The strategic awareness property would be an interesting one to measure. Which existing system would you say are more or less strategically aware? Are there examples we could point toward, like the social media algorithm one?
I don’t think that any current AIs are strategically aware of themselves. I guess the closest analogy is an AI playing ATARI games: It will see the sprite it controls as an important element of the “world” of the game, and will try to protect it from harm. But of course AIs like MuZero have no concept of themselves as being an AI that plays a game. I think the only example of agents with strategic awareness that currently exists are we humans ourselves, and some animals maybe.
The strategic awareness property would be an interesting one to measure. Which existing system would you say are more or less strategically aware? Are there examples we could point toward, like the social media algorithm one?
I don’t think that any current AIs are strategically aware of themselves. I guess the closest analogy is an AI playing ATARI games: It will see the sprite it controls as an important element of the “world” of the game, and will try to protect it from harm. But of course AIs like MuZero have no concept of themselves as being an AI that plays a game. I think the only example of agents with strategic awareness that currently exists are we humans ourselves, and some animals maybe.