This feels like a massive leap from generative video to actual generative gameplay. The fact that Genie 3 isn't just generating pixels, but understanding physics and allowing real-time interaction at 20fps, is mind-blowing.
We’re looking at the future of rapid prototyping for game devs. I’m curious—how does the model handle object permanence if you loop back to a starting location?
Congrats to the DeepMind team on shipping this!
Replies
CalPal