SymbolicFrank
Magister
- Joined
- Mar 24, 2010
- Messages
- 1,668
What would be your preference for great AI? To recap:
1. Meh. AI is no substitute for human players.
2. AI always loses against humans, so it has to cheat to provide a challenge.
3. An Impossible AI should be Impossible to win against!
4. The purpose of an AI is to lose gracefully.
Or, in other words: do you think an AI should scale to your level?
I think it should, and lose gracefully.
The main design error made in designing an AI is that it always takes the top-down strategem: an overall strategy is pushed down through the ranks.
While any General would want things to be like that, they're not. In real life, each and every independent actor takes their own decisions. And that's what a great AI should emulate: lots of individuals, all doing their thing.
Which is far easier than it seems like: in groups, people aren't all that different to ants. Simple.
But then again, when the AI should lose anyway, why spend money and resources on making a good one?
1. Meh. AI is no substitute for human players.
2. AI always loses against humans, so it has to cheat to provide a challenge.
3. An Impossible AI should be Impossible to win against!
4. The purpose of an AI is to lose gracefully.
Or, in other words: do you think an AI should scale to your level?
I think it should, and lose gracefully.
The main design error made in designing an AI is that it always takes the top-down strategem: an overall strategy is pushed down through the ranks.
While any General would want things to be like that, they're not. In real life, each and every independent actor takes their own decisions. And that's what a great AI should emulate: lots of individuals, all doing their thing.
Which is far easier than it seems like: in groups, people aren't all that different to ants. Simple.
But then again, when the AI should lose anyway, why spend money and resources on making a good one?