The use and abilities of assistant AIs

Red Alert 3 touched base mildly on this idea of giving and AI a mild stance and then relying on his defence you set him to help you with or having him put pressure on an enemy while you set up or having him attack one side while you attack another.

I personally think:

1) It would be cool to have an AI with very fine settings, that can be set from AI=OFF to the point the AI plays the whole thing as if it was his own game while you borrow resources for your small assignments.

Gives newbies a crutch, not a hindrance, since they will have something to start them off while they build their own army waves and send them out.

Gives experienced players the extra set of hands only the multi-screens and built-a-ingame-script players have. They can set the AI not to play entirely, but to use so many resources to build certain waves of units to attack steadily or harass enemy field units outside of their own base safety, while you focus on continuing building or making a specific unit group for a specific task to do yourself.

I would only like it if it could be entirely turned on, entirely turned off, and only turned partially on to do base matinence while you are away conquesting or to constantly strike at the enemy while you build massive base expansions. Hell, the main thing I would see as a benefit is them playing for you on entirely other planets, because trying to fight simultaneous battles on different planets would be rather difficult.

2) Of course, the AI had better not be capable of major fallacies, like not listening to orders, leaving itself wide open on expanding, base power neglecting, resource overtapping, throwing away units (really it should either only be able to control individual units if it is units assigned personally to the ai and even then it should play hit and return with them, while any units you control and leave idle should be left untouched by the ai unless you "give" them to ai control)

Ahh the visions of epic games i get when i hear this. Two commanders trying to outsmart each other not worrying to much about anything else. Both sides spanning multiple worlds with multiple bases.

Im in total agreement, if they included AI assistants this is how i would like them to be.

I heard this game allow another player playing the game for you if you need to be afk for a while, if that is acceptable, having a AI play the game for you should be acceptable too, no matter entirely or partly.

Thats a feature for on-going super epic battles with like 40 players.
I am always talking about regular 1v1 ladder settings.

There is one more point against such heavy-AI-helpers:
They are just not easy to make. In fact if we look at what current AI's that play the game do, they are FAR away from anything that could be useful as an assistant. Developing an actually helpful AI would probably strain developing resources quite a lot and the result might still suck.

Hmm, sigh.

Yea its very unlikely that such a ideal AI assistant would ever be made unless the development team feelt strongly for it and planned it from the begining. The work hours used would be humongous. But we can allways dream. Maybe one of the developers is a super genius that reads this and feels strongly for it and develops the whole thing in a couple of days? :roll:

Explain. Why would I allow the incompetent AI the chance to mess with my stuff if I know I can do better than it?

The thing is, this system you're proposing won't even help in the way you describe, since the real bottleneck isn't giving orders, it's information processing, and there's no way your AI assistant is going to process information for you in a satisfactory way.

Name a conflict where a commander in charge of a few hundred units has been able to instantaneously convey precise orders to any of his soldiers regardless of their positions with no chance of said orders being intercepted by the enemy, misinterpreted, mangled, lost or forgotten, and who commanded an army with no logistics that doesn't feel pain and follows all orders with perfect morale regardless of how suicidal and can give and interpret these precise orders in a fraction of a second regardless of their complexity. You can't because no real life commander has ever been in command of an army that convenient, and real life command structures are created for commands being given in sentences to imperfect humans with imperfect information over imperfect communications channels, and have nothing to do with commanding an army of perfectly obedient robots with direct infallible communication links to their commander.

And when the answer of when you should use them is "never", which it will be because the AI will be bad, all this feature is doing is creating a trap option for new players to fall into.

"If you think criminals should be punished, then you must be in favour of the death penalty for jaywalking." <- This is your argument.

UI features which automate repetitive tasks (patrols) are different to ones which make decisions for the player (your AI takeover suggestion), this has been explained to you several times. I notice in later posts you claim they're the same thing and the difference is only a matter of degree, which is still missing the point, since units auto-firing on things in range and patrols don't involve any important decisions.

I was about to post something just like this, now I don't have to!

This also applies to Factory Repeat as well, as it only repeats what you ordered it to build in the first place.

Mike

Having some advanced AI assistants in a game is like having other players that can share controls with you, a game with them is more like some kind of team game which your teammates are AIs, you can't criticize this kind of game as if it is 1v1, if you want real 1v1, the game could have a option to disable them.

The second part explains the first part. And i disagree that the bottle neck is information processing. The bottle neck is being unable to easily focus your attention at several locations at once (It can be done, but it requires aloot of your concentration that in turn makes it harder to plan those grand strategies i want).

