignore this because there is no place within walking distance you would want to get to anyway.

Note that the key to long distances is far enough that you wouldn't do this normally. If everybody's grandma (or vacation spot...) lived in the same neighborhood as mine this would fail - but then the bus/train would be a direct route and so it would be practical.

There was nothing in the article to support the notion that self-driving cars won't address the danger of cars. Sure, as long as there are human drivers there will be accidents caused by human drivers (and obviously there will be human drivers for a long time, just as horse drawn vehicles remained around for a long time after automobiles arrived [1]), but if the only argument related to self driving cars is that they won't replace human drivers 100%, that is a weak argument.

The author completely avoids taking on the fact that self driving cars will eliminate -- for that vehicle -- the major cause of traffic deaths, human error, and the hypothesis that they will not significantly replace it with machine error. Obviously some people argue they will never get that technology right, and while I disagree with that, I'd at least appreciate it if the author even addressed it.

I'm all for getting rid of most cars, but I think the most realistic path to that is driverless taxis and small busses (i.e. vans) that are economical and efficient. The latter can be highly efficient if enough people use them, especially if they reward people with a few cents discount for scheduling a few minutes or a few hours ahead of time, and for using them at least busy times of day, and for being willing to have a couple quick stops along the way.

You'd need far fewer of these vehicles total, since many would have multiple passengers at once, they would be in near constant use, and they aren't wasting driving time and valuable space by the need to park near their endpoints.

1. and remain, especially in Amish country.

Small cars and vans don't need signalling to be good. The reason is geometric: someone already in the car doesn't want to detour to your front door to pick you up (they will excuse this if you are obviously handicapped and thus need special treatment).

All transit makes the most sense when they run a regular fixed route - just like large buses and trains run. The advantage of self driving is they run every 5 minutes or less 24/7 so you don't have to check a schedule to find out when the next bus comes. The small is an advantage only in that it is cheaper to buy and uses less fuel on routes that that don't have enough demand to fill a larger vehicle on the same route. The advantage of self driving is cost of labor (drivers) is a large reason buses don't run these routes already.

Bottom line: if your neighborhood currently doesn't have the ability to support transit today having a personal self driving car is going to make more sense.

Do not confuse ability to support with currently supports: even in the non-dense suburbs if everybody took transit instead of their car there is enough density to support every 10 minute service with stops every mile down a grid, at less cost than everybody's car currently costs. Most suburbs do not have the type of grid that would support this though. Even if they did the transit system needs to be large and frequent enough for years before people will make the switch - years where transit is more expensive than just buying all the riders a used car (7 years old, free replacement every 5 years)

They don't necessarily have to detour to your front door, you can often walk to a pickup point that is a compromise: balancing amount of detouring, amount of walking, amount of waiting, amount of adapting your schedule to the schedule of the transportation. The more flexible you are on any of those things, the more you can save yourself a few cents.

The worst case is, you need the car as soon as possible, you want it to come to your door, you don't want to be slowed down at all by stops along the way. Maybe you have other requirements, like you need it to have room for cargo, you need it to have a seat for your kid, you need to take your Golden Retriever along, or whatever. In those cases, it'll cost you a bit more, but it's still the same system so it's not a big deal.

On the other hand, maybe you are good with scheduling your trip in advance, you don't mind it making lots of stops because you'll just be working on your laptop anyway, and if the weather is nice you are good with a bit of a walk on either end. The more people who are this flexible, the better the system works, so they are happy to give you a discount.

What I'm talking about is about is somewhere between a city bus, and an Uber/cab. Lots of efficiencies can be gained by balancing all these different things, and the more people using it, the easier it gets to do this sort of thing.

I can't see why people wouldn't want it to work like that.

Not by itself; but if the high-speed crash rate drops to almost nothing we can make cars much lighter and lower; reverse the SUV arms-race that has made the streets more deadly than ever.

> if the high-speed crash rate drops to almost nothing

The high-speed crash rate is almost nothing - statistically speaking.

Most accidents are on city streets at relatively low speeds.

What more, statistically speaking for number of miles driven, we already have a near-zero rate of fatalities. Can it be better? Sure, but it will never be entirely zero, even with full automation for all road vehicles.

