What Is Up With WAZE And Google Maps?


I don't know about you, but up until about two to three years ago, I implicitly trusted Google Maps and Waze, the other Google owned mapping app. They were pretty bullet-proof as far as I was concerned. But then... things started to get weird and inconsistent.

It was one thing when you were near shopping malls or on college campuses where these kinds of facilities mucked with your reception, but all of a sudden Maps started sending me to ALMOST the right destination.

For example: I'm coming up on the east side of a mall next to an entrance. Maps then had me turn take the next right, heading west, then the next right, headed north, and then the first driveway into the mall, only to drive across the entire mall to be pretty much, right where I had started out from. The Maps of old would have had me pull in at that first driveway, not go all the way around the mall.

-A second example was a set of directions that had me coming up behind an auto parts store, but maps brought me to the west side of the mini-mall it was in, where there was no driveway. And the only option I had was to move forward and turn right, or away from my destination due to the one-way nature of the road the store was on.

-The latest Google Maps snafu was when I was passing through Santa Nella, Ca, and asked Maps for directions to Dixon, Ca. Maps showed me a route that approximately 35 miles longer than what Waze showed me. Maps was pointing me to Sacramento, then turning west, for a another 24 miles. While Waze was pointing me along the shorter route for the same amount of time.

WTF?

But Waze isn't off the hook either because at times, it seems to not have my best interest at heart.

My commute only has a few options, but every now and then it seems to point me in the slowest direction or route possible, even when it knows there's a better way.   There are even times when I ignore the Waze to take the right fork in the road, go straight, take my next turn, then rather than adapt, it tells me to take the next right and go back those last three miles to get back on the road I was on. But when I check the available routes, it shows me point blank that going straight, rather than back two miles is much, much faster.

These are only a few examples of what I've been encountering lately.

I remember a few years back when Waze pulled the option of if your preference was the shortest or faster routes. I had reached out to the Waze team to ask about it and they got back to me saying the algorithm wasn't working right. But you could not prove that to me. My app was spot on for showing me faster or shorter routes. I know, I used to swap back and forth flawlessly, getting exactly what I wanted.

So what the bloody hell is up with Google Maps and Waze?

My theory is that they started relying more heavily on their machine learning (ML) AI in the back-end of the app, that sits on their servers. Machine Learning is a term for AI-like algorithms that learns from feedback, and when they installed it, the app took a step back equivalent to several years, and the app had to start learning from a somewhat recent set of data points to start delivering drivers to their destinations.

The timing of my supposition goes hand in hand with when the app started asking if we, the user, was pleased with our guided map instructions. Which in a way, is one directy kind of feedback that you'd give a Machine Learning app... teaching it that what it just did was good or not.

And we, the humans on the other end, are the test rats for this ML.

-

Google, owned now by Alphabet Inc, (Another name for Google) uses GPS data from our phones and other sources that's supplied to their ML algorithm to figure out traffic situations. They add our GPS data (speed, location, etc.) to a historical database, which tells the AI, hey, historically speaking, traffic usually moves faster (or slower) than this.

When Google bought Waze, they used this additional data source to show drivers traffic scenarios. Plus Waze has this suicide mode, where they make it possible for drivers to input data about the traffic they're in, and with this additional data, provide even more improved traffic predictive data to Maps users.

-

The ML behind Maps is not the only place they've installed it, obviously. It's also in their search code, Gmail and other spots. Of course Google Assistant is also powered by ML.

At this point, I don't trust Maps like I used to. I either double check it with Waze (or the other way around) or trust my gut instincts.

The apps have taken a step back while they let the ML kick in and figure things out but I am sure that in time, the predictive nature of this digital beast they've started installing into Google Maps will get pretty smart and be back to it's trust worthy self once again, I mean, how can it not?

- -

Resources to check out:

ltd.edc.org - big data driving google maps

medium.com - how is google using machine learning

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Comments