It has nothing to do with imperfect command structures and human being imperfect. You cant do everything yourself, it tears your focus to pieces and makes the end result less of what you could achive if you delagate some of the less important work. In every command structure delegation is necessary to reach a peak effectivness.

I disagree, the AI would help aloot, you dont agree with my examples so further argument would just be me repeating myself. But theres plenty of people here that agree with me. The only way to prove either of our standpoints would be if the included the AI assistants and we could try them out ingame.

Offcourse auto fireing at units at range and behavior in patrol routes are important decisions.

Some examples from SupCom: You have a radar stealthed base, when the turrets fire's at long range they might give the whole base away for example.
Or if your playing a 8 player game, your tactical missiles auto fireing at a enemy might start a conflict that you wish to avoid for the moment.
The patrol functions for engineers effects your whole economy by helping unit construction and repairing or following a trail of broken enemy units for mass harvesting that in the end results in the engineers death.

Every action in a game can be important depending on context, saying that the auto features that exist now dont involve important decisions is wrong.

Edit: Allmost forgot to mention that blasted long range artillery on auto fire. Leets say that you have a couple of long range artillery bases around the map, then a small enemy force invade your base, you quickly destroy them, but not before your own artillery has fired a salvo at your own base *facepalm*, the artillery hits some tier 3 power plants or your commander and WOW to fire those shots was one important decision.

The AI wouldent run free, as i said earlier it would be given missions, you would tell it what to do, just like you tell your units to "patrol from here to there".

I’m hearing a lot about 1v1, but tools for massive 40+ player battles will also be needed—why not find a solution that fits both game types? Imagine PA’s scale taken to it’s logical conclusion—battling across Jupiter’s 67 moons with 40 players (visual reference below). What tools will adequately command a play field of this scale?

And what about non-competitive play? I love huge skirmishes against the AI. In a massive 40+ AI player scenario without friends helping to command my force—I need tools to manage 30+ unique SupCom-esque battlefields simultaneously for the late-game. I don’t think area commands and patrol routes are going to cut it, and I’m not sure an AI assistant vs. opponent AI will be enjoyable.

My point is, what if we dream up more possibilities—and not worry about the details of the proposed ones. Let’s help the developers by putting are minds together brainstorming all the possibilities—and stop circling the first two obvious possibillities.

Put simply, I think that extremely large play arenas will be best suited to many players. One player vs. 40 AI players may be possible but it's unlikely to go well for that player under any circumstance. Even 3v1 or 2v1 AI can be fairly difficult in other TA lineage games when using a good AI.

Edit:

I'm into the general concept of UI innovations to more efficiently manage massive multiplanet battles, just like I'm in for the concept of robot hoverboards that double as comedians to entertain me while carrying me through the skies, but any proposals will need to be slightly more detailed than that to be useful in practice.

To clarify, I like single player vs. many AI’s in a free for all. 1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1v1 if you will.

It would be awesome if this were a practical option, but I struggle to see how a human will be able to manage 40 moons better than an AI.

Oh look, more people who are going to be severely disappointed when the game comes out and it's not of the ridiculously huge scale their imaginations have concocted. "30+ unique SC battlefields and 40 AI players" indeed. Maybe in 2018 when computers exist that can run games that big and not cost tens of thousands of dollars you guys can mod in some high level AI tools. As a feature at launch? Laughable.

I see a lot of people on this forum talking about "grand strategies" in very abstract terms, usually in the context of not wanting to play the game, but nobody ever says what they are or gives any examples of where they've been used effectively. It seems like the equivalent of claiming to be "an ideas man" in a creative field, someone with no talent but lots of ideas they think are good. Here's the thing: Ideas (or "grand strategies") are cheap, implementation is hard. You may want to see yourself as some kind of mastermind who directs battles through high level plots and schemes, but that kind of person only exists in fiction. You are not Yagami Light, Tzeench or Havelock Vetinari, and you'll never be. In practice, strategy is about both making plans and then competently executing them. If you think having to think about more than one thing at once is hampering your "grand strategy", maybe you're just bad at multitasking.

Oh no, I have evidence. Every Civilisation game has advisors and the option to automate cities and workers. A common piece of advice is to ignore all advisors and never automate workers, because you get better results by doing things yourself. If an established developer who has almost literally been making the same game with the same advisors for two decades can't make a good assistant, do you really think a team that's never made one before is going to do significantly better?