> reverse the SUV arms-race

The SUV fascination is not driven by some desire to have a safer vehicle either...

.. is not "solely" driven by the desire for a safer vehicle. That does play some part in the decision for an SUV in my experience

>The SUV fascination is not driven by some desire to have a safer vehicle either...

That is absolutely part of it. It's only one reason among many (style, higher sitting position, emissions regulations, etc) but I often hear people make comments about smaller cars being unsafe, or push others to buy an SUV when they have kids "for their safety."

I have absolutely recommended someone drive a bigger vehicle because I've seen what a smaller one looks like in a crash. I would still make that recommendation today. (Brutal honesty: If I'm talking to someone about what car they're safer driving in, I am going to suggest what keeps them safer than what keeps everyone else around them safer.)

Walking outside can result in deaths. Should we ban that too? What a silly premise.

Obviously cars confer a lot of benefits. They are fast, and in areas that are not high-density, they save time relative to ANY alternative (walking, biking, buses, trains) and therefore drastically improve your quality of life. They don't require you to wait on someone else's schedule, especially given the often inconsistent timing of buses. They don't require you to risk sitting down on dirty seats (6-year-old girl stabbed by uncapped needle on bus: https://metro.co.uk/2019/06/08/girl-6-injected-needle-hidden...). They don't require you to risk personal injury (40 to 60 teens rob train: https://www.sfgate.com/crime/article/BART-takeover-robbery-5...). They let you travel all over, including away from cities, and therefore give you a greater degree of freedom than the reach of fixed rails or transit systems that are limited to cities.

Yes cars can cause deaths. So do a lot of other things. But the rate of vehicle-caused injuries/deaths has been dropping sharply since the 80s (see CDC data). It will continue to drop as backup cameras, lane departure prevention, collision detection, collision avoidance, and other assistance features become ubiquitous or required. Nothing is ever perfect, and this is the fundamental reason why efforts like Vision Zero are flawed. Zero deaths is not the goal, and generally any such absolute goal is unrealistic.

We get massive benefits from fast, private point-to-point transport. Let's not forget that and work to retain those benefits, instead of damaging the benefits of vehicles with low speed limits and other 'road diet' suggestions.

One thing I found very useful when driving a car was buying a clip-on wide-angle mirror for $15. It made it much easier to have constant awareness of what was on the right-hand (passenger) side of my vehicle.

I’d urge you to get it if you drive in a city with cyclists and worry about killing another human.

I wish there were realistic car-driving simulators.

Arcade games don’t handle like a car and they don’t require 360 spatial awareness like a car. If you could walk into a mall, swipe your card, and rent time in a realistic driving sim, then:

* 14-year-olds could practice before they got their learner permits.

* People moving north could practice driving on black ice and performing the right maneuvers.

* Self-driving car companies could gather anonymized data about how real drivers act.

* Insurance companies could offer discounts to people who could pass a cheaply-administered test every 5 years.

* Immersive games could be written based on trying to rapidly deliver taxi passengers around a city, or trying to escape police in a stolen auto.

* People of all ages could regularly test their driving reflexes.

* People with anxiety could practice various scenarios in a low-stakes environment.

There are, just not necessarily affordable: https://trl.co.uk/driving-simulator

The current UK driving test has some sort of hazard simulation test, overlaid on real world footage. I passed long before this was introduced so know little of how involving it is.

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

You know, I've been nodding along to conversations about the dangers of vehicles for a long time but I think I just had a change of heart.

Travel has always been dangerous, and the farther you go the more danger there is of never coming home. We've done a lot culturally and technically to improve that situation. Yes, airplanes are safer, but would they still be if every trip was on an airplane? It's an apples and crabapples comparison. If we had 100 times as many airplanes, pilots, mechanics, etc, I really think it'd be the same shitshow.

