Artificial Intelligence (AI) should be improved before release of Normandy - Agree or Disagree

[RedKestrel]

Just now, J2_Bidu said:

In my opinion, with the last update the AI went braindead. This is probably a side effect from something else.

 

Example: an SE5 is on my very high 7. He dives all the way to my height, flipping upside down while at it, passes me flattening and heads perfect North unchanging while I fry him from his 6. Happened twice in a row on Flugpark.

 

I don't remember ever seeing this before.

If you take a look on the technical issues side of the forum you'll see that DED_Rapidus is posting there in a few of the AI and campaign related threads, and he mentions that they are starting some intensive work on the AI. He is the QC community liaison. They are gathering info about AI bugs. 

So my advice is when flying take tracks and try and save the mission file. Then go over and make a report in the appropriate thread or if what you see isn't covered in that thread, make an entirely new thread presenting your evidence (track, mission file, video if you have it). AI is complicated so the more evidence they have the more they can nail down the behaviour.

[photog95661]

A vote for supporting the creators of this sim would also be appropriate.  I trust the folks that run the sim to decide when to release Normandy.  Nobody will ever please everyone.  Sit back, consume an adult beverage and wait.  What we get will be usable and functional.

[J2_Bidu]

2 hours ago, RedKestrel said:

If you take a look on the technical issues side of the forum you'll see that DED_Rapidus is posting there in a few of the AI and campaign related threads, and he mentions that they are starting some intensive work on the AI. He is the QC community liaison. They are gathering info about AI bugs. 

So my advice is when flying take tracks and try and save the mission file. Then go over and make a report in the appropriate thread or if what you see isn't covered in that thread, make an entirely new thread presenting your evidence (track, mission file, video if you have it). AI is complicated so the more evidence they have the more they can nail down the behaviour.

 

Good point. I may still have those tracks. Will search, thanks.

[RedKestrel]

4 minutes ago, J2_Bidu said:

 

Good point. I may still have those tracks. Will search, thanks.

Be advised old tracks may not work on newer versions of the game. Probably best to always have a track going in case something report-worthy happens, then just clean out the  tracks folder every once in a while to keep it from getting too big. 

[Sky_Wolf]

If you look through the forums you'll find lots of customers complaining about how the AI does incredibly stupid things often and that they would like it to be improved.

 

I want the AI to be prioritized. If wanting to have the AI improved is a "beef", so be it. I am not saying the game sucks, it is the best. Just looking for better AI. 

 

AndyJWest, your argument doesn't make any sense to me. In any case, Jason said today that they have had a dedicated AI programmer for a while now (see thread below). Here's hoping the AI gets so good that people stop experiencing immersion-killing AI behavior and stop (or reduce) beefing frequency in the forums!

 

 

[AndyJWest]

5 hours ago, Pericles said:

AndyJWest, your argument doesn't make any sense to me. 

 

 

If developers doing what they have the skills for doesn't make sense to you, then I can only suggest that the problem is at your end.

 

 

Edited April 18, 2020 by AndyJWest

[BraveSirRobin]

6 hours ago, Pericles said:

I want the AI to be prioritized.

 

 

The odds of the AI being prioritized over releasing a new module are pretty close to zero.  I'm kidding!  It's zero.

[Lusekofte]

8 hours ago, BraveSirRobin said:

 

The odds of the AI being prioritized over releasing a new module are pretty close to zero.  I'm kidding!  It's zero.

 

[Sky_Wolf]

AndyJWest - I started this poll to see who wants the AI to be improved before the next release. You disagreed with the "premise" of the poll for the following reason:

 

AndyJWest wrote: "Different people in the development team are employed for their skills in different things. The person working on 3D models isn't going to be able to just switch to working on AI instead, and so on. And the only reason the developers can afford to employ someone to upgrade the AI is because they have an ongoing income stream from new content."

 

I then wrote: "AndyJWest, your argument doesn't make any sense to me." To clarify, your statement that people on the team have different skills and that the 3D modeling guy can't easily switch to AI has nothing to do with my statement that the AI should be improved. And implying that continued AI improvement is dependent on the release of Normandy is demonstrably false: there is an AI guy who consistently works on AI stuff. Unless 1CGS lives month-to-month on its income stream, you have failed to make a decent argument. 

