The use and abilities of assistant AIs

An assistant AI is an AI which can be assigned missions to accomplish certain tasks which the player would like to do but either does not have the time to do or just doesn't like clicking.

This is a concept which is being indirectly discussed in the defining micro and limited base automation threads without ever being the focus.

the concept is that you would select a task/responsibility and a group of assets/budget and the AI would attempt to fulfill that task with the given assets.

For example.
local air superiority
the task is to maintain air superiority in a given area. the AI would, if given access to appropriate units, automatically scout the area and scramble superiority fighters to intercept hostile aircraft.
the resources assigned in this case would be either a aircraft plant and a resource budget (10m/s) or an aerial task force.

Build resource base.
The task is to build a resourcing base on a remote metal group with AA and shields.
the resources would be a group of engineers, an air transport and 20m/s.
The AI would respond by designing the base, shipping out the engineers, starting construction of additional Metal extractors, reclaiming metal baring rubble for surplus resources. etc.. the AI should intelligently build a base to the specifications using all of the abilities at its disposal.

Provide base upkeep
the task is to keep an already built base in good condition and safe.
resources: engineers and metal budget
AI would repair and reconstruct buildings reclaim rubble near base etc.. while giving priority to tasks which are more important. eg. base repairs over the assisting of unit construction.

Surgical Strike
the task is to destroy specific buildings within the enemy base
Resources. T3 bombers.
The AI automatically assigns an appropriate amount of bombers for each target and the appropriate attack routes and then carries out the orders to the best of its ability.

The Idea behind having an assistant AI is to get what you want done done as opposed to what you can order done. many times I've put aircraft into a patrol around my base when what really wanted them to do was just to scout and maintain air superiority. or times where i have built a template but the engineers do things one at a time and don't reclaim local resources, buildings don't fit because of terrain etc. or times where i have wanted to do a surgical strike but I couldn't manage the proper coordination of my forces and i just did it the lazy way.

the Assistant AI would allow you to skip those problems brought on by laziness and ignorance and get on with the business at hand.

I didn't say anything in the other threads similar to this, but now there is yet another "lets make an AI that plays the game for me"

OK so a lot of people have been bringing up the whole micro vs macro thing. And that TA, SupCom, and next PA are all macro games.

When a lot of people say, "I don't want to play a micro heavy game" they are talking about game play elements that are introduced to limit the player and put an emphasis on micro.

So for example in Star Craft 1 you could only put a small amount of units in a control group (12 I think, Warcraft 2 was only 9!) The only reason for this was to increase the amount of times you had to click to get a bunch of dudes into an enemy base. Then there were these battleships that shot smaller lasers on their own, but they had a special ability that would shoot one big laser that did a lot of damage with a long cooldown. To fire it you had to select the ship, select the big laser power, and click on a target. Now they could have just fired on their own at whatever they were engaged with whenever their cooldown was up. But in order to get the most DPS out of your battleships you have to keep clicking.

Now in Supreme Commander there are a lot of tools to help you fight large scale fights. Strategic Zoom, having a factory send new units on a patrol path, selecting as many things as you want, and giant laser death spiders that can fire their own giant death lasers without being told. These tools let you do macro management, but at the end of the day you are still giving orders to dudes, and still playing the game. We don't need to add AI that plays the game for us to have massive battles.

Amen Brother ZachB.

Mike

+1 ZachB

Until the AI gets enjoyment from playing the game I'll do that part.

Indeed ZachB /thread

If included, They're not forcing you to use the AI assistants.

Me myself would not see a problem with one or two AI assistants you could give side projects while you do other things.

I mean its guaranteed that the AI will do a worse job then you would do yourself so the game would still reward players that can multi task.

But it would allow me to focus a little more on strategy if i could order a AI to take care of one of my 5 planets late game. Or order a AI to make a farm base on an astroid early game while im focused on the land combat.

Plainly speaking, if you depend to much on a AI your gonna lose. But it would be nice to have a assistant to take care of side projects. (Theres no AI in any game that can win against a human that has played the game for a while, unless its cheating. Making a AI that is awsome in a legit way would be way to resource consuming for your computer and take way to much time and money for the developers).

