The problem is how to make an AI play nice with people on the road, who are inattentive, stupid, violent, vindictive, and frequently malicious.
And once you do, how do you test it?
Rolling it out on the road, with an operator in the drivers seat, is expensive.
Just the liability insurance would be insane.
Obviously, one solution, for the software at least, is to test it in a virtual environment, but this raises an important question: Where can one find a virtual reality that even comes close to mimicking the insanity that is humans driving cars?
Three Words: Grand Theft Auto:
Developers building self-driving cars can now take their AI agents for a spin in the simulated open world of Grand Theft Auto V – via OpenAI’s machine-learning playground, Universe.Admittedly, GTA, with its hot rods, weapons, and rampant crime is only a pale shadow of commuting in Boston,* but putting self driving automobile software through its paces in the fictional burg of San Andreas, is a truly inspired reuse of code.
The open-source MIT-licensed code gluing GTA V to Universe is maintained by Craig Quiter, who works for Otto – the Uber-owned startup that delivered 51,744 cans of Budweiser over 193km (120 miles) using a self-driving truck.
The software comes with a trained driving agent; all developers need is a copy of the game to get cracking. After that, programmers can swap out the demo AI model with their own agents to test their code and neural networks. Universe and Quiter's integration code takes care of the fiddly interfacing with the game.
Video games new and old provide great training grounds for developing reinforcement learning agents, which learn through trial and error – or rather, trial and reward when things go right. OpenAI's Universe was released in December, and is a wedge of open-source middleware that connects game controls and video displays to machine-learning agents so they can be trained in the virtual arenas.
*No joke: I knew that it was time for me to leave New England when I screamed at someone for NOT cutting me off in a parking lot.