 

You then wrote: "If developers doing what they have the skills for doesn't make sense to you, then I can only suggest that the problem is at your end."

 

You remain illogical. Your statement does not follow from what has been written here and from what we know about 1CGS (that they have an AI guy). 

[BraveSirRobin]

3 hours ago, 216th_LuseKofte said:

 

Do you know what “prioritized over” means?  

[Lusekofte]

43 minutes ago, BraveSirRobin said:

Do you know what “prioritized over” means?  

Do you know how to not being a patronizing prick?

[AndyJWest]

Ok, Pericles, if you can't understand my logic, answer the following question:

 

Do you want:

 

(a) The developers to employ someone else to work on the AI alongside the existing specialist.

or

(b) The developers to move someone working on something else onto working on the AI instead. Someone who isn't an AI specialist.

 

I can't think of anything else that the developers could do to expedite improvements to the AI, given that they already have someone working on it. Someone they had to employ for their specialist skills. Someone they apparently had difficulty finding. Either option is clearly likely to have a negative financial impact, which needs to be considered when discussing priorities. If you aren't proposing either of these options, what exactly are you proposing? 

 

I like to see improvements to the AI as much as anyone else. We have already seen some. From what Jason says, we will see more. How long such improvements take will depend on how complex they are, and on the skills of the programmer. What other people in the team are doing is likely to have minimal effect on how fast it gets done. Arguing that it needs to be done 'before the release of Normandy' makes little sense in this situation. 

 

 

 

Edited April 18, 2020 by AndyJWest

[Rei-sen]

17 hours ago, BraveSirRobin said:

 

The odds of the AI being prioritized over releasing a new module are pretty close to zero.  I'm kidding!  It's zero.

Then we'll have another lifeless module with defunct AI, but with some new planes. Yay, I'm so excited..

[Sky_Wolf]

On 4/17/2020 at 10:36 AM, BraveSirRobin said:

The AI was improved with the last update.  Problem solved!

You are in the minority if you are satisfied with the AI improvements in the last update and believe that no further improvements are necessary. But I don't imagine you actually believe that...

 

As for you, AndyJWest, to answer one of your questions, what I am proposing is that the AI improves more before the release of Normandy. 

 

So far the poll is 80% in favor of prioritizing AI over releasing new theatres/content. 

Edited April 18, 2020 by Pericles

[AndyJWest]

Stating that the AI should 'improve more' is an utterly facile response to my question. Your refusal to explain what you are asking the developers to actually do makes me suspect that you don't know yourself.

[BraveSirRobin]

5 hours ago, 216th_LuseKofte said:

Do you know how to not being a patronizing prick?

Right back at ya...

13 minutes ago, Pericles said:

You are in the minority if you are satisfied with the AI improvements in the last update and believe that no further improvements are necessary. But I don't imagine you actually believe that...

Whether I’m happy with the AI or think it needs improvements is completely irrelevant.  They are working on the AI all the time.  But they are never, ever, ever going to delay a module while they wait for AI that is “good enough” for you, or anyone else.  The sales of modules pays for that AI work.  So delaying sales to work on AI isn’t an option.  It seems that this should be so obvious that it doesn’t need explaining, but here we are...

19 minutes ago, Pericles said:

 

So far the poll is 80% in favor of prioritizing AI over releasing new theatres/content. 

All that means is that 80% of the responders have no idea how software development works.  That’s probably the most accurate poll ever posted in here.

36 minutes ago, Arthur-A said:

Then we'll have another lifeless module with defunct AI, but with some new planes. Yay, I'm so excited..

So am I!  BoN should be great!

[Raven109]

37 minutes ago, AndyJWest said:

Stating that the AI should 'improve more' is an utterly facile response to my question. Your refusal to explain what you are asking the developers to actually do makes me suspect that you don't know yourself.

 

That's simple, without going into detail, just look at what a good player would do under different situations (actually, just let me lower the bar, to make it "somewhat" simpler, and go for what an average player would do). Have that as a target for what the AI can do.

 

So, take one of the devs who are better at fighting in game, and try to model the AI after him.

