and there's me thinking that the maps just gave you directions from A to B in the most direct route
so surely, if the traffic is less now, the easiest route between two points IS a straight line ?
Machine-learning models used to direct the journeys of Google Maps users have been retrained to adapt to changing traffic conditions during the coronavirus outbreak. “Since the start of the COVID-19 pandemic, traffic patterns around the globe have shifted dramatically,” said Johann Lau, the product manager at Google Maps. “We …
Country lanes and side streets are shorter, but motorways and bypasses are faster.
Even ignoring traffic, roads have different safe speeds, and junctions different time costs.
The trick is to somehow get a sufficiently accurate estimate of those costs into the graph. Quite why that's AI or even "machine learning" is beyond me though. Measuring real data from real events and inserting that into a rolling average wasn't called that when I was learning about such things.
The other point is that those costs vary, with some predictability and some unpredictability.
It's worth avoiding a school at pickup time even if it adds minutes to the average journey time, but it's fine to pass it at other times.
If there's an accident on the quickest road, it's better to take a different route early to avoid the whole area.
With local knowledge, these decisions are trivial. When working it out for the whole globe, I can see why they'd outsource the learning to machines.
> I can see why they'd outsource the learning to machines
How can machines guess the school starting and closing times at [Far, Far Away]? It's still knowledge you either have or not.
Anyway, reading the statements of the Googlers I understand they just look at the traffic flow in the past weeks, average it, and forecast the future situation from that data (that's the "AI" part). Of course reducing the sampling window that much will yield some interesting artifacts around big, several days-long events (big holidays and such)...
Biting the hand that feeds IT © 1998–2020