Self-driving cars are emerging as the Next Big Thing in tech, but striking out on your own can be difficult, meaning you’ll probably need a gig at Google, Uber, or a major auto maker to really dig in and make progress—until now.
The DeepDrive project came up with a unique way to let tech-savvy types tinker with autonomous driving tech—it repurposed popular video game Grand Theft Auto V as a self-driving car simulator. Companies like Intel have been playing around with it, but now it’s open source thanks to an integration with Universe, which emerged last month as “a software platform for measuring and training an AI’s general intelligence across the world’s supply of games, websites and other applications.”
The Universe integration with GTA V comes courtesy of OpenAI, a non-profit research company launched in 2015 by Elon Musk and others. “To use it, you’ll just need a purchased copy of GTA V, and then your Universe agent will be able to start driving a car around the streets of a high-fidelity virtual world,” OpenAI said in a blog post.
That “agent” was “trained via imitation learning on 21 hours (about 600,000 images) on the game’s AI driving,” according to OpenAI. “The baseline agent can drive in a variety of different weather conditions, react to traffic, and keep to its lane. This agent is a start, which we invite the community to improve upon!”
Those familiar with the game might wonder if the simulation will stumble upon prostitutes and gang members looking for a fight. Sorry, adrenaline junkes, the “integration modifies the behavior of people within GTA V to be non-violent,” said OpenAI.
Not to worry; there’s still plenty to explore minus the gun battles.
“GTA V gives researchers access to a rich, diverse world for testing and developing AI. Its island setting is almost one fifth the size of Los Angeles, giving access to a broad range of scenarios to test systems,” OpenAI said. “Add to that the 257 different vehicles, 7 types of bicycles, and 14 weather types, and its possible to explore a huge number of permutations using a single simulator.”