Offcourse they have been talking about allowing multiple players to take controll of the same commander and its forces, the result would be the same but with humans instead of AI assistants.

(Edited some spelling mistakes and weird sentences, but the content remain the same)

Amen Brother Deloi.

Seriously, why are you people against an option that you won't even use anyway?
I want to play a Commander, not a Sargent with oversized squads. Building yet another economic extension with yet the same pattern that the eleven previous ones, assigning each bomber its target so they don't all waste their bombs on the same mass extractor, needing peaks at 2000 APM just to make coordinated attacks... You guys may enjoy doing it, but I'd like the option to avoid the high-APM low-reflection boring stuff, thank you very much.

If I wanted to play against the AI, I'd make a game on my home computer. If I'm playing online, I want to play against a person, not someone hiding behind a collection of AI assistants. People using this option online cuts down the pool of potentially interesting players, especially as someone using an AI assistant as a crutch will never learn the part of the game they're not playing, and so never get better at it.

There's a difference between "The game has a good UI so it doesn't take a lot of keystrokes to get it to do what you want it to do" and "The game has a button you can press to make it play itself".

You still would, those AI assistants would be tools just like everything else (The robots, the commander, the patrol features etc).

Theres no commander in the history of humanity that has done everything himself without delegation. Well maybe one that failed and is forgotten.

And since AI is worse then humans you wouldent be able to use the assistants on any main front or point of intresst unless you wanna lose. They are just for the work that happens away from the front lines that you tend to forget while doing 20 other things. It would for example be catastrofic to use the AI to tend your main base at the start, since any player worth his keyboard can take down a AI. Its just (for me atleast) for use in those extremly large games were you start to forget things and cant handle everything at once. Or possibly something to take over the defense on your main base while you visit the bathroom. Things like that.

Exactly. All it does is hold bad players back, because if you're any good, everywhere is a point of interest ('intresst', for the benefit of those in the audience who can't spell). Perhaps you'd like an animated paperclip to appear and say "It looks like you're trying to construct a firebase. Would you like assistance with this feature?"?

The game will most likely have some sort of ranking system for ranked matches so that complete newbies wont have to face those who are extremly skilled and have played for a long time. Therefor, if as you say, the AI will hold players back anyone that depends on them would end up in a lower ranking, and you would not have to face them.

Me myself belive that even pro players could have use for such a tool, you obviously disagrees.

Since (If what you say about them limiting players are true) you wont have to face them or use them i fail to see your reason for disagreeing to include them for those of us that cant multi-task as well as you. Not every future in the game has to be included for your convenience alone, there are other players. Not to mention that the top players make up a minority of the players using the game (But they can be considered more loyal supporters).

And im sorry for the spelling mistake, im sufficient in english to make myself understood and to understand what im reading, but it's not my primary languish (Im Swedish), so i do sometimes make mistakes.

how do you feel about the robots of your army shooting on their own initiative? would you rather their AI not control these actions.


the basic idea expressed in this thread is the ability to construct a unit that could be assigned tasks, ie. a sub-commander, it can carry out orders just like any other unit but on a larger scale, instead of the simple commands available to simpler units it would have a more complex set of commands related to construction (ie. construct resource base, construct production base etc...) or maintenance or unit production etc...

I don't like the idea of this type of AI that will make complex decisions about when and where to use groups of units (such as in your air superiority example).

What I do support is predictable, basic unit behaviours, such as auto-kiting for skirmishers and auto-chase for raiders, that can be configured to several states (the TA movement states were one version of this automation).

These are two very different concepts.

Instead of suggesting a bunch of AI maybe you should define what you want the game to be about. As in there is no point adding AI to automate a core part of what the player should do. Either it will play the game for you or be bad in which case nobody should use it. I suppose the option is generally good but I doubt I would ever use these AIs.

There seems to be an extremism problem with these AI debates. A lot of people seemed to pick either of these two sides:

• Anti-AI: Against just about everything more intelligent than in FA.
• Pro-AI: Supports whichever AI suggestion comes up no matter how extreme.

To me neither of these positions make sense.

It is hard to be consistently anti-AI because every command has some level of intelligence behind it and everything a player can do the game is effectively a command. I think there is a continuum of 'command smartness' which people are not appreciating. There is no sudden jump in unit smartness past the 'simple commands' which the anti-AI people are fine with. For example most people seem to think that area commands would be a good thing but they are a bit more intelligent than the FA commands.