 

Edited April 18, 2020 by Raven109

[BraveSirRobin]

1 minute ago, Raven109 said:

 

That's simple, without going into detail, just look at what a good player would do under different situations (actually, just let me lower the bar, to make it "somewhat" simpler, and go for what an average player would do). Have that as a target for what the AI can do.

 

So, take one of the devs who are better at fighting in game, and try to model the AI after them.

 

That sounds really easy.  I wonder why the devs never thought of doing that?

But just to flesh things out, maybe you need to provide more details.

[Raven109]

I never said that the devs did not think about it, that's your assumption. I was only providing an answer to Andy's query. He's the one who asked about what the developers should implement, so he's the one who can answer your question, @BraveSirRobin: "I wonder why the devs never thought of doing that?"

 

I'm sure that the devs have thought of many things, yet the discussion here is not about what the devs are thinking.

[BraveSirRobin]

Just now, Raven109 said:

I never said that the devs did not think about it, that's your assumption.

I only asked the question to try to get you to consider why the AI doesn’t work like “a good player”.  Apparently that didn’t happen.  So I’ll tell you why.  It’s a ridiculously complex problem.  It’s not “simple”.  That’s why.

[DD_fruitbat]

I know the world is all messed up right now, it has to be because i'm largely going to agree with BSR here.

 

The Devs know and are working to improve the AI, it has improved some since release, as much as i'd liked, no but thats what it is. It is however never going to make any business sense to prioritize ai over a new module, because those sales are going to pay the ai coder's wages. What i would like to see is a simplification of non 'pilot' ai though, to free up cpu cycles for more planes and objects in missions, but that seems unlikely to happen, so i just get on with ity

 

And if by some cunning and impossible trick the Devs made an AI that was just like a human, a huge proportion of offline players would just stop playing very quickly I suspect.

[Raven109]

No one said it was simple to implement the AI... again, I was only providing a simple answer to Andy's question. He never specified the level of detail he wants to go into when discussing AI development.

 

But if we want to get technical, the details are in the behavior of the target the devs choose to follow. So, for example, let's say they choose the graphics engine guy as a target, and try to model his play style. That model will offer the details.

 

Even if I answer the question with more details, you can just go, "ok, but what if"...  and we end up defining the entire behavior here, in a forum thread. I'm not here to prove anything, i'm not the developer.

[Sky_Wolf]

45 minutes ago, AndyJWest said:

Stating that the AI should 'improve more' is an utterly facile response to my question. Your refusal to explain what you are asking the developers to actually do makes me suspect that you don't know yourself.

 

First of all, it is highly presumptuous for any of us to go on talking about how 1CGS manages its operations - I have deliberately not done that. I started a poll about improving the AI before the next release (i.e. prioritizing the AI). The majority of the community seems to support that opinion.

 

But to indulge you: There is a dedicated AI guy now. So if he is working full days, full weeks, on AI, then I am happy with that. But there are serious problems with the AI and it has been this way since release of Stalingrad. So to answer your question...

 

5 hours ago, AndyJWest said:

Do you want:

 

(a) The developers to employ someone else to work on the AI alongside the existing specialist.

or

(b) The developers to move someone working on something else onto working on the AI instead. Someone who isn't an AI specialist.

 

I choose (b) - now even though this person might not be an "AI specialist", as long as he knows programming the specialist could manage him. But again, it is highly presumptuous for any of us to start talking about 1CGS management. It is enough to communicate to the devs that the majority of the community wants AI improved before the release of new content (i.e. that AI becomes a major priority). If that isn't possible, so be it. 

 

BraveSirRobin - you should change your name to SarcasticCowardRobin or something... it would suit you better. I'm not being sarcastic or sly, I'm being direct. 

 

[AndyJWest]

Raven109, I wasn't asking for specifics regarding which AI improvements should be worked on, I was asking how 'prioritizing' it was going to be achieved. Who would be made to work on it that wasn't already doing so?

 

Periclese, you should probably read The Mythical Man-Month. The idea that in software engineering you can always get things done quicker by allocating more people to the job has repeatedly been shown to be questionable. It takes time to learn how complex things (i.e. gaming AI) work, and teaching someone new how to do it necessarily slows down those already working on it. You might see a benefit in the long term, but short-term all you are doing is delaying things.

 

Perhaps we should just ask the AI developer to emulate comrade Stakhanov, and do eighteen hours work in every eight-hour shift, for the glory of the Motherland. ?

