Hacker News Reader

Comment by seibelj | Hacker News Reader

Driverless cars in urban areas are decades away

I find it interesting how the same people that have no problem accepting that the last 10% of a random software project accounts for a significant portion of the total time spent, can at the same time be so unreasonably optimistic that a whole sphere of tech can go from like 70% to 100% (where 100% is also the minimum amount required) in a really short period.

Probably for the same reason software developers can't estimate tasks. At least for me the problem is that I figure out the algorithm at a high level and base my estimate around how long that will take me. But inevitably there are platform gotchas, or the API doesn't work the way I expect, or I forgot/didn't know about edge case. (I've done this enough that I have an idea of what kind of things can go wrong now)

I think the optimists are thinking "oh, the high-level problem of image recognition and building a 3D model of the area around the car is pretty much a solved problem, great, done!" But in fact, it's all the edge cases that will kill you. What happens when it is dark and rainy and the car in front of you is black, with no license plate? What happens when the car in front of you hits something in front of them unexpectedly? What happens when the car gets a flat tire? What happens when someone doesn't look in their mirrors before changing lanes? What happens when your "there's a problem" solution (like, say, stopping or slowing down) is actually not a good solution?

You would have to defend that humans make good decisions in such cases. I would argue that, for the majority, they do not.

I have been in all of those situations except for the black car, and the humans involved all made spectacularly poor decisions.

People who claim that self-driving cars will have failure modes far underestimate the failure modes that humans have.

I think the disagreement is what the last 10% are. I see urban driving as part of the 90%. My 10% are rural gravel roads with puddles, snow covered roads and similar bad road and weather conditions.

I see those stupid robot cars all over the Phoenix metro area these days, whole packs of them roam Chandler/Tempe/Scottsdale areas.

Now they have permits to run around 100% humanless apparently.

Right. That must be why a Waymo drove itself past me in the Bay Area two weeks ago.

5 years ago I was told the same thing, and 5 years before that. Maybe 5 years from now, but it’s been the same story forever.

You saw a driverless car (with a safety driver) drive past you five and ten years ago?

In October 2010, Google announced that they'd logged 140,000 miles with a self-driving car, with a driver, including through San Francisco and a lot of California [1]. So eight years ago is possible, and I'm willing to grant a little hyperbole here.

[1] https://googleblog.blogspot.com/2010/10/what-were-driving-at...

Honestly, it really hasn't been the same story forever.

In the 2004 DARPA Grand Challenge, of 21 teams entered, 15 qualified to race on the course, and none finished the 240 km course. The best result was 12km. That was off road, with no requirement to obey traffic laws or other cars to hit.

Since 2009, Google/Waymo has 5 million driverless miles under their belt.

Is it 5 years away? I don't know, but this is not the same story forever. This is iterating and improving at a breakneck pace.

That's the problem with exponential development. It seems like you're getting nowhere for decades, then all of a sudden it's starting to amount to something, and in the span of maybe a year or two it goes from impractical to plausible to obvious and ubiquitous.

The problem with predicting these curves is there's usually a confluence of factors that lead to that tipping point. When the right algorithms, hardware, and software are combined it'll be obvious in retrospect, but right now we're still fumbling around with primitive solutions.

When will there be a TPU-type device suitable for in-car use? When will there be adaptive deep-learning algorithms available that can work in the demanding real-time environment of a vehicle? When will enough testing be done that we know such a solution can work without endangering people?

It will happen, but pinning down when is very hard. You'll only know when you're close, and by then you're already flipping from impossible to inevitable.

I saw one yesterday. There was a driver but like... come one, decades?

When driverless cars are common on midwestern highways every night, they will start moving into complex urban areas. We are very far from cars navigating downtown NYC, Seattle, SF, Boston, etc.

Do you live in NYC, Seattle, SF, or Boston?

Yes, Boston, and when can I order my driverless car to pick me up?

A few months ago if you ignore the safety driver [0].

If cars with safety drivers don't count, nearly a year ago if you move to Pheonix and get lucky [1]. They're apparently planning to launch in more cities "soon" (though I can't find a list or date) [2] and they've recently got permission to start charging passengers and intend to by the end of they year [3].

So I think "decades" is an exaggeration.

[0] https://www.theverge.com/2017/12/6/16742924/lyft-nutonomy-bo...

[1] https://www.theverge.com/2017/4/25/15415840/waymo-self-drivi...

[2] https://www.theatlantic.com/technology/archive/2018/01/waymo...

[3] https://www.digitaltrends.com/cars/waymo-now-a-full-ride-hai...

Re "midwestern highways": haha, I don't see a "self-driving" car navigating the I-90 in Chicago anytime sooner than SF. ;) At least SFBA has super mild weather that makes the obstacle detection problem much easier. Here there are only two seasons: winter and road construction.

(But yeah, we are very far from cars navigating urban areas, or indeed anyplace where weather is inclement and so are people.)

Pretty sure driverless cars are being tested in almost all of those cities already.

The Falcon Heavy took five years to develop and it's basically just a bigger version of the same Falcon 9 rocket they've been sending up for years.

Even if you have a driverless solution today it'll take years to test it and get it certified.

A gas-powered, highway-capable, driverless car feels callous. I hope those algos are good.

Urban areas are easy. Snow is a hundred times harder.

I see a lot of driverless ubers here in pittsburgh

Reply on news.ycombinator.com