When two cars are facing each other, there is a lot of human interaction and understanding about who should go first, who should pull off into a parking spot, or perhaps who should reverse into the intersection to let the other pass.

This is one of thousands of edge-cases that are going to be extremely challenging to automate. I'm not sure it's even possible without some understanding of the counterparty, their intentions and body language.

One problem with developing these technologies in California is that California has very wide, very regular roads. High visibility and dry conditions are the norm. There are few uncontrolled intersections (in Seattle they are the rule, not an exception). I think California's history as a motorist culture has conversely made it one of the easiest places to automate.

We have those in California, too.

> One problem with developing these technologies in California is that California has very wide, very regular roads.

Except where it doesn't.

> High visibility and dry conditions are the norm.

San Francisco, for instance, has fog about 1 in 3 days of the year.

This is one of the things that self-driving cars WILL do better--and fairly quickly.

Having a sensor system that isn't blinded by aerosolized water will be a huge benefit to driving.

It has few enough that you can route around them and still get where you're going.

I think this is a misconception. California has many residents who live in areas that don't have the super wide roads you're thinking it has. It has a wide gamut of roads and many residents (such as myself) are surrounded by more narrow roads. I'm not even in San Francisco. I'm in a suburb in the bay area. (San Carlos)

The following seems to show a road next to a playground in San Carlos.[1] You think thats too narrow a road for an autonomous vehicle to navigate without endangering kids? Have you familiarized yourself with the latest in autonomous / self-driving tech?[2]

I think your fears are overblown or purposely exaggerated.

[1] Laureola Park San Carlos, CA Playground

https://www.youtube.com/watch?v=3g3ITuPd7Gw

[2] The $800M Robo Taxi That Could Beat Uber

https://www.youtube.com/watch?v=OjDLwnTyybo

https://goo.gl/maps/vokEswX4S8z

I live on that street. The streets adjacent to it are also plagued by a plethora of cars on both sides. Much worse than what streetview is showing because streetview is done midday on weekdays when everyone is gone. (Notice how most of the driveways are empty and recycling cans are out - it's the middle of a Thursday that this was taken) It used to be that people would park with their cars on the curb to give more access but I don't see people doing that anymore.

Certain weekends and at night these streets are very full with cars. Much more than google street view shows. It's deceptively smaller than it appears. It looks as if two cars might be able to slip by but frequently not the case. I can't actually remember the last time I saw two cars go by each other side by side without one pulling over to wait for another to pass on my street in particular. The more main street to the south has the same problem but is even wider.

The streets in particular are pretty wide for this issue but because of street parking being available on both sides, it's an issue.

Actually I am curious what europeans honestly think of the self driving car craze, since getting the bus, the metro, and.. ehrm walking and biking is much more common and pleasant there.

I think we'll need general artificial intelligence to handle that.

There are roads near me that people struggle to drive down. The amount of negotiation with other drivers, the ability to bump up a hedge or pavement, or reverse hundreds of meters ... or sometimes the need to assertively barge through. But we have nice roads, so geo-fencing to AI certified roads sounds fine.

Isn't dealing with corruption and corrupt officials rife in the quotidian lives of Italians even in 2019? Don't you think its high time Italians figured this dysfunction out rather than expect the modern world to bend to the moribundity of your culture and ways of doing things? [1] [2]

[1] The Impossibility of Italian Politics

https://www.the-american-interest.com/2018/03/07/impossibili...

[2] Italian Elections: More of the Same Political Dysfunction

https://americas.nikkoam.com/articles/2018/02/italian-electi...

I like Musk's proposal for mostly autonomous tunnel driving because it actually seems much more feasible than handling every edge case known to man.

Maybe I'm just not familiar with the specifics of that proposal, but I don't think tunnels solve the problem you presented either. If I need to go down a one lane street to get to my house, I'll still need to do that whether we have tunnels or not, unless for some reason the tunnel exit is right in front of my house.

Personally, I think the solution is to make only make certain large, predictable roads self driving and force a user to take over on smaller more complicated roads. As technology advances, you can slowly add more and more of the smaller roads to the self driving map.

According to Elon, his investment in cheaper drilling tech will mean we will have many tunnels with relatively cheap to build entrance and exit points that are significantly smaller than subway stations.

These entry points will allow you to drive to and from your house, enter the tunnel system where some sort of conveyor belt/computer controls your car. You exit the tunnel and drive normal.

So the idea is that the tunnel exits will be _very_ near to your home. Otherwise it's dumb. I think what you are saying about large predictable roads is basically what Elon Musk is proposing but underground.

https://www.boringcompany.com/faq

- There are going to be certain places where you can't dig a tunnel for geological reasons. For example, half the population of Florida lives less than 4 feet above sea level.