 

Smash the Fascist AI bugs!

 

Forward to the glorious future of robotic aviation!

 

 

[Raven109]

1 hour ago, AndyJWest said:

Stating that the AI should 'improve more' is an utterly facile response to my question. Your refusal to explain what you are asking the developers to actually do makes me suspect that you don't know yourself.

 

20 minutes ago, AndyJWest said:

Raven109, I wasn't asking for specifics regarding which AI improvements should be worked on

 

Ok, then, I must've misunderstood the 1st statement quoted above.

 

Yes, that's true, at some point in a project's development cycle, you will get delays by adding more resources on the same task. Are we past that point?

 

Regarding your pics, why are most of you taking every suggestion/bug report thread like it's an order to the devs, taking it to the extreme... they are mere suggestions, the devs are always free to ignore them, no one can force them to improve the AI. No one is asking for it to be done by tomorrow, not sure where the sense of urgency is being picked up from by some of you.

 

Almost each and every suggestion/bug report thread turns into a witch hunt of the OP: 

 

 

Edited April 18, 2020 by Raven109

[AndyJWest]

6 minutes ago, Raven109 said:

...at some point in a project's development cycle, you will get delays by adding more resources on the same task. Are we past that point?

 

I don't know. Though I do know that Jason has told us that he had to recruit someone especially to work on the AI. Which makes me think that expecting a programmer with no knowledge of the subject to pick it up quickly may be rather optimistic.

 

This whole thread is silly, since it seems to be based on the premise that things get done faster by making them 'priorities'. They don't. Getting things done faster involves making specific decisions based on the resources and information you have available. Since we don't have that information, and since Jason has told us that they already consider AI work a priority,  it seems that the only actual proposal being made here is that Normandy shouldn't be released until an unspecified something-or-other is done with the AI. And since the developers will no doubt have done something-or-other with the AI by then, there is no point in having a poll asking them to.

[Raven109]

11 minutes ago, AndyJWest said:

Getting things done faster involves making specific decisions based on the resources and information you have available

 

Yes, and this decision is up to the devs. But we, as customers, can make suggestions about what we would like to see in the sim. After all, the customer's needs is what drives a product's development direction, and I bet one of the reason this forums exists is for the dev team to get a feeling about what people want.

 

According to the poll so far more want an improvement in the AI department than a new module. It doesn't mean that the devs should start coding tomorrow. The poll just shows what people want so far between the two choices.

Edited April 18, 2020 by Raven109

[BraveSirRobin]

54 minutes ago, Raven109 said:

No one said it was simple to implement the AI... 

 

1 hour ago, Raven109 said:

 

That's simple, without going into detail, just look at what a good player would do under different situations

 

So now it isn’t simple?

 

the problem, of course, is that it’s basically impossible to get AI to do what a good human player can do.  There will always be “tells” that it’s AI.

 

Edited April 18, 2020 by BraveSirRobin

[Enceladus828]

I believe that improving the AI for IL-2 1946 and Cliffs of Dover “before the release of Normandy” as stated by Pericles is more of a priority than in IL-2 Great Battles.

[AndyJWest]

5 minutes ago, Raven109 said:

 

According to the poll so far more want an improvement in the AI department than a new module. It doesn't mean that the devs should start coding tomorrow. The poll just shows what people want so far between the two choices.

 

There aren't 'two choices' in the poll. 'Improving the AI' is already happening. So is work on Normandy. And the developers aren't going to hold up the release of Normandy just because 24 people (as of now) want to see something nonspecific done to the AI first. 

 

Sadly, some people on this forum seem to think that buying a product gives you a seat on the board of directors. It doesn't. 

[Raven109]

10 minutes ago, BraveSirRobin said:

 

So now it isn’t simple?

 

 

It's simple to answer Andy's implied question, it's not simple to implement AI. Not sure how this is related to a poll which asks what people want.

 

Andy's statement: Your refusal to explain what you are asking the developers to actually do makes me suspect that you don't know yourself.

My response: That's simple, without going into detail, just look at what a good player would do under different situations

 

I.e., it's simple to say what needs to be improving: take a human player, define a series of behaviors, etc...