The extreme pro-AI people are similarly nonsensical because there has to be something for players to do.

Think about what the player should do, what the game is about, instead of what the AI should do for the player. The remaining things can be either abstracted or AIed.

...thus reducing the size of the pool of skilled players and decreasing the overall quality of the player base. Like I already explained.

You seem to be under the impression that "better than the AI" and "Pro player" are equivalent statements. I wasn't even very good at SC (I struggled to get into the Top 1000), and even then the AI was a joke (and I had no trouble multitasking, despite not being a pro player). I suspect that you could be better than the AI and still be below the average.

I already addressed this in my first post in this thread. Sometimes it feels like people on this forum only read one word in ten.

Im sorry, that was a faulty assumption on my part. But most players that play games like this (Drawing from my experiance and logical assumptions of supreme commander and not any actual hard data) cant multi task good enof to handle everything in one of the larger games.

Me personaly would have aloot easier to handle complex strategies if i something like this was added. In supreme commander the strategies i could handle while handling economy, defense and offense of several bases cant be descriped as anything other then simple and crude. And these kind of games (TA, Supcom, PA) is supposed to be the real grand scale strategy games, right?

Unlogical assumption.

For all we know the AI assistant would make the game more accessible and increase the player base greatly therefor also extend the skilled player pool.

Theres not enof data to draw a conclusion that the "pool of skilled players" would be different (lesser or greater).

You get better at things by practicing them. These tools prevent you from practicing because the AI does all the work. So people who use them won't get better, and will never become good. QED.

Theres no reason to be good at multi tasking to this extent if there are possibilites for tools like the one suggested. Not to mention that this game will have multiplie playing feilds, most likely increasing the need for multi-tasking even futher then supcom.

Its like saying people dont need body armor in war, just more practice with dodgeing.

Its a strategy game, improving your strategy should be the priority. The sollution suggested makes it easier to do just that. Your capability to multi task should offcourse be important, but it shouldent be the top priority. Your capability to come up with strategies that defeat your opponent should.

(Small edit to fix some spelling mistakes)

while I like the idea of AI sub-commanders in principle, I don't think PA will have the scale (on a per player level) to make them a requirement. With a good UI you can and probably want to manage those tasks yourself (for fun and better efficiency).

to stay in the terms and scales used to describe PA, a game where you would manage multiple solar systems at the same time a "base" control mode instead of a unit control mode might be an interesting case. But that wouldn't be the same game type and not what is planned for PA.

~~~~~~~~~~~~~~~~~~~~~~~
on the general topic of AI, I still would like to see the option to run a game with just multiple AIs competing against each other. That would be nice to test different AIs, and just for testing/benchmarking servers in general.

i posted this topic because as we all know. AIs are stupid. thick as a yard of lard and have the creativity of a coprolite. but they also are very good and have excellent coordination for simply defined hard tasks.

I am a lazy bastard. i don't like to click very much when i can avoid it. but when i have 10 bombers and a whole bunch of targets i want to take out in order without wasting valuable bombers.AI job right there.

I want my base maintenance engineers to actually repair a base during enemy attacks instead of doltishly continuing to reinforce unit production. AI job.

build a base with 3 metal patches, a single constructor and a 5m/s in the optimum build order including construction of nano towers/factories/other engineers etc. AI.

Automatic coordinated bombing runs and scouting on base approaches. AI.

As i said AIs are stupid they have little creativity and when given a complex job with hazily defined parameters they will get confused and lose every time. but their strengths in order optimization targeting, maneuvering and strongly defined micromanagement both exceeds the capabilities of normal players dramatically and allows expert players to move to something more important.

Most simple commands which micromanagement takes care of relies heavily on pattern repetition, quick reflexes and optimization of movement all things which even expert players are too hamfisted to do with the same speed and accuracy as a computer.

I firmly believe that the use of assistant AIs to accomplish narrow tasks which require a lot of fiddly bits will increase the quality of competitive play and allow more units to function to their fullest potential more of the time than even the most talented micromanager could manage while at the same time allowing for more innovative and grand strategies.