GTA V Teaching Autonomous Cars?
GTA V is one of the simulation platforms that researchers and engineers are using to test and train AI to eventually drive us to work. Car companies are touting that there will be a no hands on model in 3 years but there’s still a lot to learn and a lot of algorithms to respond to. Which is where GTA V comes in, scientists from Darmstadt University of Technology in Germany and Intel Labs have developed algorithms that have been tweaked for use in self driving cars.
This is because there’s more time to do more things in the virtual world and cars never run out of petrol. It will never be a substitute for real world environments but the game is the game is the closest and most ‘realistic’ virtual environment to compare real world possibilities to. (With 262 types of vehicles, more than 1,000 different unpredictable pedestrians and animals, 14 weather conditions and countless bridges, traffic signals, tunnels and intersections.)
Most things are done through trial and error, and we’re spitballing here, exploring countless different scenarios and adjusting how the cars would react to said situations countless times!
It may seem out of this world but realistic video games are able to generate very similar data to what AI agents can obtain from the road. AI software has been using games such as Super Mario Bros/Angry Birds tackling problems in controlled environments, learning through simple trial and error.
Waymo (A self driving car company) uses it’s simulators to create an infinite number of situations that it’s engineers can muster up. Such as having three cars changing lanes at the same time at an assortment of speeds and directions, what’s learned virtually is applied physically and problems encountered on the road in real life is then studied in simulation.
So it’s not so crazy afterall.
To listen to the guys chatting about this please click here!