Jump to content
The Dark Mod Forums

More Intelligent Enemies


goingsupersonic

Recommended Posts

  • Replies 78
  • Created
  • Last Reply

Top Posters In This Topic

I think that article perhaps misses the point that the purpose of most computer enemies isn't actually to kill the player, but merely to provide some challenge or interesting fun; they should provide enough difficulty to make the player think they're in danger, but not enough to likely kill them. Notice that in Starcraft, the singleplayer AIs function entirely different from the multiplayer AIs - the singleplayer AIs periodically throw survivable waves of enemies at the player to pretend to attack them, the multiplayer AIs actually try to kill them.

 

In any case, using genetic algorithms in games isn't anything new - I think designers have intentionally avoided it for exactly the reasons described above.

Link to comment
Share on other sites

Yes, we'll also be fully modeling human consciousness using a macroscale quantum computer based on Josephson Junctions in low-TC superconducting material with current loops. You'll need to connect to our central server to utilize the quantum computing resources though, and we require a credit card or bank account number as proof of purchase before this step can be completed.

Link to comment
Share on other sites

It'd be nice as an option for the toughest difficulty, better for replayability ... but I doubt devs are going to line up to implement something like this, with the heavy load of work, if it's not going to be used much.

 

By the way, my undergrad cogsci/AI professor (Risto Miikkulainen) at U Texas has been involved in this sort of project for years now, engineering this sort of stuff, or one version of it, for video games and the military, the NEAT project (Neuro-Evolution of Augmenting Topologies).

 

E.g., they have a demo game out called NERO (Neuro Evolving Robotic Operatives), which plays with putting NEAT into real time... you have to defeat little invading robot armies, RTS style, that get progressively smarter and learn from your defenses with each advance, and also a project for FPSs, right now apparently focusing on goal-directed navigation and, very cool I think, a platform for a uniform interface to a library of computer game environments.

Edited by demagogue

What do you see when you turn out the light? I can't tell you but I know that it's mine.

Link to comment
Share on other sites

This has been a well kept secret since the first stages of development of TDM, but due to this TC being close to completion, I've finally been given permission to share the one feature that will seperate this mod from all other video games in history:

 

Simply put, as soon as you create a FM TDM, every person who subsequently downloads your FM will also upload thier I.P. to my computer, after which, if that player is spotted by a guard and killed, I will go to their house, garbed in full medieval armor and long sword, and kill them (not before screaming something like "have at you, knave!!" or "taste my long one!!").

 

The feature has been cleared with the law enforcement agencies of over 30 countries, and just under 100 more are still pending.

 

Of course, a waiver is being drawn up as we speak. Downloading and playing a TDM made FM will be considered acceptance of terms.

 

Now THAT's realism!!!

 

PHEAR TDM BITCHES!!!111!!

Link to comment
Share on other sites

No need for that. If you remember, I proposed a strategy to achieve a similar effect way back, and it is not even hard to implement.

 

I'm not sure if that is sarcasm or not Sparhawk <_<

 

goingsupersonic:

 

The nature of this kind of technique is that it produces unexpected behavior, a cursory examination of the article reveals that they accidentally managed to create bots that didn't attack at all but only hid. Stuff like this is extremely time intensive to test and debug, especially if you don't have a fixed criteria or an easy way of grading the results - for example a guard that checks common hiding spots at random might quickly drive even dedicated players crazy.

 

The games STALKER and Fable are good examples of the problems that can be encountered, both made wild claims of evolving/learning opponents competing with the player but for various reasons fable delivered nothing, stalker looks to achieve only marginally more and even something as simple as SIN episodes dynamic difficulty shipped essentially broken.

Link to comment
Share on other sites

I proposed a technique where you can have self learning AI, which adapts itslef to the player. This means, that if the player is weaker the AI will also be weaker, while with stronger players the AI will also be stronger. If this is tuned well enough, then the AI could be always challenging enough but never fully beatable.

Gerhard

Link to comment
Share on other sites

I proposed a technique where you can have self learning AI, which adapts itslef to the player. This means, that if the player is weaker the AI will also be weaker, while with stronger players the AI will also be stronger. If this is tuned well enough, then the AI could be always challenging enough but never fully beatable.

