On one hand this describes the way forward, doesn't it? In order for transformers to truly beat humans at tasks, they need to be free to try these tasks in the real world.
If you do this in a game world, they can explore the game world. But what happens if you let them loose in the real world?
Is there a difference? Ie imagine a robot with only one video feed. No other sensors. Would a realistic game world be any different than the real world?
Our current tech falls short for simulations I’m sure, but I suspect the differences between real and simulation will revolve around the inputs the machine has available. Thoughts?
If you do this in a game world, they can explore the game world. But what happens if you let them loose in the real world?