How much would you pay for an improved AI upgrade

[sburke]

Let's put it this way. If BF provided or outsourced a match making service on a subscription basis how much would you be willing to pay? $5/month? $10/month?

Nothing, I'm already married.

[A Canadian Cat]

You can call me AI, send your checks directly to me

 

Great idea Al

 

That would be one tough AI though - we would have people here complaining that they cannot win scenarios no matter how many times they play them.

Edited March 12, 2015 by IanL

[cool breeze]

If you all are so willing to pay for smart AI.

 

I can solve your problem.

 

You can call me AI, send your checks directly to me and I will quite my job and will be available for your services each and every day.

 

I promise you to provide a high level of play and you will have complete control over any battle and situation you would like to play.

 

I even promise to never speak a word  so that you have that true AI feeling of not having to deal with another Human being.

 

Hahahaha  great idea some here

[Stagler]

If you all are so willing to pay for smart AI.

 

I can solve your problem.

 

You can call me AI, send your checks directly to me and I will quite my job and will be available for your services each and every day.

 

I promise you to provide a high level of play and you will have complete control over any battle and situation you would like to play.

 

I even promise to never speak a word  so that you have that true AI feeling of not having to deal with another Human being.

 

 

If I could play you at real time any time I saw fit then I would pay for this service.

[Jock Tamson]

64bit Windows release

What's the most RAM you have ever seen CM use?  If you have 64 Bit Windows the 32 Bit exe can use up to 4GB.

[slysniper]

Great idea Al

 

That would be one tough AI though - we would have people here complaining that they cannot win scenarios no matter how many times they play them.

 

Yah, there is only one problem to my concept. I think I would have to adjust my cost of living Budget.  It does not look promising with the $10 a day income I could possible make becoming AI.

 

It appears I might have to keep my present Job. (sooo Sad)  

[womble]

If BF implemented an AI like suggested above BF would be crucified. I would vote on the poll but the pricing is completely unrealistic. Even $60 is likely not going to get you much.

$60 would mean Charles and Phil working on nothing else than AI for the hours that they'd spend working on other engine improvements for any other family release, with the commensurate testing effort focused entirely on improving or teaching the thing rather than churning out new textures and models. I have confidence that they could give the pixeltruppen a few simple inputs that could produce some apparently-complex emergent behaviours in that time.

 

People make a good point about multi-core and 64-bit support. That should probably come first, to provide the horsepower to run and store the additional algorithms.

 

Still, this small sample of a small sample of the more-than-averagely-keen-on-CM does seem to indicate some willingness to pay the price of a new "family head" product or more to get AI improvements. Maybe using a crowdfunding pledge scheme to mitigate the financial risk by getting some of the money up-front would also give both a better idea of the broader player-base's views and give CM itself a publicity boost.

[Stagler]

A starter for ten would be putting the generic plans from CM1 back in.

 

The objective areas and starting areas marked out on the map would have to serve as a base for this AI to function on without a plan specified, and that must of been how the CM1 generic AI functioned.

 

Im guessing that a move to occupy objective areas was their primary directive.

The AI would have to do three things.

 

• Be able to split its force randomly into groups based on roughly equal point sizes.
• Know each category of unit - recce, infantry, mbt, support. This would be done by coding a unit category into each unit entry within the game config.
• Know the length of the scenario - this is specified in the QB menu, so that would have to be communicated that with the AI.

 

A convincing generic plan would be to split the force as stated above, and move in intervals towards objective areas, or the enemy zone if there wasnt objective areas specified (there always are in QB so this could be ommitted).

Recon would move forward in the first quarter of the game time. - the AI already knows how to call in artillery when your units are spotted.

Then MBT and Infantry category units into the objective areas in the "middle" of the game time, so the second two halves of the game time.

Support would stay in the starting zone i.e GBAD and on-map indirect fires for the entire game time.

A clear disadvantage would be the inability for the AI to pre plan fires as there was no fire zones allocated as there was no AI plan.

 

The set up above however would mean that when the QB AI picks its force each time it would be slightly different as it may decide to send each part of its split force to a different objective on the QB map. By hand picking the enemy AI force then further variety could be added. Also by changing the force to either infantry/mix/armour etc would add more variety.

[Skwabie]

For me $500-$1000... so my choice is not on the list. But since I'm possibly one of few, as economics 101 goes I guess even $1000 is far from enough. Lottery, here we go...

[Stagler]

Yeah 1000 dollars probably isnt enough, i realise it would take a significant investment for BF to step down the path of AI development, and probably not be able to develop other things in leiu, but it is the choice that they have to make whether to improve engine programming, or produce new content.

One would think that a robust engine with as much longevity as possible is required first.

[Childress]

Nothing, I'm already married.

 

Evidently not for long.

[womble]

Yeah 1000 dollars probably isnt enough...

Doesn't that rather depend on how many people are prepared to pay that much?

 

In the end, one or two people being able and willing to drop a cool grand on a game upgrade is worth one or two 60ths of  a thousand people being willing to part with $60. A very cool indeed million (i.e. if those thousand people all spent $1000) should be enough to pay for some improvement in the AI. I can't believe that it's "as good as Charles and Phil can make it regardless of further hours spent".

