Waymo: Google's self-driving car company • Hacker News Board

Re: Waymo: Google's self-driving car company

#521

Post by walrus1066 » Tue, Dec 13, 2016, 10:52 PM UTC

yohui wrote: post #494

walrus1066 wrote: post #339

It feels there's a mountain of hype around autonomous vehicles. I think the core challenges to a fully autonomous vehicle (with no human backup), are still far from solved. They need to reliably deal with an almost infinite number of edge and corner cases, each quite different from the last. For example: - communication with other human drivers. In London, this is required all the time, like when parked cars block th…

https://www.wired.com/2016/12/google-self-driving-car-waymo > By early 2015, Fairfield thought they were getting close: The cars were clocking full days at Google’s testing grounds without needing human takeovers. They had driven 1.2 million miles on public roads and could pull over for emergency vehicles. They could detect and brake for squirrels, and read hand signals from cops and construction workers. They knew w…

The car-human interaction bit isn't from a quote, is there any evidence of this actually happening in the field? Would be great if there was a video or something.

Re: Waymo: Google's self-driving car company

#522

Post by agumonkey » Tue, Dec 13, 2016, 10:54 PM UTC

walrus1066 wrote: post #339

It feels there's a mountain of hype around autonomous vehicles. I think the core challenges to a fully autonomous vehicle (with no human backup), are still far from solved. They need to reliably deal with an almost infinite number of edge and corner cases, each quite different from the last. For example: - communication with other human drivers. In London, this is required all the time, like when parked cars block th…

All these are old "problems", either solved or in progress.

Something I've never seen adressed is how SDV fleets behave when change occurs. Are there chaotic oscillation patterns that emerge from AI trying similar avoiding behavior.

There's research in massive multi agent behavior, but I've never seen it tested on the field with SDV makers actual models.

Re: Waymo: Google's self-driving car company

#523

Post by aetherson » Tue, Dec 13, 2016, 10:55 PM UTC

mypalmike wrote: post #508

aetherson wrote: post #384

Earlier quoted context omitted.

Not without seriously increasing the costs for car rental. The cost savings for car rental are predicated on increased utilization. Increased utilization is predicated on each car having more people using it on any given day. And these things don't scale infinitely -- each car can only serve a particular locality. The number of variations that you propose is a fast geometric progression. Kids 0-12 months or so need d…

> The number of variations that you propose is a fast geometric progression. Only if you're an engineer. To a marketing person, you bundle groups of features together and price them at various tiers.

Er, I wasn't talking about pricing complexity, I was talking about the ability to actually get the vehicle that meets your needs. It doesn't do you any good to know that "a vehicle with three or more constraints costs $1.50/mile" if you get a vehicle whose three constraints are "4WD, Bike rack, Dog area" when what you wanted was "carseat1, carseat2, large storage area."

And if the fleet can successfully meet arbitrary needs around constraints, then the fleet is going to be heavily overprovisioned, and thus expensive.

Re: Waymo: Google's self-driving car company

#524

Post by semi-extrinsic » Tue, Dec 13, 2016, 10:55 PM UTC

jcranmer wrote: post #312

joeguilmette wrote: post #49

I'm interested to see where all the major players end up in 5yrs: • Tesla bootstrapping a ride service on the backs of buyers • Waymo directly rolling their own fleet • Uber trying to get a self-driving fleet up, burning mountains of money to maintain their "monopoly" on Uber for rides • Lyft working with GM to get a fleet up • All of the other car manufacturers trying to get autonomous vehicles going, presumably hop…

> Also: Private car ownership is going to fall off a cliff shortly after autonomous ride services arrive. Which probably means general demand for vehicles will fall off a cliff. I've seen this sentiment so many times now that I have to ask a serious question to anyone who echoes this sentiment: Do you own a personal car right now? If so, why? All of this... excitement over the transformations that self-driving vehicl…

I have to agree with you skepticism. Let's e.g. replace "car" with "apartment". Imagine an apartment where you pay a premium to have all the cooking, cleaning and maintenance done for you. This obviously exists today, yet >99% of people don't use it.

I believe people like owning things, and using them to do stuff. This is a basic fact of being human. Probably goes right back to the stone age. I doubt autonomous cars will change that pattern.

Re: Waymo: Google's self-driving car company

#525

Post by riprowan » Tue, Dec 13, 2016, 10:56 PM UTC

brwr wrote: post #486

riprowan wrote: post #308

A lot of comments here and I haven't seen one that addresses the problem of impulsivity. Many / most of the times I need a car are not schedulable events but impulses. For example when I'm working on a project around the house, it's typical to need to run up to the hardware store 3 times in a day. I think autonomous transport will be very popular but shared vehicles will be hardly more popular than Uber / taxis are n…

I think you may be experiencing recall bias. How do you get to work every day? How often do you go to the grocery store? What about going out to dinner? For myself, living in the Bay Area, anything related to driving or owning a car is a nightmare -- not that that has stopped me from owning a car. Driving from Mountain View to San Mateo takes about 45 minutes in the morning at 75 minutes in the evening. I went to SF…

I don't think you read my comment clearly. I think autonomous cars are going to be quite popular, but I don't understand how that necessarily leads to shared instead of privately owned autonomous cars.

