Shouldn't AI really stand for Artificial Insanity?

A Canadian courier company who delivers for Amazon had two packages, one for me and one for my neighbour. We are in houses and share a double driveway right at the top of a "T" with our street being the horizontal part of the "T" at the north end.

The AI used by the courier directed the driver to deliver to my neighbour first. Then he was to drive a block west on our street, then a block south, then a block east, then a block north (on the vertical of the "T") and park in the exact same driveway and deliver my parcel.

Instead of a 10 meter walk, the AI forced the driver (they are monitored) to drive 800 meters and then do the 10 meter walk.

Is the "I" really "Intelligence" or something a bit less?

Since it's probably not notetaking related, I invite you to share this story on other social media such as lemmy:

Have a good day, cheers!

So sorry. Lots of people talking about AI here just thought I'd share my experience with AI. Won't do it again.

Byre.

You are in the correct place for off-topic conversations, literally says it in the category:

image

Seems weird to use AI for that kind of thing at all, I thought we had route planning down pretty well already...

Not sure if it's AI or just poorly planned deliveries. Maybe their software looks at each delivery in isolation instead of looking at where the driver is supposed to go next too. Although that's Amazon so you'd think they know what they're doing

This thread is bringing me nightmares of graph theory classes in uni.

I don't think you'd use AI for this kind of thing, pathfinding amongst nodes has been a thing for decades already, an AI would surely work out the optimal path is always as a crow flies, or maybe quantum tunneling, because it's both clever and dumb simultaneously.

More than likely it'll be silly things where the the weight (perceived effort, but not necessarily just "distance", e.g efficiency or ease of transport could be factored) between two places literally next to each other is just done wrongly, most likely because mapping takes effort and I'd assume it's more likely if AI was used, it's being used to convert satellite imagery into vectors that get used for the actual pathfinding, than it is to be part of the pathfinding itself. Did it detect some roadworks on the image that day? Those nodes are now disconnected entirely, despite being on the same street.

I would say when I'm using Google Maps in the car, it refuses to ever say "Do a U-Turn" unless there's literally no other options, and would rather detour you 50 miles if it could instead, presumably because the computer perceives a U-Turn as difficult or unsafe, but please Google, why!

Similarly having bought a new car and comparing the built in satnav with Google, the built in sat nav is basically always prioritising distance at the expense of ease, Google will happily let you take a minute longer if it knocks off 20 turns.

With all those little nuances, I can easily see how someone trying to build around API's provided like that could struggle to get good results when there's 100 stops involved.

I think if it tells you to go one way, and you go a different way it takes that into account and finds a new itinerary that's roughly in the direction you're going. My old car navsat wasn't like that - if I decided not to follow its direction it would keep telling me to go back and do a U-turn at the first opportunity. Google/Waze seems to be smarter or at least tries to be, and doesn't nag the driver as much

For me, it does instantly recalculate a route and let you keep going, it just picks routes that are significantly worse than going back to the exit you missed (even when that's incredibly easy to do). 50 miles is an exaggeration, but there's definitely a "take 3 lefts and go round a neighborhood" feel rather than just take 30 seconds to go back to the exit you missed.

I wouldn't be surprised if it was psychological. If you can avoid making the user feel like an "idiot" for missing the exit, it's less irritating than what your old car sounds like. "The Scenic Route TM". This could honestly be a big factor since an irritated driver is a less safe driver.

The issue is that U-turns are a problem in North America, because the laws are different from state to state or province to province. e.g. in Ontario you can make a u-turn at a traffic light, but not so in Alberta. And there are other restrictions depending on where you are.

To be honest, a lot of traffic laws make no sense in North America, mostly because they are meant to counter the ineptness of drivers. (This is not only the case for traffic laws, but many other areas as well. I have never seen so many illogical laws as in North America.)
Getting a driver's license in North America is way too easy. I grew up in Europe. Apart from the fact that a license there costs around $1200, you have to take a first aid course and learn even the basics of car maintenance.
Additionally in Europe there are roads where you have the right of way, either implicitly or explicitly. This is of great help, while in North America traffic comes to a crawling halt, because a busted traffic light means a 4-way stop. In Europe you just go (without reducing speed) if you are on the road that has the right of way. In fact in many cities they turn off the traffic lights (switch them to a blinking yellow light) at 1am (sometimes even as early as 10pm) until the morning.
There is also something called the principle of reliance. Which basically means you can rely on the fact that others know the traffic laws. It is part of the traffic law. This does not exist in North America.
These were just a few examples, but there are a lot more.

I've been living in Canada for almost 2 decades and I have also driven in many states in the US. I am not saying people in North America are all bad drivers, but you can only do what you have learnt. And unfortunately that's not a lot when it comes to driving a car.

Sorry for this, but it is the lounge category. :wink:

3 Likes

Not where I originally considered for a first post but the lounge it is.

About 25 years ago, before prevalent mapping, I navigated around a chunk of Europe with a simple Magellan GPS, its surprisingly large list of cities and towns with lat/long, and good paper map for other grid refs. It is remarkable how successful you can be with a bit of blu-tack to keep the GPS in place, showing an arrow ("try to head thataway") and a straight-line distance to target ("you're getting there!").

It would have solved the problem in the opening post. :slight_smile:

2 Likes

I'm increasingly apt to label whatever runs Facebook's automatic detection of "community standards" violations as an AU(Artificial Unintelligence), as there seems to be no intelligence at all to its results.

OTOH, I have had Google Maps (which the Swype-type keyboard on my phone wanted me to enter as Giggle Maps) tell me to make a U-turn after I made a wrong turn, sometimes repeatedly as I kept going past the U-turns it wanted me to make, when I've decided to take a different route than suggested.

You can, though, sometimes take advantage of having it automatically recalculate the route. One December, when I was going to go to a ham club meeting on the way home, I discovered that the route Google maps wanted me to take was an exit off of the top of I-285 leading to a major mall (anyone who has driven Atlanta's Ring-O-Stop- And-Go-Traffic will know immediately of which exit I speak), the meeting place being behind said mall. Due to it being December, said exit was packed full of cars moving quite sluggishly. I quickly decided no way am I going that way, and took the exit after that, letting Google maps figure out how to get to the meeting place from there. Saved me from an hour of (not) driving, I'm sure.

Speaking of phone keyboard apps, I sometimes wonder if there's a somewhat malign AI in the one I use. It's not unusual for it to repeatedly misinterpret my swyping, both in the primary suggested word, and in the alternates it suggests. When this happens enough that I instead revert to typing one letter at a time, I often find that I only have to enter 2 or 3 letters before it suggests exactly the word I want. It's as though it is saying "I knew exactly what you meant. I just wanted to make you do it one character at a time".