Microsoft’s new interactive AI “world model” still has a long way to go
1 min read
Summary
Microsoft has released its new World and Human Action Model (WHAM), which uses AI to generate interactive gameplay footage from sample videos.
It can infer understanding of in-game behaviour to create fully interactive videos based on video footage alone.
WHAM was trained on Bleeding Edge gameplay footage, a 4 vs 4 online brawler, alongside player inputs since the game’s launch.
After 1 million training updates, the model demonstrated an understanding of complex gameplay interactions, such as the flight abilities of a specific character and items exploding after being hit.
Microsoft tested the model by asking it to generate new gameplay sequences using real gameplay footage and simulated inputs.