 

Again, no one is attacking the devs... why are you turning all suggestion threads into this silly semantics game is beyond me.

[BraveSirRobin]

2 minutes ago, Raven109 said:

 

It's simple to answer Andy's implied question,

No, it really isn’t.  It’s like saying that the answer to traveling to another star is to build a warp drive engine.  Simple.   

[AndyJWest]

'Don't release new content until you have improved the AI' is a silly suggestion in the first place. 

[Raven109]

15 minutes ago, AndyJWest said:

 

There aren't 'two choices' in the poll. 'Improving the AI' is already happening. So is work on Normandy. And the developers aren't going to hold up the release of Normandy just because 24 people (as of now) want to see something nonspecific done to the AI first. 

 

Sadly, some people on this forum seem to think that buying a product gives you a seat on the board of directors. It doesn't. 

 

But there are two choices. You're assuming that this thread is about the devs, and what they can do. This thread, at least as I see it is about what people would like to see. We can't really say what's possible for the devs (because we don't have that information), but we can say what we would like to see implemented (because, we do have that information).

 

Regarding your last statement, about the board... that's miles away from what I said...  a product's success depends a lot on whether it is compatible with people's needs or not... whether you like it or not, it's how markets work. Demand and supply... There wouldn't be a niche sim market if there wasn't a need for it. 

9 minutes ago, BraveSirRobin said:

No, it really isn’t.  It’s like saying that the answer to traveling to another star is to build a warp drive engine.  Simple.   

 

Again, this thread is not about how easy it is to implement AI. You really like your semantics, don't you? Do you want me to start defining the behavior of neural networks just because I want better AI?! So if I want a red car, do I have to also define how to implement said car? When you wished for a Spitfire to be added to the game, did you define to the devs what a Spitfire means?

 

When it comes to markets, there are people who have needs (demand - customer) and there are people who can satisfy said needs (supply - dev). If the customer can define in detail how to do the stuff and also does it, then he becomes the supplier, and there is no need for the other suppliers.

Edited April 18, 2020 by Raven109

[AndyJWest]

 

There are no choices. The developers haven't offered us any, and they are the only ones in any position to.

[BraveSirRobin]

3 minutes ago, Raven109 said:

Again, this thread is not about how easy it is to implement AI.

I’m fully aware of that.  And your response about how you’d like it be better really has nothing to do with how it could be better.  It’s just silly grandstanding.

 

They have an AI programmer.  That person will make the AI better.  There is absolutely no chance that BoN is going to be delayed until the AI reaches a certain level that is satisfactory for the people who voted that it needs to be better before BoN is released.  Because that level does not exist.

[Raven109]

2 minutes ago, AndyJWest said:

 

There are no choices. The developers haven't offered us any, and they are the only ones in any position to.

 

You don't know that. The only thing we know is that there is a guy working on AI. If the demand is great enough, maybe they might reconsider and add another AI dev to reduce time spent on improving the AI.

 

Anyway, whether your stament is true or mine, it doesn't make any difference, the poll is asking what people want, it's not asking people to say what they think it's possible... Only the devs know what's possible and even that might change.

[BraveSirRobin]

2 minutes ago, Raven109 said:

 

the poll is asking what people want

‘Yes, they want the AI “better”.  There is already an AI programmer working to make it “better”.  Problem solved.

[[I./JG62]steppa]

More capable and less hardware taxing AI and all the benefits that can come from that could lead to more sales and  would increase my enjoyment of the game. But without new maps, aircraft artwork etc. it can´t attract new customers by itself. Both need each other.

[Raven109]

@BraveSirRobin, I'm sorry, but that's not what the poll is asking.

 

"Artificial Intelligence (AI) should be improved before release of Normandy - Agree or Disagree:"

 

It's asking the users of this game whether they'd want the AI to be improved before BoN is released. As you can see, it doesn't ask whether we think that the devs can do it, or whether the devs should start tomorrow if more people Agree than Disagree. It just asks whether we'd like this.

The devs can look at this poll, see what their milestones are, see if they have the necessary resources, see what benefits they could get from a better AI long term (i.e more people buying the game in the future), compare this prediction to the prediction about profits from releasing BoN.. etc... no one is forcing them to do anything via a poll.

Edited April 18, 2020 by Raven109