Thursday, November 21, 2024
8.7 C
New York

Grand Theft Auto Teaches Self-Driving Cars a Thing or Two

So, we all know that certain artificial intelligence systems have the ability to use deep learning techniques to self-teach itself and learn things quickly and efficiently. Well, now researchers have found that using some computer games can help this learning in self-driving cars and have employed the use of Grand Theft Auto to try and teach the intelligent self-driving motors a thing or two.





The realistic sceneries of Grand Theft Auto help to provide a fantastic learning platform for the artificial intelligence systems as includes real world situations. Using machine learning, the computer is able to identify faces and recognize speech just as a human can. Collaboration between researchers at Intel Labs and Darmstadt University in Germany bought about a technique in which extracted data from the game, Grand Theft Auto. They created a layer of software that sat between the game and the computer’s hardware and worked by automatically classifying all that it could see within the game (such as different objects and road scenes). The classifying entails labeling each item which can then be fed back to the computer via a machine learning algorithm and teaches it to recognize cars, people, and objects shown.

 

 

An image from Grand Theft Auto in which different elements have been automatically annotated.
An image from Grand Theft Auto in which different elements have been automatically annotated.

In a complex task such as automated driving, it takes thousands of hours to collect and collate data and is almost impossible to feed it every possible situation that may occur. Other research confirms the theory that games are useful in learning real-life situations and two scientists from the University of British Columbia have just proved that video games also provide an easy way to vary the environmental factors that currently exist within the training data. Ph.D. student Alizera Shafaei who is the co-author of the research says, “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. We showed that this synthetic data is almost as good, or sometimes even better, that using real data for training.”





More News To Read

Hot this week

Brooklyn Defendants Charged in Rideshare Hacking Scheme: Jailbroken Phones Used to Exploit Uber

Brooklyn federal court has charged two defendants, Eliahou Paldiel...

Detecting Defects in Next-Generation Computer Chips: The Future of TMD-Based Semiconductors

As technology advances, the demand for smaller, more powerful...

Merging Galaxies in the Early Universe: The Birth of a Monster Galaxy

Astronomers have recently observed a fascinating event in the...

Topics

Brooklyn Defendants Charged in Rideshare Hacking Scheme: Jailbroken Phones Used to Exploit Uber

Brooklyn federal court has charged two defendants, Eliahou Paldiel...

Detecting Defects in Next-Generation Computer Chips: The Future of TMD-Based Semiconductors

As technology advances, the demand for smaller, more powerful...

Merging Galaxies in the Early Universe: The Birth of a Monster Galaxy

Astronomers have recently observed a fascinating event in the...

NASA’s Roman Space Telescope to Uncover Galactic Fossils and Dark Matter Mysteries

NASA’s Roman Space Telescope is set to transform our...

Black Myth: Wukong – A Game that Gamers Love Despite Media Backlash

In a gaming industry increasingly influenced by social agendas,...

Gravitational Waves Reveal a ‘Supercool’ Secret About the Big Bang

In 2023, physicists made a groundbreaking discovery that could...

Related Articles

Popular Categories

Send this to a friend