This is what happens when people post ideas without thinking them through.

 

This is a sneaking game. How would the AI adapt to the player successfully sneaking? By becoming progressively more and more sensitive? That would suck to an ungodly degree. It would render ghosting almost impossible.

Link to comment
Share on other sites

Evolving A.I. I feel goes against basic gameplay mechanics period. The player, once the AI is encountered, doesn't know what to expect at first. After observing the AI (fighting it, whatever) the player learns the laws governing that AI, its movements, attacks etc. Then through processing this information they develop a strategy to overcome the AI. However, if the rules change, then the player will never get to the point where they can process the information and overcome the AI. The player stops trusting your game because the rules change.

 

An example of this is if you played chess. Except the pieces didn't have any rules attached to their movements. No, their movements were random. Sometimes the Knight would move in the 'L' shape. Sometimes a pawn. Sometimes the Queen would only move one space forward and the King like the Rook.

 

The game wouldn't be any fun. Likewise with board games. We're talking about games here, not real life. In games, enemies have certain movement points per turn, certain hit points, certain spells, saving throws, etc. The player knows how close they can get before being detected, they know how to react to an enemy. If all enemies were random in these fields yet seemingly the same, the player would never know how to react. One might say "duh...that's the point, they always have to adapt." But that isn't the point of a game. Without unchanging rules (the laws of the game) the game becomes unplayable.

 

The player will never be able to predict the outcomes of their actions and never be able to determine the best course of action. There is no mastery and there is no game. The player will always be caught and think to themselves "But the last guard didn't see me when I was that close to him!" This isn't any fun when you can't predict the outcomes of your actions. Suddenly it isn't your mistake for being caught (by not utilizing the best strategy). It isn't your mistake for losing the chess game against a pal.

 

It's simply that the chess game offers no strategy without rules and therefore, the player will blame the game, not themselves because there was no way for them to predict the outcome of their actions. In game theory, this is necessary.

Link to comment
Share on other sites

The player stops trusting your game because the rules change.

 

An example of this is if you played chess. Except the pieces didn't have any rules attached to their movements.

 

I don't really agree with your analogy. Making the AI smarter (whether by adaptive/evolutionary algorithms or whatnot) is not the same thing as changing the rules of the game. Likewise, playing chess against a more skilled opponent is not the same thing as the game pieces having randomly allowed moves every turn.

 

I guess I don't really see why people don't want smarter AI. If the AI get too clever for your skill level and the game is no longer fun, turn back the "AI cleverness" dial in the game preferences. It's the same thing as how online gaming sites pit players against other players of their own skill level.

Link to comment
Share on other sites

Ombrenuit is right, but we've probably all experienced what Komag is saying as well. It's important to be able to predict your actions in a game, but not necessarily exactly. In TDM we're including small random variables, so that AI won't act exactly the same all the time. They may not take exactly 25 seconds to walk down that hallway on patrol--maybe once they'll stop and look out the window, so it takes longer. Guards will have a slight randomness added to their acuity as well. You'll be able to predict--more or less--what you're up against, but like real life there will be some randomness to keep you on your toes.

Link to comment
Share on other sites

Ombrenuit is right, but we've probably all experienced what Komag is saying as well. It's important to be able to predict your actions in a game, but not necessarily exactly. In TDM we're including small random variables, so that AI won't act exactly the same all the time. They may not take exactly 25 seconds to walk down that hallway on patrol--maybe once they'll stop and look out the window, so it takes longer. Guards will have a slight randomness added to their acuity as well. You'll be able to predict--more or less--what you're up against, but like real life there will be some randomness to keep you on your toes.

 

I do like this. Predictable variablity is something that I think adds a unique element to the game. I was making the case that randomness can be taken too far, to the point where the rules of the game are undermined. Because variability in itself can become a rule. I see a window and as an experienced player I know there is a chance that the guard might stop and gaze out at it. As long as randomness isn't taken to an extreme, it isn't used to punish the player. I like tension when sneaking up on a gaurd because I might know that he may be more alert then another. Or that you can't "count" (a skill I've practiced a lot in Thief) a guards path to know the moment when to spring from the shadows and out the window.