- Even if you increase the speed you can dig a tunnel and lower the cost, it's still cheaper to pave a road above ground.

- There are plenty of places where NIMBY folks just wont let you dig a tunnel (looking at you LA purple line extension).

Ignoring all that though, even if you can build an entry point every half mile, you still run into the issue of needing to drive the car that last stretch, which means you need to either solve the problem of self-driving on complicated roads, or force the driver to take over. The tunnel system shortens the amount of time you would spend on small roads, but it doesn't eliminate it.

Not reasonable. You'd be parking the car at least half a mile from where I live because it's full of these streets. Usually they're one car at a time because cars are parked on both sides of the street causing the streets to become just big enough for one car to get through.

For reference, I live in San Carlos (Bay Area). I've encountered this same issue in many other parts of the USA too. It's not uncommon at all.

If your car could get to your location without needing to use those streets, it could park itself anywhere, half a mile or further. There is also the very reasonable possibility of banning parking conditions that lead to that kind of issue, made more palatable by my earlier point.

I've learned to always think in "options", not "absolutes".

You're doing well until you say this. The state of California is just as diverse as Seattle when it comes to roads. I lived in both quite a bit...

California has some bad roads with poor visibility. Recently, we've had a lot of forest fires. Guess what - visibility gets pretty bad. It's like driving in fog (and we have that too). We also have snow! (The bay area got snow this winter - skyline did its thing and fell off the face of the earth again from it too) We had so much snow that most of Tahoe was closed for winter. The roads aren't super well maintained and if you go do some trips to Tahoe (good luck getting there or back!), you'd find out that the markings for lanes just disappear altogether. (Snow or not) We have single lane with 2 way traffic either intentionally or because there are so many cars parked on the road that only one car can fit through. There are insane hills with no nearly no visibility unless you creep forward into the intersection past the stop sign when at the top. Same with going down them - no god damn idea what's below until you just jump that cliff. You have some of those in Seattle but they're really steep in SF. Incredibly twisty and poorly paved roads with no road markings that suddenly change from having lines to not having lines randomly or permanently (e.g. Alpine, Old La Honda Rd, Kings Mountain) and here's the kicker - they're pretty popular roads with lots of people using them all the time with a wide variety of people on them. It's common to see a cyclist on the road or a damn horse and having to overtake them when there isn't a ton of room (and there's a very real risk of oncoming traffic) unless you want to be stuck behind them for 30 minutes during their little workout or trot up the hill.

So... don't write off the whole state. There's a lot of varied conditions.

Will self-driving cars be applicable EVERYWHERE in EVERY condition shortly? No. But even if they're only available in ideal conditions in some places -- that can still be almost unfathomably huge.

On the contrary, this is very easy to automate, it's a simple shared route planning communication between the cars.

The challenge remains interpretation and reaction to irregular environmental objects and structures.

Ah yes, so simple.

Obviously the cost of this is prohibitive. But at some point we have to get it through our heads that for a computer to excel at a task, the task needs to be redesigned.

I personally shudder at the thought of redesigning our cities any more around cars, and instead hope the next decade will continue the trend of adapting car usage to living in cities and not the other way around.

---

Self-driving cars can and should be an amazing complement to city life: owning a personal vehicle would be unnecessary if the city were dense enough to support a fleet of ever-roaming self driving cars, allowing more space for pedestrians, residents, and workers to enjoy the city and reclaim all the land dedicated to street-parking and parking lots (Los Angeles proper has more area dedicated to on street surface parking than the entire city of San Francisco is big). That's more housing, more parks, more schools, more activity, more events, you get the idea.

Most of that promise would dissipate if we had to further redesign the city around the car, and for what: so the few of us lucky enough to own a self-driving car could live far away from everyone else and putter around on their phone while their car is stuck in traffic on ever wider and wider highways?

American city planning underwent a fundamental shift in the 1950/60s which prioritized cars (I'd argue for the worse), and I think it's past time we undergo another shift away from ever-extending sprawl.

At this point, why not just have tracks? Really, you can automate so much by just putting the damn thing on rails.

Of course that would require centralization and working with the government. The government has no incentive to do this because it would just create another problem for them (mass unemployment).

Actually the government DID try to do this. The U.S. DOT had an entire decades long program researching the following:

1) Costs and plans for updating roads with sensors

2) Upgrading traffic boxes (the things that control traffic lights and such) with a module to communicate back and forth with surround cars

2) Forcing manufacturers to install a module in all cars that would allow them to communicate with surrounding cars + road sensors + traffic boxes.