That's the realistic standpoint on the situation. Things are "good", mind the quotes, with trains, buses and airplanes because their volume is soooooooo much lower compared to car travel. Once you ramp up their volume, problems end up being logarithmic, not linear. For example, it was either Sweden or Norway where they tolled a major highway because it was just a giant parking lot of a traffic jam. All the toll did was lower the amount of cars by ~10%-15%. Traffic jams disappeared. Back when there were a handful of cars on the road in the early 1900s, no need for speed limits or lights and there were few to no problems. Then the volume increased and the likelihood of stupid increased. Back in the day, pilots in airlines could take whatever route they wanted. They use to go by the grand canyon, even if it was out of the way, to show passengers the view. Back when drinking some whiskey was mandatory to pilot a plane. Then the number of flights per year increased and some assholes ended up crashing in those areas and no one could find them because they took random routes. FAA is established and an end to unscheduled pilot routes.

Volume matters.

The whiskey thing was a joke, for you prudes out there.

> Figuring out how to get people to drive less by providing safer, more sustainable alternatives to the car.

100% this. I could go on forever about how unsustainable private cars are, especially in urban areas, but instead I'll drop some positive news:

The 14th St busway in NYC is real: https://nyc.streetsblog.org/2019/10/03/live-from-the-busway-...

It's a start!

There was an article yesterday also dubiously claiming "Cars are killing more people – and driverless cars won't solve the problem." It seems ironic that just as self driving tech is getting good enough to maybe make a difference you get all these articles asserting it'll never work.

Looking down the list of accidents in the article, basically all of them apart from maybe deliberate road rage could be prevented by self driving tech within a decade or two. Stopping them by taking most cars of the road in the same period on the other hand isn't going to happen.

Are there current or projected estimates of death toll from accidents vs from climate effects?

Unfortunately "Self-Driving Tech" is only in the headline and missing from the actual op-ed, aside from the one line, "So far, I’m unpersuaded."

Self-Driving technology is a broad term that also includes active safety features such as emergency braking and pedestrian detection and avoidance. The TM3 recently won a 5-star rating in AU, in main part due to their performance on self-driving safety features. [1]

Beyond the active safety measures designed to assist an attentive driver, the allure of FSD (Level 4 or 5) as a way to increase safety and reduce or eliminate fatalities is very strong.

The NHTSA estimates that the direct economic cost of car accidents is $277 billion per year, but if you include the economic harm from deaths and injuries (pain and suffering) the estimate rises to $870 billion. Per Year. [2]

What we have here is a trillion dollar cost of doing business that could conceivably be almost entirely eliminated with the right advances. That is just the benefit of eliminating the gruesome side effects of driving. With a technological solution, you don't just reduce cost but dramatically increase value, productivity, convenience, etc. That's roughly another trillion dollar proposition.

A $2 trillion dollar annuity which pays in perpetuity, at 3% discount rate, is worth $2 trillion divided by 0.03 or roughly $66 trillion dollars NPV. That's the scale of the investment that would pay off if it resulted in a mandatory-use nationwide Level 5 autonomous driving system which could eliminate 99% of traffic accidents.

By the way, if you can only eliminate 50% of accidents, and provide 50% of the productivity benefits, the annual savings are worth merely $16 trillion dollars.

I believe Level 5 autonomy is one of, if not flat out the greatest challenge for software in the 21st century, and it is no less than a moral imperative to achieve it.

IMO a significant portion of that investment is not to just somehow create AGI, but several trillion in infrastructure improvements, new standards and regulations, which allow an extremely advanced AI system circa 2025/2030 to be able to function at Level 5 nationwide on highways and local streets.

There's no question in my mind that within 50 years we will look back and wonder how it was possible people actually manually operated these 2-ton machines. It's a bet against humanity to think otherwise.

[1] - https://twitter.com/ANCAPsafety/status/1156060826404438016?s...

[2] - https://www.cooneyconway.com/blog/economic-and-societal-impa...

by this logic, life is a death machine

Cars don't kill anywhere near as many people in Europe, though. The US is an outlier when it comes to first-world countries.

Still, I would like it if the power was taken back from from cars on the road. Cars should be a third-class road user, below bicycles and pedestrians.

My idea for improving driver quality is to do what pilots do. Before you are even allowed behind the wheel of a car you must have proven experience using the road on a non-motorised vehicle like a bicycle. Then you can apply for a small car. After some experience with that you can apply for a large car if you need to. It's insane that we let kids drive cars before they've even learnt how to use road.

