An AI system had succeeded in mastering classic 80s video games, including Atari titles such as Montezuma’s Revenge, Pitfall, and Freeway. The AI’s algorithms could one day be used to help robots navigate real-world environments such as disaster zones, according to its creators.
Similar to disaster zones, many hard exploration games present a series of obstacles that must be avoided and paths that must be navigated in order to reach a goal. Previous attempts to create AI capable of solving such games have not worked out, due to the complexities of free exploration.
For instance, many AIs use reinforcement learning – which involves rewarding successful actions – in order to complete a task. The problem with this approach is that rewards tend to be very sparse, making it difficult for a system to achieve its objective.
The only way to solve this issue is by creating an AI that can actively navigate its environment. And to overcome these issues, researchers created a “family of algorithms” which they have called Go-Explore. In a nutshell, this system works by continually archiving every state it encounters, thereby allowing it to remember the paths it chose to discard at each point in the video game. It is then able to immediately return to any one of these promising saved states, thus overcoming both detachment and derailment.
Go-Explore was able to surpass the average human score on Pitfall, a game in which previous algorithms failed to score any points. It also achieved a score of 1.7 million on Montezuma’s Revenge, smashing the puny human world record of 1.2 million points.
Artificial Intelligence Learns to Beat Classic 80s Video Games
