Google Maps just fucking SUCKS now

[Scott5114]

It's probably a form of AI, but (a) I disagree that it's slop, and (b) image analysis technology is not new in GIS.

I doubt they're doing image analysis or AI. It looks procedurally generated to me (the line does connect to a logical place, it's just wrong, which says to me that the routine that drew that line was probably casting just a bit too wide a net on places to draw lines).

[jakeroot]

It's probably a form of AI, but (a) I disagree that it's slop, and (b) image analysis technology is not new in GIS.

I doubt they're doing image analysis or AI. It looks procedurally generated to me (the line does connect to a logical place, it's just wrong, which says to me that the routine that drew that line was probably casting just a bit too wide a net on places to draw lines).

To be clear, I'm thinking the image analysis (a form of AI as far as I'm concerned, if not actually) is assessing the placement of the lanes using the markings and then, as you say, procedurally generating markings on top of that layer to the best of its abilities based on the earlier analysis. Unfortunately it seems to get things wrong a lot.

I would actually like to see them generate markings based on the actual markings on the ground, as opposed to using whatever rules are generated by their analysis and then their default markings applied to that situation. Eg, in states where transverse crosswalks are used, not using GMaps' default zebra style. That level of analysis might actually help them correct other errors...like yellow lines erroneously going through an intersection.

[jakeroot]

AI slop, I'd imagine.

Looks fine on my end:

It's probably a form of AI, but (a) I disagree that it's slop, and (b) image analysis technology is not new in GIS.

Well, as they say, one man's trash is another man's treasure. 

A lot of this stuff is trash, but I don't think Google, Apple, or any other mapping company has the wherewithal to maintain a database of road markings created entirely by hand. Given how often they change and how many there are to map, it has to be done using artificial tools of some kind, and then corrected by hand where possible. Or we just don't do any of this stuff. But I think people like it more than they dislike the AI shit.

[Scott5114]

It's probably a form of AI, but (a) I disagree that it's slop, and (b) image analysis technology is not new in GIS.

I doubt they're doing image analysis or AI. It looks procedurally generated to me (the line does connect to a logical place, it's just wrong, which says to me that the routine that drew that line was probably casting just a bit too wide a net on places to draw lines).

To be clear, I'm thinking the image analysis (a form of AI as far as I'm concerned, if not actually) is assessing the placement of the lanes using the markings and then, as you say, procedurally generating markings on top of that layer to the best of its abilities based on the earlier analysis. Unfortunately it seems to get things wrong a lot.

I would actually like to see them generate markings based on the actual markings on the ground, as opposed to using whatever rules are generated by their analysis and then their default markings applied to that situation. Eg, in states where transverse crosswalks are used, not using GMaps' default zebra style. That level of analysis might actually help them correct other errors...like yellow lines erroneously going through an intersection.

It would be far simpler to have the vector objects just include the number of lanes for a given stretch and then render markings accordingly. That's something you could do with the OSM dataset, for example. Of course OSM has had "image analysis" done too, it's just that the hardware used is nerd eyeballs.

[Rothman]

From north of Salina, KS:

[GaryV]

The Google Lady has lately been telling me to take a right at the next traffic light - ignoring the fact that the next light is for a school crosswalk where there is no street whatsoever. This has happened in 2 or 3 locations.

[freebrickproductions]

Apparently Google Maps has decided that Pigeon Forge is a more important city in Tennessee than Knoxville:
https://www.google.com/maps/@36.5108351,-82.8164586,6z?entry=ttu&g_ep=EgoyMDI2MDQyOC4wIKXMDSoASAFQAw%3D%3D

[kphoger]

I have no idea who is responsible for the change, but the entirety of Mexican federal highway 2—both the western and eastern iterations of it—is now labeled as 2D instead.  This means that, in locations where there really are both a free road and a toll road, both of them are now labeled 2D.

[NE2]

Apparently Google Maps has decided that Pigeon Forge is a more important city in Tennessee than Knoxville:
https://www.google.com/maps/@36.5108351,-82.8164586,6z?entry=ttu&g_ep=EgoyMDI2MDQyOC4wIKXMDSoASAFQAw%3D%3D

This may be true, for certain values of importance (e.g. tourists using Google Maps). But more important than Memphis?