 

But all game mechanics must have a perscribed purpose and stick to them: in this case it is for added tension because the player knows there is some variability, and with each degree of variability the player will react accordingly.

 

But it's funny how most people think of skill level and game difficulty as a linear relationship (as player skill goes up, the game difficulty goes up with it at the same rate). Then there is no mastery! Games like this fail because the player can never enjoy the fruits of their progression because the more they progress, the difficulty is constant! They never are rewarded for their labor.

 

They are afraid to open the next door because they've been taught by the game that it will just keep throwing harder and harder obstacles at them, there is no fun. Games should have a skill level and game difficulty relationship with valleys and peaks, in otherwords a modulating positive relationship. That there are sections of the game where the player enjoys mastery. Where the game gets easier for a while before stepping it up. Where the game rewards the player.

 

The best example of this for me was Resident Evil 4. The game was terrifying for the first half simply because you couldn't deal with what was being thrown at you when it was first introduced. Then as you went on, you slowly got better and better with your encounters (especially in the castle) with the Los Illuminados. Eventually there came a point (for me after the Gardens) where your fire power grew to the point where you could deal with those obstacles: this was the valley. The game became more action then scary because it wasn't about "How am I going to survive?" but rather, you could come up with creative ways of dispatching foes and dealing with them in general.

 

You didn't dread each encounter because you had the tools at your disposal. Likewise, the game progressed further and there was another peak in difficulty, followed by another valley. Developers need to reward players as they progress to encourage them to progress further. Games that are made to punish the player throughout will obviously not be completed by many. (There was a phase in the early 90s where developers would gloat about how few had completed their games. Is this something to be proud of?)

Link to comment
Share on other sites

Ombrenuits talks bullshit, because it's not the rules that are changing. The rules stay always the same. To use your analogy with chess appropriately it would mean, that once you have learned to beat an oponnent, you would never be allowed to play against a different oponnent who is stronger or uses different tactics. It doesn't matter wether I play against my daughter or against Kasparov, because the rules are always the same and don't change, just because I changed my oponnent.

Gerhard

Link to comment
Share on other sites

I think it's more like "reasonable human" measures, like "Guards can see me at this 'sort of near' distance...", where you have a good idea of normal human capabilities so you can hopefully judge the guard's capabilities at least close. When you've played so long that you know exactly the sight distances and sound thresholds of the AI, it becomes a computer game instead of a great experience.

shadowdark50.gif keep50.gif
Link to comment
Share on other sites

For the record, this debate reminds me of the debate in behavioral economics that "intellegent" actors, at least when they're doing strategic behavior, are self-consciously a little random (predictably random!) just so their competitor can't predict exactly what they'll do, while they can still be optimal at the same time.

 

This is one reason why Pavlov was wrong (all actions aren't s/r reflexes, arguably none) and why building some of this into video game AI makes for good gaming.

 

A slightly different debate here is on rule following and generativity (being "original"), which works for some types of rules like chess and language. Too much to explain here, though.

What do you see when you turn out the light? I can't tell you but I know that it's mine.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recent Status Updates

    • Petike the Taffer

      I've finally managed to log in to The Dark Mod Wiki. I'm back in the saddle and before the holidays start in full, I'll be adding a few new FM articles and doing other updates. Written in Stone is already done.
      · 4 replies
    • nbohr1more

      TDM 15th Anniversary Contest is now active! Please declare your participation: https://forums.thedarkmod.com/index.php?/topic/22413-the-dark-mod-15th-anniversary-contest-entry-thread/
       
      · 0 replies
    • JackFarmer

      @TheUnbeholden
      You cannot receive PMs. Could you please be so kind and check your mailbox if it is full (or maybe you switched off the function)?
      · 1 reply
    • OrbWeaver

      I like the new frob highlight but it would nice if it was less "flickery" while moving over objects (especially barred metal doors).
      · 4 replies
    • nbohr1more

      Please vote in the 15th Anniversary Contest Theme Poll
       
      · 0 replies
×
×
  • Create New...