Hacker News Reader

Comment by absentmoon | Hacker News Reader

Pressumably things will start to snowball if these smaller scale trials demonstrate adaquate safety

One question is to what degree safety is dependent on testing in these specific good weather areas for literally years. And even the SF public self-driving area excludes what is almost certainly the somewhat harder area of SF for self-driving (though opening up much of SF is more impressive than Phoenix generally).

Seems like fair weather tech at best at the moment. I just can't see these things working in a place like a Canadian winter without a few generational improvements. It starts with the basics, like knowing you have to start slowing down 3 blocks before the intersection if you don't want to slide through it, but also involves knowing how the driving culture changes during a storm.

We can install like IR reflectors in the road so that they know where the lines are under the snow, but people don't know where the lines are, and so during a heavy snow day the lanes change and take on people's best guess and it becomes the new default choice as more people take the new paths.

As is often repeated the most dangerous thing to be on the road is unpredictable, and I don't know if the self driving cars would have the ability to see the changes in lane positions and adjust, or if they'd be trying to follow the old lanes in a crowd of cars making their best guesses to make it home. Seems one of those "it won't work well until the vast majority of the cars are automated" type deals.

> the most dangerous thing to be on the road is unpredictable

Google already has the biggest hive mind for predicting what's ahead. One reason I turn on Waze even when driving a familiar route is that I can see traffic and conditions ahead of me. Totally saved my butt when I noticed dozens of accidents ahead in what seemed like a light rain. The road had iced over. I got off at the next exit and waited it out, using Waze to know when traffic was moving again. Passed some horrifying big truck accident scenes yet to be cleared by tow trucks.

You are right. You wrote: <<3 blocks before the intersection if you don't want to slide through it>> It sounds like road maintenance needs improvement in your city.

That's funny, with that framing it seems as though it won't be the robot cars fault for screwing up even though the majority of drivers can get home safe in those conditions.

When it's -30 for a week you end up in a situation where road salt doesn't work and the exhaust from cars polishes intersections to an icy sheen. It's hard to protect against that.

Humans suck in these conditions too. The solution is to go slower, pull over in the worst conditions, take an off ramp, don't drive in the first place, etc. These are all things humans have been known to do.

Road lines are important but physical obstacles like other vehicles are more important.

Will the car follow through with these solutions as well? Are they set up to do that now? Can they make that assessment?

Sounds like the solution here is just to not run the cars when people are going to be most likely to want to use it.

Even if it's safe...what's the short term benefit for riders beyond curiosity? Presumably cheaper fares + more throughput but the math is not obvious to me that owning a fleet of robotaxis will be a lucrative business to be in anytime soon, even if you are the only robotaxi company in town.

Seems like they need to license this tech to have it payoff.

cheaper fares is a really big deal. if you can deliver the same service 30% cheaper with 20% higher profit margins, that's massive. driver pay is a significant component of cost so cutting it is pretty massive.

I guess, but if I'm getting a taxi, replacing a human with a computer is not a huge benefit to me, as I'm already not driving. If I'm driving myself to work every day, replacing that repetitive and boring task with a computer is a huge user benefit and thus I'd be more willing to pay and at a higher price for the tech.

I just think that's the only real way to make this thing pay off. Also potentially it might work for long haul trucking.

the benefit to the user isn't the computer, it's lower cost. the bushes model makes sense because you can lower user cost while increasing profit

The benefit absolutely is the safer computer driver. I'd pay more for that.

Maybe? Not that I use Uber/Lyft/taxis much but, when I do, I of course prefer them to be cheaper but +/- 30% pricing pretty much wouldn't affect if I take them or not. It's still a significant premium relative to driving myself at home and, if I'm traveling, I may not have much of a choice.

> but +/- 30% pricing pretty much wouldn't affect if I take them or not

But it's enough to prefer Waymo over Uber and that's all it takes.

if +/-30% pricing isn't significant, then the benefit isn't lower fares, it's higher profits for the company operating the cars. either way you slice it, not having to pay drivers would be a significant win for a taxi company.

Not necessarily if the technology (both development and hardware cost) to replace the drivers costs enough. I don't doubt it will get there someday but Waymo certainly isn't making money off current riders.

License it to who? If they prove the model then investors will be lining up to give them the capital to expand.

Car manufacturers and trucking companies.

> Seems like they need to license this tech to have it payoff.

Seems like that's the plan. Volvo and some other car OEMs are using Android on the dashboard. A total package of infotainment, autonation, and the cloud services behind the in-vehicle systems is a likely destination for this and other products than can be sold into cars at the OEM or end-user level.

Reply on news.ycombinator.com