Re:Consciousness and the Laws of Physics
by gus_massa on Saturday July 17, 2021 @02:22PM
(#27867684)
AlphaStar[1] is an AI program that plays StarCraft against humans. The second version, that has a limited vision of the map and a limited number of clicks per second, plays quite humanlike.
> * In order for the human animal to function, we need to understand the past, predict the future,*
AlphaStar can do that (or at least simulate that). It it has seen a invisible helicopter, it remembers that. It also understand that the opponent has build an airport to make the helicopter and invested to research the invisibility feature. So other research are probably delayed. Also it expect to see more invisible helicopter in the future, so it builds detector for invisible units. Does it understand the past and the future or only simulates understanding?
> understand the motivations of others
If it sees an airport that is researching invisibility, it probably guess that the opponent wants to make invisible helicopters, and want to use the helicopters to attack it's units or buildings.
> and come up with stories about the world we use to convince others and facilitate collaboration. Human society wouldn't be possible without consciousness,
AFAIK there is no multiplayer version of this AI, but it would be interesting to see one. It will be a difficult problem to choose how much communication to allow between to AI so they can play together, but not as a four-handled player. Also, how to interact with a human teammate. But it doesn't look impossible.
Re:Consciousness and the Laws of Physics
by skohan on Saturday July 17, 2021 @03:56PM
(#27868320)
I suspect things like this will converge towards AGI, and at a certain point, AI is going to be so embedded in our lives that whether or not it's truly conscious is going to seem like a pedantic distinction.
> * In order for the human animal to function, we need to understand the past, predict the future,*
AlphaStar can do that (or at least simulate that). It it has seen a invisible helicopter, it remembers that. It also understand that the opponent has build an airport to make the helicopter and invested to research the invisibility feature. So other research are probably delayed. Also it expect to see more invisible helicopter in the future, so it builds detector for invisible units. Does it understand the past and the future or only simulates understanding?
> understand the motivations of others
If it sees an airport that is researching invisibility, it probably guess that the opponent wants to make invisible helicopters, and want to use the helicopters to attack it's units or buildings.
> and come up with stories about the world we use to convince others and facilitate collaboration. Human society wouldn't be possible without consciousness,
AFAIK there is no multiplayer version of this AI, but it would be interesting to see one. It will be a difficult problem to choose how much communication to allow between to AI so they can play together, but not as a four-handled player. Also, how to interact with a human teammate. But it doesn't look impossible.
[1] https://en.wikipedia.org/wiki/AlphaStar_(software)