Google Maps ‘perfects’ through DeepMind AI-model

Google Maps is one of the most popular product by the company. Thanks to its ability to predict traffic jams in our future routes, many of us utilize the app on a daily basis to ease the hassle during commutes. In fact, Google says that more than 1 billion kilometers of road are driven with the app's help.

And now latest blog post shows that Google Maps will get more accurate by implementing machine learning tools from Deep Mind. For those who are not aware, DeepMind is a London-based lab owned by Google’s parent company Alphabet.

In the blog post, researchers from both – Google and DeepMind explained how they take data from different sources and have it analyzed by machine learning models to predict the traffic flows. The data they collect includes live traffic information, speed limits, and presence of construction sites, along with the quality, size, and direction of the given road.

All this information is fed into the networks created by DeepMind, that picks out patterns in the data and uses them to predict traffic flow. Google claims that updates to its AI models have improved the accuracy of Google Map’s real-time ETA by up to 50 percent in some cities.

Regarding COVID-19, Google commented that they saw a 50 percent decrease in global traffic when lockdowns were imposed in early 2020. They also had to make updates to their AI models to make them more agile with the change.

Google further explained that the AI tools work by dividing maps into ‘super segments’ or clusters of adjacent streets that share traffic volume. Each of these is paired with another network that makes predictions for that specific sector. As of yet, it’s unclear how large the super segments are but Google claims that they are dynamics in size and are adaptable to the fluctuations in traffic.

They also make use of a special type of neural network known as Graph Neural Network, which according to Google is one of the best way to process this sort of mapping data.

Read next: Chrome 86 Will Introduce A New Feature Called Quick Focus Highlight Which Improves The User Experience When It Comes To Working With Focus
Previous Post Next Post