Black & White from Lionhead Studios released originally back in 2001 went on to great reviews, and won various awards. Sadly, it's one game that has also been lost to time but the openblack project plans to revive it with open source.
Demis Hassabis is a genius in the field of artificial intelligence, but the AI of Black & White is 20+ years old and runs on pretty limited hardware (by today’s standards).
Sure, you could probably write an awesome new AI for black and white but you could also write an awesome 3d renderer for the original Mario World.
My point was that the AI is really core to the game, and I am not sure how they would replicate it in an accurate manner. If you wrote a new AI it would be a different game.
What I meant was that it will not be that hard to replicate. It might feel complicated and deep but most likely isn’t that complex. Sure - without analyzing the original source code (which isn’t legally available as far as I know) - it might be impossible to recreate the exact behavior, but it should be possible to mimic the behavior - it might just take a few attempts to fine tune the input parameters.
That’s the thing - AI isn’t about size, it’s about categorizing the state of the world. If you can understand an action in context and possible responses, a Markov chain can learn and respond appropriately with the processing of a calculator - it really doesn’t take much
Demis Hassabis is a genius in the field of artificial intelligence, but the AI of Black & White is 20+ years old and runs on pretty limited hardware (by today’s standards).
Sure, you could probably write an awesome new AI for black and white but you could also write an awesome 3d renderer for the original Mario World.
My point was that the AI is really core to the game, and I am not sure how they would replicate it in an accurate manner. If you wrote a new AI it would be a different game.
What I meant was that it will not be that hard to replicate. It might feel complicated and deep but most likely isn’t that complex. Sure - without analyzing the original source code (which isn’t legally available as far as I know) - it might be impossible to recreate the exact behavior, but it should be possible to mimic the behavior - it might just take a few attempts to fine tune the input parameters.
That’s the thing - AI isn’t about size, it’s about categorizing the state of the world. If you can understand an action in context and possible responses, a Markov chain can learn and respond appropriately with the processing of a calculator - it really doesn’t take much