[freebrickproductions]

Apparently Google Maps has decided that Pigeon Forge is a more important city in Tennessee than Knoxville:
https://www.google.com/maps/@36.5108351,-82.8164586,6z?entry=ttu&g_ep=EgoyMDI2MDQyOC4wIKXMDSoASAFQAw%3D%3D

This may be true, for certain values of importance (e.g. tourists using Google Maps). But more important than Memphis?

Even then, Knoxville being home to the University of Tennessee would still make it pretty important (along with being one of the four largest cities in Tennessee, IIRC).

[kphoger]

Along those lines, check out how important it thinks Sedona is compared to Flagstaff, AZ.  Heck, it seems to think it's more important than Amarillo or Lubbock or Colorado Springs.

[Scott5114]

I have no idea who is responsible for the change, but the entirety of Mexican federal highway 2—both the western and eastern iterations of it—is now labeled as 2D instead.  This means that, in locations where there really are both a free road and a toll road, both of them are now labeled 2D.

2D or not 2D, that is the question...

[vdeane]

This means that, in locations where there really are both a free road and a toll road, both of them are now labeled 2D.

Clearly those roads must be flatter than Illinois.  If they weren't, they would be 3D.

[NE2]

Along those lines, check out how important it thinks Sedona is compared to Flagstaff, AZ.  Heck, it seems to think it's more important than Amarillo or Lubbock or Colorado Springs.

Google should use control cities as a measure of importance. Move over Lubbock, here comes Limon!

[GaryV]

as a measure of importance

I think you misspelled impotence.

[JayhawkCO]

From north of Salina, KS:

The Overseas Highway does just kinda go on forever.

[kphoger]

From north of Salina, KS:

What?

[hobsini2]

Actually what is getting me more than anything is the mispronunciations of names of roads and places by Google maps. For example, there is a road in the southwest suburbs called "Gougar Rd". Google treats it like cougar when the correct pronunciation is Gow-ger.

[kphoger]

Actually what is getting me more than anything is the mispronunciations of names of roads and places by Google maps. For example, there is a road in the southwest suburbs called "Gougar Rd". Google treats it like cougar when the correct pronunciation is Gow-ger.

Voice directions have always been a bit silly, and I'm sure they'll always be.  Like pronouncing McLean as "em cee lean".

[hobsini2]

Actually what is getting me more than anything is the mispronunciations of names of roads and places by Google maps. For example, there is a road in the southwest suburbs called "Gougar Rd". Google treats it like cougar when the correct pronunciation is Gow-ger.

Voice directions have always been a bit silly, and I'm sure they'll always be.  Like pronouncing McLean as "em cee lean".

I haven't tried it yet but I would bet it screws up Oconomowoc and Waukesha. I know it screws up Wausau.

[GaryV]

The Google Lady finally found out how to say Dequindre a few years ago. I think she still struggles with Schoenherr.

A long time back we had to go to a place on St. Joseph Road near Lansing. She said it, "Street Joseph Road".

[kphoger]

Dequindre

Sounds like a popular name for kids born in the 1980s.

This is Janae and her sister LaKeisha.  And this is DeQuindre, who just turned eleven on Thursday.

Schoenherr

Sounds like someone had a stroke partway through the name.

Umm, excuse me, Mister Schoenthaler? Excuse me?  Excuse me, Mister Schoenherrrrrrrrrrrr  *thump*

[DenverBrian]

Actually what is getting me more than anything is the mispronunciations of names of roads and places by Google maps. For example, there is a road in the southwest suburbs called "Gougar Rd". Google treats it like cougar when the correct pronunciation is Gow-ger.

Voice directions have always been a bit silly, and I'm sure they'll always be.  Like pronouncing McLean as "em cee lean".

I haven't tried it yet but I would bet it screws up Oconomowoc and Waukesha. I know it screws up Wausau.

I bet YOU screw up Oconomowoc on occasion. 

[hobsini2]

Actually what is getting me more than anything is the mispronunciations of names of roads and places by Google maps. For example, there is a road in the southwest suburbs called "Gougar Rd". Google treats it like cougar when the correct pronunciation is Gow-ger.

Voice directions have always been a bit silly, and I'm sure they'll always be.  Like pronouncing McLean as "em cee lean".

I haven't tried it yet but I would bet it screws up Oconomowoc and Waukesha. I know it screws up Wausau.

I bet YOU screw up Oconomowoc on occasion. 

Maybe when I was 5. I grew up in Wisconsin. I know that one well. 

[kphoger]

I grew up in Wisconsin.

Then I bet you you screw up 'pillow'...