Live data from Hacker News

Geofencing Will Enable L5

semiengineering.com

71–76 of 76 posts

Re: Geofencing Will Enable L5

#71
post #28
post #16

Earlier quoted context omitted.

Same reason you don't directly control the elevator going up and down.

The possibility space of elevator motion seems somewhat more constrained than that of an automobile...

Yes, that's a large part of why we were able to automate them so much earlier than cars. But we've since removed the lever than elevator operators used to use to control the elevator and noone really seems to mind.

Re: Geofencing Will Enable L5

#72

L5 is impossible by this definition: >“operating the vehicle on-road anywhere that a typically skilled human driver can reasonably operate a conventional vehicle.” People who think otherwise must have spent their entire lives in metro areas with little to no weather. Moderate amounts of snow will certainly make L5 driving impossible until we have AI where reasonable people start thinking about giving civil rights to.

>Moderate amounts of snow will certainly make L5 driving impossible until we have AI where reasonable people start thinking about giving civil rights to.

The list of problems for which that sort of claim has been made but which have since been solved without anything resembling AGI is long enough that I'd tend to doubt new ones.

Re: Geofencing Will Enable L5

#73
I imagine that we will see the following:

1. Cars that are good enough to drive themselves most of the time.

2. Remote driving centers, run by the car manufacturers, that will have a human take over as soon as the car hits a spot it is unsure of. To the passenger, this is invisible and is considered self driving. (As long as the self-driving is good enough, one operator can handle a lot of cars. If there is an open standard, one remotely-led car can lead [or at least prompt] all the cars in back of it.)

3. Areas, such as urban centers, that only allow self driving vehicles and have no parking (as the vehicles can leave after drop off).

4. Improvements to self-driving in those areas, as they are geofenced, don't have human drivers, and have lots of cars hitting any given snag - all of which make debugging easier.

5. As problems are solved, and the pressure to move to self-driving goes up (due to said areas) the self driving part will handle more, and remote-control less, in ever widening circles.

Re: Geofencing Will Enable L5

#74
post #30

Earlier quoted context omitted.

OK, you build trains from Sidney, Montana, to Lewistown. I'm not interested in self-driving cars, but thinking trains solve all problems is a very narrow, urban view.

I don't think trains have to just be an urban thing. Maybe you wont get a train stop in McDermitt, NV or Loa, UT or Mackay, ID but there are plenty of places with higher density than that. It took me a long time to realize the history of the various interurban trails I've seen. This tweet I saw recently has some interesting maps of the networks of people focused rail to even small towns near Boston. https://twitter.c…

Before WW1, companies building housing developments used to also build tram lines so that people could commute to jobs in the city from those new homes.

Post-WW2, building suburbs without any sort of public transit was an additional way to discriminate against "the poors" as virtually all of those suburbs had race & religion restriction covenants in the property deeds.

Re: Geofencing Will Enable L5

#75
post #55

Earlier quoted context omitted.

so, trams?

I think they're describing something more like a PRT system: https://en.wikipedia.org/wiki/Personal_rapid_transit

Oh yes and maybe that was even in my head because I had heard about it before. It does seem like an area we could do much better. I was imaging something even more hybrid though. More like an elevator, especially the fancy ones where you select your floor before getting on, where it groups multiple people if they're taking similar trips. A lot of it seems like the sort of pure software problem that we should be able to solve easier now. You scan in on your phone or with a touchscreen at the station and then an algorithm works out which vehicles to dispatch to get everyone where they need to go which could involve lots of ride vehicles stopping together at station in the downtown, but then splitting off in the outskirts to drop smaller loads of people off

Obviously that gets a bit pie in the sky so maybe facing a reality of something simpler and more traditional public transit like is ok, but one can dream

Post reply on HN