Re: Waymo: Google's self-driving car company

#526

Post by ageofwant » Tue, Dec 13, 2016, 10:56 PM UTC

thaw13579 wrote: post #436

notatoad wrote: post #354

Earlier quoted context omitted.

Do you really think the companies building autonomous cars haven't thought about all these things? They may not have solved them yet (If they had, we've probably see their products on the road) but I'm sure it's in progress.

I agree that it's safe to assume folks at these companies are aware of these issues. But if I'm going to risk my life by using one of these systems, I'd want to have good answers to how they have solved them. Until then, I think it's fair to keep asking to avoid unrealistic hype.

It only needs to be a better driver than you for you to be safer, and its that already.

I can guarantee you that on a one to one driving skills shootout between you and the software you will loose 98% of the time.

Re: Waymo: Google's self-driving car company

#527

Post by gist » Tue, Dec 13, 2016, 10:56 PM UTC

What's the story in the video with 'pulling at heartstrings' by featuring a blind person? The problem of a blind person getting a ride is obviously solved. But more importantly it's an edge case of a need for a vehicle that is self driving. And actually I wonder whether someone who is blind actually would rather have a human driver in the car and feel safer generally that way. Seem to me almost like a strawman in a way. In other words 'if you feel that you have to help the handicapped you need to be onboard with self driving cars'.

Re: Waymo: Google's self-driving car company

#528

Post by deadbunny » Tue, Dec 13, 2016, 10:56 PM UTC

woofyman wrote: post #496

ticviking wrote: post #466

Earlier quoted context omitted.

That's a key point here. The cars don't have to be perfectly safe, just safer than the average driver.

It has be better than the best human driver or they'll get sued out of existance. You can't sell a product that you know will hurt or kill a certain number of people.

I'm pretty sure the car industry has been doing that for years.

Re: Waymo: Google's self-driving car company

#529

Post by walrus1066 » Tue, Dec 13, 2016, 10:57 PM UTC

jasonwatkinspdx wrote: post #507

walrus1066 wrote: post #475

Earlier quoted context omitted.

I think safety is fine, especially in cities with not too crazy weather. It's more that if an autonomous car with no human stalls, due to a traffic light that's stopped working, or an oddly shaped plastic bag in the road, it's stuck, and would cause a traffic jam. Not sure how to resolve such a scenario, can't rely on someone remotely taking control, as mobile data doesn't have 100% coverage, so a human who is allowe…

First, some baselines: humans suck at driving too. Next, coverage isn't an all or nothing thing, it'd be quite easy for an autonomous uber like service to restrict itself to areas that permit remote control when cars need assistance. Last I'd encourage reading some of the literature behind what folks are doing: it's a very different sort of system, one that blends a lot of complex probability estimates. It's not like…

Machine learning tools also cannot cover corner cases, if they're not present in it's training data. Communication with human agents (other cars, police officers) is also extremely tough.

Re: Waymo: Google's self-driving car company

#530

Post by spuz » Tue, Dec 13, 2016, 10:57 PM UTC

walrus1066 wrote: post #339

It feels there's a mountain of hype around autonomous vehicles. I think the core challenges to a fully autonomous vehicle (with no human backup), are still far from solved. They need to reliably deal with an almost infinite number of edge and corner cases, each quite different from the last. For example: - communication with other human drivers. In London, this is required all the time, like when parked cars block th…

I think a good way to think about the last 5% automation problem is to look at other examples where automation has replaced human work. In every case that I can think of, the work of the human was never 100% replaced. Automated check outs at the supermarket still require a number of attendants. An automated assembly line still needs factory workers. Automated rail systems are still controlled by humans in a control room. You can also look at how the environment has been adapted to better suit the new technology. For example to allow big heavy horseless carriages roll along roads we replaced dirt with tarmac (asphalt).

Driverless cars will never (until we have general AI) be able to fully replace a human driver. However we don't need to get to 100% for the technology to become useful. It might be that driverless cars will be restricted to only driving on certain mapped roads. Or every car's movements are tracked in a central control room and if a problem is detected either by the car or the passenger it could be controlled remotely. Or let's say a GPS sensor breaks an actual human driver might be issued to come and rescue you. Lastly we might even accept a certain failure rate. Society is willing to tolerate failure of technology to a certain degree in cars and planes today. All we need is is for the technology to make us all better off on average for it to be successful.