The problem was the amount of push back the government got from car companies was really absurd. And of course Republican funding games. Republicans love their private companies so they basically sapped the program's funds.

It was able to pick up momentum again from around 2012 onward though:

* https://www.its.dot.gov/about/federal_its_program.htm

* https://www.its.dot.gov/research_areas/automation.htm

* A practical example of its research: https://www.its.dot.gov/pilots/florida_agencies.htm

start designing roads specifically for ease of use. design lanes that are extremely easy to see. Traffic lights that are easy to parse. Street signs standardized and easily readable.

Good luck when it snows. And it's 3 AM. And no one else is on the road, so there are no context clues.

Um, a human sucks in this case, too.

There is a reason we put reflectors on the lane markers and rumble cuts in the road.

I'm from the snow belt, and these are invisible until the plows come through.

We'll eventually embed sensors in the road, but that's not really general purpose self-driving imo.

If the tech industry wasn't so bubbled, this would have been the plan from the beginning.

Autonomous cars are interesting for the trucking industry - for everyday driving, reducing traffic via remote work and smarter city planning is what'll make people happy imho.

I don't mind a 10 minute drive. It's the 100 minute drive in bumper to bumper traffic that makes people dream of self-driving cars. The real answer is not being in that position to begin with.

And, I suspect that give a 100 minute drive route now, if self-driving cars lower the bar of being in traffic, that the same commutes will actually get longer with even more traffic.

That to me is the crux of the issue. Self-driving cars are a poor solution to an entirely self-inflicted problem. The only viable long term solution is to reduce commuting.

In regards to world hunger - many of this planet's woes, would be helped by advancements in science, agriculture AND not endlessly having children.

What it doesn't do is satisfy the fantasy of those who live in dense cities that they'll never need to own a car or even learn to drive. But hitting the point where a car can reliably drive itself around a chaotic urban landscape in all sorts of weather conditions by itself is almost certainly a much harder problem.

I really want autonomous cars to be a thing in my lifetime, and feel like this sort of optimism is more damaging to their future than people being skeptical or even down on the idea. The more people push the narrative that it's a done deal and just around the corner, with no infrastructure changes required, the more backlash (fair or unfair, it doesn't matter) is created whenever there's an accident involving autonomous vehicles.

That said, I think "decades" is overly pessimistic. There have been some great strides in computer vision and decision making over the past few years and the hardware is pressing ever onward. That said it will be decades (if ever) before we're willing to rip the controls out of the cabin entirely, but a level 4.5 system where the people in the car can do whatever while it drives and only need to intervene when the computer alerts them that there is a situation it cannot handle (but can stop a safe distance away from if necessary).

We give plenty of indications. It's just that they're hard to define objectively.

for some values of plenty < ε. As cyclist I can tell you that human drivers often don't even use their turn signals at intersections where it is important for me to know where they're going. And it's not like I'm sneaking up on them, it's when they're facing me at the other side of the intersection.

I disagree. It's not up-front, there's been decades of work on autonomous systems such as drones, UAVs, trains, robots, etc.

It's just that it's a hard, if not impossible problem. Some companies thought they could successfully navigate this problem with enough data. I don't think it's a data problem, I think it's impossible to give real situational awareness to a machine, and that's the problem.

FWIW I know of at least one company that's developing their autonomous vehicles with a remote piloting feature. In the event that the vehicle runs into a situation it cannot handle control is passed to human operators in cubicles that take over control of the vehicle on the spot and release control only after clearing the obstacle and the on-board autonomy gives the green light. They also had ambitions to record what the human operator did and program that into other vehicles into the fleet on the fly as they approach the obstacle. It sounded very ambitious (and data hungry) to me, but is probably the only way to fully remove the controls from the vehicle cabin in the foreseeable future.

https://en.wikipedia.org/wiki/Moravec's_paradox

> As the new generation of intelligent devices appears, it will be the stock analysts and petrochemical engineers and parole board members who are in danger of being replaced by machines. The gardeners, receptionists, and cooks are secure in their jobs for decades to come.

Indeed, much of the trading world has been replaced by machines.

There has been great progress made and I don't see any reason why we won't see them in less than a decade. However, this is only going to be in very specific areas. Really complex driving, like roads with no lines, a single lane but allows two directions, etc; I think that type of driving will take decades.

So in 8 years I can totally see getting off a plane, jumping into an autonomous car and having it drive me somewhere specific. But I can't see driving through all roads until much, much, much later.

That is a flawed premise. Humans do not operate cars perfectly either. The highest requirement you could put on them is to be as good as the average human driver. And given the other benefits that self-driving fleets offer we can even go a little lower and still come out positive according to various utility metrics.