[sburke]

The description above is an interesting glance at just the most rudimentary AI and essentially not far removed from what is already in the QB AI configuration, but then look at how the suggestion breaks down unit "types". In only the most rudimentary form does that exist now. Vehicles and infantry. The suggestion is first to create group types that do not exist and then assign patterns of behavior to them. BF's programmers are not lazy. Take a look at the effort to introduce UAVs. It is cool they are in, but there is a lot given up in their implementation that runs counter to most everything else in CM. Adding different unit types and behavior and then applying AI behavior specific is far from simplistic. Just the work to alter AT and Inf gun behavior has been a difficult process and they still do not function at all in an attack formation, just one of the many weaknesses in QB behavior.

Anyone who thinks that this effort to produce even an incrementally better AI wouldn't bankrupt BF is just not facing reality. BF is not going to go years without income to add something that from what little the poll shows wouldnt even net the same amount as a base game for a new family.

Contrary to expectations, the BF team does have their own bills to pay, expenses of raising families etc etc. Our wish lists are not ever going to trump that.

[gunnersman]

How about another poll asking what features you are willing to do without while the AI is being developed? What are you willing to sacrifice?  It seems Battlefront is always coming out with improvements in each new game or module.  I cant imagine what is already in the works.  Although, personally I would much rather see an improvement on the LOS levels to allow for proper use of weapons such as the Krizantema or the TOW Stryker behind a defilade. Same for recon/observation vehicles.  Maybe even smaller action spots.

[Skwabie]

Time... sburke. Eventually with the growth of economy and tech, ppl will start to plug their central nerves into a virtual reality to play games.

A competent AI should happen sometime in between now and then.

 

Secondly I'd like to think there're outliers in the past. Because we don't want AI because we want AI. We want it because we want replayability and diversified and challenging gameplay. As in scripted missions only offers so much. Combat Mission: Campaigns almost became an outlier. There were other games that actually did. However the game industry then imploded. But that doesn't mean it won't revive again. Time. The tech is already there, unlike Matrix level sci-fi tech. Atm we're simply bound by limits of capitalism and quality of life, which will improve.

 

Thirdly, what do y'all think triggers are? Why when AI is mentioned one immediately associates with pixeltruppens acting like intelligent Terminators. Small baby steps count. Even in the hardest times.

 

All in all, try not be so short-sighted and negative, tis my notion.

[sburke]

Time... sburke. Eventually with the growth of economy and tech, ppl will start to plug their central nerves into a virtual reality to play games.

A competent AI should happen sometime in between now and then.

 

Secondly I'd like to think there're outliers in the past. Because we don't want AI because we want AI. We want it because we want replayability and diversified and challenging gameplay. As in scripted missions only offers so much. Combat Mission: Campaigns almost became an outlier. There were other games that actually did. However the game industry then imploded. But that doesn't mean it won't revive again. Time. The tech is already there, unlike Matrix level sci-fi tech. Atm we're simply bound by limits of capitalism and quality of life, which will improve.

 

Thirdly, what do y'all think triggers are? Why when AI is mentioned one immediately associates with pixeltruppens acting like intelligent Terminators. Small baby steps count. Even in the hardest times.

 

All in all, try not be so short-sighted and negative, tis my notion.

Actually agree with all the above, but note triggers are still part of the script. More triggers with different variables would be fantastic. For example right now triggers are based on actual movement of a unit. Would love to see for example a trigger based on spotting a unit or if/then options. I am sure folks could likely think of dozens more. Getting them though is a whole other issue. I have no idea what is feasible in the coding.

[A Canadian Cat]

A starter for ten would be putting the generic plans from CM1 back in.

Are you <several dozen attempts deleted because could be considered a personal attack> kidding?

The AI in CM1 was, well a joke.  OK perhaps that is a bit harsh - no, no it is. It always did the same plan and half the time the plan was not particularly good either. In a QB I suppose it could be made to work fairly well.  But if you factor in a QB with more than one scripted plan available the scripted plan method smokes it. In a scenario it just cannot complete with the scripted AI we have right now. A scenario designer can create AI plans that take into account terrain features, objective priorities etc. and match the forces chosen for the AI groups.

 

It is theoretically possible to create an autonomous AI that can do all these things but that is a crazy amount of work.  If BFC spent a fraction of that work in adding new features to the current AI scripting (additional triggers if/then paths) plus some editor improvements, scenario authors could create an even better experience at a fraction of the cost of your fantasy AI.  You are just dreaming and you are way off base.  Spend your time convincing BFC to augment their existing script based AI.  Since they actually want to go in that direction you will have more success.

[kevinkin]

Silly question(s) but how do polls get set-up and are there any past ones to view?

Kevin

[sfhand]

What's the most RAM you have ever seen CM use?  If you have 64 Bit Windows the 32 Bit exe can use up to 4GB.