This is why the Hold Fire option exists. Also TMLs don't autofire in SC, so that example is irrelevant. When not to fire (Hold Fire) is an important decision, but the default action of firing on enemy units when there are enemy units in range is not, so the default behaviour is, sensibly, to fire on anything unless instructed otherwise. Units with ammunition do not autofire precisely because there is a cost associated with their firing.

Whether or not to set up a patrol route is a decision. Once you have acknowledged that the patrol route has been made, there are no important decisions being made for you.

You have already said that the AI would design bases for you. At least try to be consistent.

I like the idea of contextual commands depending on zoom level

In practice this might include unique commands available at the system, planet and usual SupCom scales.The more zoomed out you get the more broad your commands become—thus sacrificing specificity and gaining time from less clicks. Perhaps production at a planet scale could look something like this:



The specifics don't matter much to me, just the end goal: smooth gaming at large and small scales.

Making grand strategies arent that hard, if you dont have aloot of other things you need to keep track of at the same time. Do however note that "grand" is a realtive term, i meant it as grand in comparison to the tactics used today in RTS games, i dont see myself as a tactical genius or anything like that. But anyone of us would be able to create and use aloot more interesting strategies if we werent tied up in all the minor details.

I dont consider myself great at multi-tasking but im certainly not bad at it either, my favorite multiplayer rts of all times being Dawn of war and thats more in the direction of Starcraft with every unit having several abilites and such (offcourse the economy is simplfied in that game using capture points around the map).

You do rember that from my point of view the bottle neck is that the player cant focus his attention everywhere at once, right?

Then your using a turn based game as evidence that a human player in a real time strategy game wouldent benefit from a AI assistant? Uhm... in a turn based game you can plan and wait for as long as you want and be everywhere at once (since time is paused)... so your evidence isent realy evidence? Or did i miss something?

Okey ignore my TML's comment (Seems my memory is alittle fuzzy, now that you mention it i was like "Offcourse, why didnt i remeber that?"), but the same situation could develop from artillery fire.

I did mention that you dont have to use the theoretical AI assistants right? So we have a auto fire function on for example artillery that can make great decisions by itself that you dont have to use... How is that different from a AI assistant that makes decisions that you dont have to use?

Whether or not you give orders to your AI assistant is your decision. Once you have acknowledged the order has been made, theres no important decisions being made for you. You need to tell it what to do, it dosent do things automaticly.

For example, you tell the AI Build a standard attack force using this base and at a maximum 50% of my income within one hour. How is the work it does anything more then a slightly more advanced building que?

And how is that a important decision in your view, since you dont consider building layouts tool (Im including it since you said that the auto features that was included in SupCom didnt do important decisions, even if you didnt mention building layouts specifically), auto fire, and patrolling important decisions? You tell it what kind of base you want, you tell it where you want it, its basicly just a slightly smarter version of the building layout allredy in SupCom isent it?


Edit: Or defending your base while you visit the bathroom, isent that just a slightly more advanced and easier to use version of a patrol?

Edit2: Changed the text in last part.

And yet you've consistently failed to actually come up with any examples of these so-called "grand strategies".

How is it being realtime going to make the assistant not suck? It's hardly saving the poor starved attention of your little mind if you have to continually check up on it because you can't rely on it to make good decisions.

"You don't have to use it" is something you keep saying, I've responded to it in the same way, but I'm going to have to spell it out for you: Options in the game which are always a bad choice are a sign of bad game design. If the only saving grace of this so-called 'feature' is that you don't have to use it, then it's a bad feature and exists only to trap new players. Trap options are bad game design.

What the hell is a "que"? And whether or not to give orders to your AI assistant is not a decision because the answer is always "No" (Or, it's only a decision in the same way that whether or not to self-destruct your ACU at the start of the match is a decision). But you thinking there's such thing as a "standard attack force" shows just how little you know about the decisions you make in a game of SC. By telling the AI to build you an attack force, you're automatically handing over all the decisions about force composition to it, which it will proceed to make wrong because it doesn't know anything about what the enemy is doing or what you're doing. If you think there are no decisions to be made between deciding to attack a player in an hour and actually attacking a player, you're a bad player. And your example is of some horrendous play, since nobody plans attacks an hour in advance because situations never stay stable for more than a few minutes at a time.

