The research is not just about beating video games with AI, but also about understanding how algorithms might better interpret the world around them. The researchers believe that these algorithms, which proved to be efficient at exploring all the nooks and crannies of video games, could also be adapted to make debugging code easier or to play through a video game to ensure there aren't any glitches.
Key takeaways:
- AI researchers are exploring how to give algorithms a sense of curiosity, enabling them to learn without any human guidance.
- OpenAI, in collaboration with researchers from UC Berkeley and the University of Edinburgh, found that when an AI algorithm is given a simple definition of curiosity, it can explore more than 50 video games and even beat some of them without any human-provided information.
- The AI algorithm was rewarded for seeing new things, which sometimes led it to die on purpose just to see the Game Over screen, or become obsessed with a fake TV and its remote, flipping through channel after channel to find something new.
- The research aims to understand how algorithms might better interpret the world around them, with potential applications in areas such as debugging code or playing through a video game to ensure there aren't any glitches.