I have to disagree with you.
I know someone who studied IA things during his students years. He once told me (10 years back at the very least) that there are a lot of very potent IA models that are just not used by the industry. Why ? We know a bad IA doesn't make a game a failure (sales wise).

Hmm, have you ever played the release version of Master of Orion 3? The intergalactic conquest game with an AI that was incapable of invading a single planet, because it always forgot to put the soldiers into the transports before it sent them away?

I think you're both correct. More research into AI would be great, but the publishers currently don't even use the approaches that we already have.

Basically, the regular publisher's mindset is this:
- Developing a good AI is costly and much more risky than creating great graphics (or other assets).
- A good AI won't give us good reviews because reviewers don't play a game long enough to assess an AI's competence anyway. Great graphics, on the other hand, make for enticing screenshots.
- Players like to win.
- If the AI is too good, players won't have fun, and complain that the AI cheats.
- Therefore, an ideal AI is one that creates a convincing illusion of a challenge (rather than actually providing one), but lets the player win.
- Coincidentally, such an AI also happens to be much cheaper to produce than an actually competent one.

You mean the intergalactic spreadsheet simulator ;)

On another note, this discussion reminded me of this: [url=http://en.wikipedia.org/wiki/Nosferatu:_The_Wrath_of_Malachi]Nosferatu: Wrath of Malachi[/url]

Yeah, Ageod games have very decent AI. But I am not talking about wanting a challenging AI from Total War series, what I was talking about, the AI is apparently braindead even for TW standards, it's incapable of attacking your cities with armies, it never declares any wars, it goes suicidal in most battles etc etc.

It's obviously broken. You can see how much broken it is in Angry Joe's video (i personally never liked the guy, but his Rome 2 review is honest and it lists tons upon tons of problems, not only with AI.

In general, good AI is not a goal to achieve for most publishers. In FPS games, for example. We had great commando AI in Half-Life, we had nice Skaarj AI in Unreal, we had good AI in FEAR and first Far Cry and that's about it.

Most of the games don't even come close to what was already introduced in games many years ago.

And pretty much what Psyringe told - publisherd don't even want good AI, because most of the players would be asswhipped by it (probably including myself), and whiny bitches would flood the internet with "this game is cheating" rants all over the place. Most of the mainstream gamers don't want a great challenge, they want to win. There are tons of articles about "why so few people finish their games?" - finishing the game, winning it, it's the ultimate goal since arcade games.

This is why I slowly started to get interested in board games - first, the only limitations in intelligence is your and your friends capabilities, also, modern board games more and more rarely focus on winning, more on having a good time and adventure. Of course there always is a winner in board games, but winning, "completing" the game is not the main goal. At least not for me and most of my friends.

I must admit I preordered this game thinking that it is a Frictional game. I guess I should have done my homework then.

But I don't feel so bad, I'll make my judgement when I eventually play the game. I haven't yet fully played the Penumbra or Amnesia games either, but now I know I shouldn't necessarily treat this as an Amnesia sequel.

What interests me the most is what happened to the people who made Korsakovia? Clearly something happened to them, since Machine for Pigs is ridiculously underwhelming. All that potential squandered...