Zaleukos
Bum
- Joined
- November 4, 2006
- Messages
- 2,013
I realise I have the compulsion to turn every post I make about the TW series into a rant over the AI
At one level I think that from a players point of view one can dismiss the inner workings of the AI and treat it as a black box. A simpler model can produce as believable results as a more complex one (in the RPG world I think of Gothic with it's simple schedules vs the more dynamic Oblivion AI) and it's that output that matters. Getting that right is among the toughest challenges in the genres that I play (TBS and RPGs) and the difficulty of the task makes me fairly forgiving, but I at least expect to see some improvements from developers between titles Ultimately I guess one would want an AI that passes the Turing test, but that is pretty darn difficult for a game as multidimensional as TW...
Then I'm simply a bit jaded when it comes to (any) developers' claims regarding AI. In the TW case the AI behaviour has been fairly constant in all the games I've played, which is from Medieval 1 onwards. I also dont think the problem merely lied in the underlying model (even if it was flawed in the "divided" sense), but also in CA:s idea that challenging the player means constantly throwing hordes of whatever the AI has (if only peasants) at him, and at least in Medieval 2 the problems were made worse by horrible tuning. Changing the AI framework will not necessairly resolve all the issues.
That said I'll of course be very happy if the new model turns out to be an improvement in practice. Two weeks to a month of dedicated players dissecting the game should be enough to know if the game is less suicidal.
At one level I think that from a players point of view one can dismiss the inner workings of the AI and treat it as a black box. A simpler model can produce as believable results as a more complex one (in the RPG world I think of Gothic with it's simple schedules vs the more dynamic Oblivion AI) and it's that output that matters. Getting that right is among the toughest challenges in the genres that I play (TBS and RPGs) and the difficulty of the task makes me fairly forgiving, but I at least expect to see some improvements from developers between titles Ultimately I guess one would want an AI that passes the Turing test, but that is pretty darn difficult for a game as multidimensional as TW...
Then I'm simply a bit jaded when it comes to (any) developers' claims regarding AI. In the TW case the AI behaviour has been fairly constant in all the games I've played, which is from Medieval 1 onwards. I also dont think the problem merely lied in the underlying model (even if it was flawed in the "divided" sense), but also in CA:s idea that challenging the player means constantly throwing hordes of whatever the AI has (if only peasants) at him, and at least in Medieval 2 the problems were made worse by horrible tuning. Changing the AI framework will not necessairly resolve all the issues.
That said I'll of course be very happy if the new model turns out to be an improvement in practice. Two weeks to a month of dedicated players dissecting the game should be enough to know if the game is less suicidal.
- Joined
- Nov 4, 2006
- Messages
- 2,013