Researchers are using Grand Theft Auto to help self driving cars learn

RESEARCHERS at Intel Labs and a German university have proved it’s possible to learn good habits from playing Grand Theft Auto.

Grand Theft Auto has a surprisingly positive use for researchers.

IT MAY be violent, it may be anti-social and it may be misogynistic, but it turns out there’s a surprising use for the notorious video game Grand Theft Auto.

The hugely popular video game franchise has seen its fair share of criticism. There was the time it courted outrage by including an update which allowed gamers to have sex with a prostitute through a first-person view. Shortly after a campaign led by sex workers succeeded in getting retail giant Target to stop selling the game in its stores because they said it encouraged players to “commit sexual violence and kill women”.

The game requires players to drive around, carrying out crime-related missions and terrorising the city and its roadways as they see fit. But no one will be mad about its new-found application.

Researchers at Intel Labs and a German university are using the game to teach self-driving cars to better navigate the road.

According to the MIT Technology Review, the in-game experience offered by the latest GTA can be used to “train algorithms” used in autonomous cars.

Machine learning techniques require huge amounts of curated data that can be immensely time consuming to produce. But the high quality graphics and environment of GTA can provide the computer with the information it needs to better perceive elements in the real world.

According to a paper published by the team, researchers created a layer that sits between the self-learning computer and the game which automatically classifies different objects in the GTA environment, such as sidewalks, buildings, pedestrians and other vehicles.

While there is plenty of sophisticated technology required to make the process work, the use of computer games holds the promise of providing a way to satisfy the huge amount of data required for powerful machine-learning algorithms to process and ultimately learn.

“With artificial environments we can effortlessly gather precisely annotated data at a larger scale with a considerable amount of variation in lighting and climate settings,” Alireza Shafaei, from the University of British Columbia told the MIT Technology Review.

“We showed that this synthetic data is almost as good, or sometimes even better, than using real data for training.”

So there you have it, it’s possible to learn good habits from playing Grand Theft Auto.

Who knew you it was so useful?

Originally published as Researchers are using Grand Theft Auto to help self driving cars learn