> Cars don't kill anywhere near as many people in Europe, though. The US is an outlier when it comes to first-world countries

Do you have some data for this assertion? The US actually has a pretty low fatality rate per mile driven. Although, anything above zero isn't acceptable to some people...

> It's insane that we let kids drive cars before they've even learnt how to use road.

We don't, at least in the US. You have to go through "Driver's Training" first in order to get a Learner's Permit. This training teaches all the road laws, best practices, scenario training, and how things work. Then you get your permit and are allowed to practice only with a licensed driver in the passenger seat.

> We don't, at least in the US. You have to go through "Driver's Training" first in order to get a Learner's Permit. This training teaches all the road laws, best practices, scenario training, and how things work. Then you get your permit and are allowed to practice only with a licensed driver in the passenger seat.

that's how it works in many places. OP was probably talking about letting people <18 years old drive.

> You have to go through "Driver's Training" first in order to get a Learner's Permit.

Not in my state. Here, there is no prerequisite to get your learner's permit. However, completing driver's ed is a high school graduation requirement.

> Not in my state. Here, there is no prerequisite to get your learner's permit.

And

> However, completing driver's ed is a high school graduation requirement.

Is the same thing.

Just your state decided to standardize what was taught, instead of letting private companies also offer it. For what it's worth, in California we have both.

It's not the same thing. You can still get your learner's permit even if you don't complete high school or the driver's ed graduation requirement.

That sounds like a special provision rather than the norm - there isn't a high percentage of youth not graduating from high school in the US.

> Do you have some data for this assertion? https://en.wikipedia.org/wiki/List_of_countries_by_traffic-r...

Of the countries with data for fatalities per billion vehicle-km (not that many), the US (at 7.3) is equal to Belgium, lower than Czech Republic (11.5) and high than the 13 other European countries. In addition, counting fatalities per distance travelled will make driving less look like it has no safety benefits even though fewer people actually die. If we look at fatalities per capita instead (to attempt to adjust for country size, but not policies and behaviour) the US rate is 33% higher than Europe, more than twice as high as Canada and most of Western Europe and 4.5x Norway.

Compared to its peers, the US really doesn't have a low fatality rate.

Isn't that what learner's permits are? In my state, you need a minimum of I believe 40 hours of driving with an over-21 adult that they sign off on, then completing a driving class with an instructor, and over however many sessions they judge your driving, and after that you get your license. In my experience the student drivers were the ones staying strictly to the speed limit, always using indicators, etc. When I first started driving I was hyper-focused on not breaking the rules, paying attention to the road. The ones who are the worst are the people who think they know how to drive well, so they are on their phone or half-asleep while driving.

You need to look no further than sports cars in the 70's. American companies concentrated on straight-line performance for our seriously long and pretty straight roads. But dodging things is one of the reasons you have to go somewhere other than straight ahead.

Meanwhile you had British roadsters with literally 0 body roll on corners because the CoG is below the top of the axle.

Petrol is more expensive and it seems that owning a car is less of a requirement in Europe than the US. Most major cities in the US that are exclusively designed around the availability of automobile transport. People live in the suburbs and commute in, and it is prohibitively expensive to live in non-car commuting distance.

Paywall: is this perfect being the enemy of good?

Planes, trains, bicycles, people die on all of them. Would she argue airline autopilot was a bad thing? Well, paywall so who knows.

Self-driving tech with convergent evolution will absolutely reduce deaths. Tesla claims a statistical advantage with their speciously tested technology.

> Planes, trains, bicycles, people die on all of them.

By that argument, you can claim that people die eating strawberries, which I'm sure must have happened at some point in history, and it would be equally true and equally worthless to point out.

You know that car deaths are much more prevalent than plane or train death, and how many of the bicycle deaths are because they are hit by a car?

> Would she argue airline autopilot was a bad thing?

Would you argue airline autopilots were even remotely comparable?

During takeoff and landing, both pilots (yeah, two people!) have full, undivided attention to flying the plane. There are "sterile cockpit" rules which explicitly forbid unrelated chatter. Even with a full auto landing (something mostly only done in very low visibility conditions), the pilots have no relief of attention.