 

I'm currently testing a battle on an 8x2 km map. In the force deployment 3D preview in the editor, on a machine with only 4GB ram (remember the x64 OS is using ~1GB), it crashed until I enlarged the swap file. After enlarging the swapfile it can run but is too choppy to deploy forces. So, in short, before any battle calculations were taking place it was using more than 3 GB but less than 4 GB. Add to that flying lead coupled with the proposed enhancement of the AI and future engine enhancements and you might see how one could be concerned about needing more memory. It is true that so far on my 64 bit machine with 16GB there have been no problems yet. It is also a non-issue for Macs since the Mac version is already 64 bit.

[Stagler]

Are you <several dozen attempts deleted because could be considered a personal attack> kidding?

The AI in CM1 was, well a joke.  OK perhaps that is a bit harsh - no, no it is. It always did the same plan and half the time the plan was not particularly good either. In a QB I suppose it could be made to work fairly well.  But if you factor in a QB with more than one scripted plan available the scripted plan method smokes it. In a scenario it just cannot complete with the scripted AI we have right now. A scenario designer can create AI plans that take into account terrain features, objective priorities etc. and match the forces chosen for the AI groups.

 

It is theoretically possible to create an autonomous AI that can do all these things but that is a crazy amount of work.  If BFC spent a fraction of that work in adding new features to the current AI scripting (additional triggers if/then paths) plus some editor improvements, scenario authors could create an even better experience at a fraction of the cost of your fantasy AI.  You are just dreaming and you are way off base.  Spend your time convincing BFC to augment their existing script based AI.  Since they actually want to go in that direction you will have more success.

 

QB needs to be fast. It takes a while to make your own QB map. With an existing AI programme, you could make a QB map, put an objective in it, and away you go in 15-30 minutes. The utility of QB is its speed, and it is part of the long term playability strategy of the CM series. There is only so many shipped missions and campaigns, and there will only ever be so many user scenarios, eventually interest dies or everyone will play the newest latest bestest battlefront thing and CMBS will get forgotten about, then all you really have is QB and the editor then, and some people cant/wont edit in depth.

 

You can call it fantasy as much as you want, Battlefront does not have this, other games do, it is obviously possible - albeit a large investment of man hours and their wages. Its up to them to choose their development path and whether they listen to customer feedback or not, but yes, I am just dreaming of a day where they do decide to put resources into creating a proper AI. Is there anything wrong with dreaming for better days? I don't think so.

Edited March 16, 2015 by Stagler

[agusto]

IIRC each of the stock QB maps has 8 (?) different AI plans so replayability is within acceptable ranges, at least for me. I play a lot of QBs vs. the AI because i often dont have the time to play 1-2 hours continuously and i want to keep the well made scenarios for when i have the time to fully enjoy them. The trick for an enjoyable QB IMO is to 1) manually select the forces for the AI and 2) pick the side with the inferior equipment for yourself and 3) pick a wide, open map so the AI can take full advantage of its superior equipment. In most of the QBs i play i am playing a russian Company (+) sized element vs. a US Company (-) sized force. Usually i pick a platoon of M1A2s + 2 platoons of Bradley infantry for the AI. The M1A2 is so damn hard to kill and the Javelins are so horribly deadly that, despite all the negative things you might says about the AI, i still regularily find myself in situations where i struggle not to loose.

Edited March 16, 2015 by agusto

[Stagler]

AI Ukranians are shocking in the attack. Just move forward and die in droves sadly.

Edited March 17, 2015 by Stagler

[BletchleyGeek]

We have the technology! But we dont want to spend alot of money...

 

As a practitioner, I can't other than agree with Stagler.

 

The technology is there ready to be used provided that 1) you have the money to pay for the skills, and 2) you have the money (or time, or willpower) to eventually and incrementally rework your codebase in order to integrate those algorithms and techniques.

 

Note the and - both things need to be true at the same time - and the eventually - a project manager, who needs to focus on the short term - paying the salaries and running costs every month, etc. - won't be very keen about such major surgery operations.

[dan/california]

As a practitioner, I can't other than agree with Stagler.

 

The technology is there ready to be used provided that 1) you have the money to pay for the skills, and 2) you have the money (or time, or willpower) to eventually and incrementally rework your codebase in order to integrate those algorithms and techniques.

 

Note the and - both things need to be true at the same time - and the eventually - a project manager, who needs to focus on the short term - paying the salaries and running costs every month, etc. - won't be very keen about such major surgery operations.

When every dollar you ever made as a company is a couple of zeros short of what the project would cost, the improvements in question cost might as well be on a moon of Saturn.  If you could build it, and that is a question mark, it might take a scientific computing cluster to run it.  If the Pentagon decided to rain money on BFC it still wouldn't go this way. It is just not happening.

Edited March 18, 2015 by dan/california

[Stagler]

As a practitioner, I can't other than agree with Stagler.

 

The technology is there ready to be used provided that 1) you have the money to pay for the skills, and 2) you have the money (or time, or willpower) to eventually and incrementally rework your codebase in order to integrate those algorithms and techniques.

 

Note the and - both things need to be true at the same time - and the eventually - a project manager, who needs to focus on the short term - paying the salaries and running costs every month, etc. - won't be very keen about such major surgery operations.

 

Well put.