If you don't think there are any important decisions in base design, you're just a bad player. There are dozens of tradeoffs to do with space efficiency vs volatility, pathing efficiency (In TA at least) build order for immediate benefits vs long-term benefits, both in terms of output and survivability, defence positioning, establishing a hard defensive line vs not bothering because of anticipated future expansion, positioning buildings to take advantage of any particular geographical features, locating priority targets relative to potential lines of enemy attack, et cetera et cetera. But no, none of these decisions are important to you, you just slap down whatever buildings wherever, which is what you'll get if you hand over to the AI. And before you cry "OMG PARAMETERS AND TWEAKING", any dialogue box with enough parameters to sufficiently specify the design of an entire base will have so many sliders, check boxes and drop-down menus that you may as well have built the whole thing yourself by the time you've set them all properly. And that half of those requirements will have changed by the time the AI gets round to them so you'll have to do it all again.

If by "the building layout" you mean SC's template feature, nobody ever used that for anything but surrounding things with pgens or storage for the adjacency bonus or surrounding PD with walls. You can hardly compare surrounding one type of building with another type of building to building an entire base automatically.

You never asked me to, you mentioned that you have never seen one mentioned, you didnt ask me to give you a example.

Thats another insult, your behavior seems rather childish dosent it?

I have explained my point of view on the use of Assistant AI's and theres been plenty of posters in this thread that agrees with me that they would be usefull, just becuse you fail to see the use of such AI and dont agree with my examples, dosent mean that there are no use for them.

Were in disagreement that it would be a bad option to use these AI assitants. The point is invalid in the argument since you wont accept my reason for there use and i wont see them as useless as you do.

The reason i mentioned that you dont have to use them is simply a response to your comment that you dont have to use auto fire on artillery (they have a cease fire button).

Hmm should it be "queue"? This is one of those times were my knowledge of the english languish temporarly failed me, resulting in a mistake. A set of actions after each other.

A "Standard attack force" would be a build premade for the AI just the same as a premade building layout that is a soft counter against any enemy army (a balanced mix of units). Im sorry if that the to me obvious fact werent as obvious for you. The hour timelimit was as said a example not something that i shouldent do in a game so using it to call me a bad player isent correct.

Ohh come on, your grasping straws now. As i said, its simular a more advanced version of the base layout tool, there could be a number of settings added to insure that the base you send your AI assistant to build would be acceptable to your standards.

Even if there were no such settings you could still send it to build bases of a lower priority were you weight the base effectivness of the AI's layout against the time saved that you can use for other things. (In a 1 vs 1 this would most likely not be necessary, but the game isent all about 1 vs 1).

Thats a lie, i use if often, i know of aloot of people that do. Incase you talking about the top ranked player what you say might be true, but they are still a minority of the player base and the game shouldent nessearly be developed for their sake alone.

You keep calling me a bad player, i disagree, but let's theorise that your right. Would me being a bad player automaticly make my comments invalid? Or features i would want in the game denied? No, becuse the game is developed for the entire player base. Not just you or those you consider to be good players.

Well, now you know you've been called out on this, yet still no examples.

You have already admitted that it would be worse than playing yourself. This is not in dispute. Unless of course you're just changing your position as time goes by so you can 'win' the thread.

So in other words it's something you should never actually choose, since the benefits of actually having an army that's capable of reacting to an opponent's moves are huge, and the benefits of a few seconds of time are insignificant.

As I have already stated, and you ignored, any set of settings that would be sufficient would render the time savings minimal due to the complexity of the problem. And saving a few seconds to get a crappy haphazard base in a "low priority" area is a pretty horrible tradeoff. You never know when a low priority area becomes important, so betting on an area always staying an uncontested backwater out of laziness is a poor gamble to make.

Why do you want a tool whose only use is keeping bad players bad? Do you just want a population of bad players that you can trounce easily?

The task they peform would be worse then done yourself, that dosent say that using AI assistants would make your play worse.

You just dont get it do you? They might build a slightly worse base then you would, but the free time gained that you can use for other things outweights the small performace loss.

The whole concept behind the idea is to lower the amount of multi-tasking so that you get this free time to increase the strategy part of the game and not be stuck building those secondary lower priority bases.

A few seconds can win or lose any game. So no a "few seconds" more is not insignificant.

Offcourse if you scout your enemies base and discover that he builds aloot of one unit type you could also tell your commander to build a different premade army layout, the basic force was a example. Its basicly a building queue that has more settings (maximum resources used, a timer).

Not if these settings are saved between games, you need to set them once to your liking, might even be able to do it in the options meny outside of games. Your mind is realy closed about this idea you know, you jump at the problems without even trying to see a way around them.

Yet again a invalid point, we didnt agree that the AI would make people bad players. This is just your assumption that you have given no proof of, just guesses.

Edit: Well im off for the night, but unless the thread is locked or you simply stop posting, il continue the argument the next time i return to these forums.