At cruise altitude, it is true that the autopilot relieves the pilots from the attention of having to fly the plane (it does not relieve them from still scanning the instruments, managing the flight, and many other tasks). But that's because at cruise altitude, you might have noticed that there are no roads, trees, or any other objects whatsoever. Including other planes.

NY Times paywall is wimpy. In Firefox, there's no problem if you hit reader view as soon as the icon appears.

It might change them if human drivers are outlawed or removed completely.

It may lower to death toll of avoidable accidents.

The author isn't wrong that tons of metal moving at any speed can be absolutely fatal.

Practically anything “can be fatal”. If you argue in tautologies you’re never wrong, but you also never actually say anything.

boggles my mind that people who know the first thing about software can actually think that self driving cars are viable technology

Software technology isn’t static, most of us are only arguing about when, not if.

And I'd argue that "when" is very, very far away. Unless you remove human drivers, and significantly change the infrastructure for self-driving cars, but at some point you are not so far from (existing) self-conducted trains, then.

And you offer no argument against “self conducted trains”.

Look, people are horrible to one another. They bring their brutalist egos into their cars and drive at the whim of it. Very few people drive the safe distance away from cars, drive safe distances, let alone reacting to perceived slights by other drivers in defensive ways. No, instead they hug bumpers, cut one another off, drive so close a simple pat on the breaks would cause an accident, and so on.

No, I don't, and it was not my intent to lessen the value of "self conducting trains" at all. Having been raised in Germany, I'm a big fan of public transport.

I just wanted to point out that "self driving cars", in isolation without any change to the infrastructure they operate in, are, by my intuition, a pipe dream.

I think we are aligned here.

We already have self driving cars running around Mountain View with little problems. Extending that progress to more extreme conditions doesn’t seem far fetched, especially when we continue to have technology advances in sensors (LIDAR) and AI control. 5 or 10 years might sound aggressive to many people, but 20 years doesn’t sound far off at all.

Driving in Mountain View just isn't that representative of a lot of the US, much less of the world. You'll never have training data that covers every scenario as we're constantly creating new ones -- from spray painted temporary signage to unclear directions from police officers to random obstacles that force breaking rules to navigate around. 20 years sounds pretty optimistic to me.

Why does some people assume the technology will start standing still when it has advanced so far in just the last 5 years? None of those problems are insurmountable, when the tech stops getting better, maybe you’ll have a point.

20 years is incredibly pessimistic to me. Heck, we might even have GAI by then.

Why do you assume that those "little" problems that stand in the way are surmountable with any realistic rate of advancement in technology? Because some ill-defined "amount" x of advancement has happened in the last 5 years, so in 10 years it will be 2 times x?

Some problems are really hard, some are essentially impossible. For example, the halting problem is impossible to solve. All its corollaries are, too: You cannot for arbitrary cases tell whether a C program adds two simple numbers together, period.

Voice Assistants are an example for "problems so hard, they are essentially impossible": A human housemate has no problem understanding what I mean when I say "it went yellow, can you order this time?", a voice assistant likely never will. If it can understand that one, it cannot understand another way to phrase it that I can come up immediately. It might get good enough to unambiguously understand "order the water filters for the fridge" every time, but I can convey the same thing to a human in infinitely many more ways.

In driving, unfortunately, you don't get to choose which way a situation is presented, the way you can with crafting a sentence to a Voice Assistant. The next time you drive through a city, look out for any even remotely non-standard situation. Now imagine how a computer should be supposed to react without the arsenal of context clues that your lifetime of human experience provides to you, and that tells you to "just swerve left after slowing down for a few feet" with barely a conscious thought.

I've detailed why I think that this blind extrapolation of "progress" is rather arbitrary and far-fetched below, but I also just realized your mentioning of "AI control", so I want to ask:

With what you know about current AI, how it's "trained" instead of "programmed", at the result of something that can at best be analyzed to a certain degree: From the bottom of your heart, would you really, really trust that AI blackbox component to navigate you through real-world traffic situations at dangerous speeds?

Reply on news.ycombinator.com