We might need to have have the tank role added to every class at this point

Community General Discussion
301

This statement, while true, is so obvious its not even a discussion for me.

99.999% of cases inventing a new technology falls into reserch institutions. Companies usually dont invest anything in conceptual R&D. They just invest in their application. This applies especially to Pharma and Tech companies (which includes gaming companies).

Simply put, hiring a team of mathematitians with a super computer (assuming such a thing can be done with current hardware) to devlop a new AI nobody has ever seen or thought about before… so that Brann can stop being a looser… is a guaranteed money dump. Its such an obvious money loss that hoping for a better Brann is beyond Copium. Its religious fanatism at this point. Cause it wont happen. Ever.

As for reserch institutions. Why would they invest their time in a novel AI specifically for controlling NPCs in games ? You need to wait for a very, very gamer Uni Professor for that. And even IF such a person existed, because modding is illegal in WOW all his reserch and testing would be in illegal servers. Which is a problem on its own.

You know who you could ask for an AI that actually mimics people ? The botters. Those guys. Those are the only people that are activelly investing in AI technology to mimic human behaviour in WOW to prevent Blizz from catching them.

So at best, you can get a Brann that is very good at farming gold. :slight_smile:

IMO, if Blizzard really wanted to make a good Solo Experience they should not add any NPCs. They should either make 1 map per spec (like mage tower) or ban tanks and healers. Only DPS experiences and tune the damage accordingly.

302

I’m not convinced that this is the case, really. It all feels way too predictable for it being the result of AI. AI often leads to something completely random, weird, unexpected.

The followers don’t act that way. They follow the same route in the dungeons, every time, while performing the same tactic at every single pull and having quite specific action->reaction behaviors programmed. The most clever part is that they know where they are allowed to walk…and that doesn’t require AI in any shape or form, just a set boundary per zone.

Same counts for Brann:

  • Aggro → disengage
  • Still aggro → run away, swim away
  • I stand in goo → Yell that I shouldn’t do that

It may as well be dynamically scripted, perhaps pretending it’s AI as it’s a hot-topic lately. It’s what our marketing dudes have done as well, selling some (fancy) static scripts as “AI”. It looks like AI, but it really isn’t.

1 Like
303

Brann needs to go :dracthyr_nod: :angry:

304

No let bran heal my warlock demon. That guy got nerfed so hard he needs those heals

305

You dont seem to understand how AI works. Its easier to explain the power of AI with a counter example.

If you wanted to code by hand what the tank in follower dungeons would do… you would put a ton of bolean conditionals in your code. Sort of like this:

If you see mob, cast taunt.
If you have Shield Block available cast.
If you have >30 rage cast Ignore Pain…
ect…

And you could have 1000 or more conditionals like that. Or even nested conditionals (an IF statement inside an IF statement)… with is worse. For each and every one of the NPCs that are with you.

The problem is that ALL computer programs read that code one line at a time. So the question is, what is the optimal sequence of conditionals that will produce the best results?

AI is an algorithm in which you give it your desired “best result” and it can back-calculate the order of those bolean conditions. And once you have done that, you end up with a “script” of sorts that you can execute over and over again and it will give you your desired result.

And that is what you see. You see this “script” that is the result of the AI training. The results of which depend 100% on the training samples you gave it. And how you train it is the crux of the AI problem here.

You got two options: You can train it with real data (recorded player dungeon runs). OR… You can train it with synthetic data (an AI training with another AI). And you need 1M or more training samples to get a result. And the result of that training needs to converge to a result. If your algorithm diverges you will not get any results.

The problem with WOW is that you got 1 human playing cooperatively with the AI. So you cannot use synthetic datasets. And real player datasets are in insufficient quantity, and way too random and variable for the algorithm to converge into a good solution.

And that is why you get all that whacky behavior. You get some “good enough” script from the very limited training you give your AI. Its the best AI tech can give you with current technology.

TLDR of all that wall of text is:

It is predictable because of the limitations of AI training. But it is still an AI.

THIS:

Is diagnostic of a deterministic model. Or a badly trained AI. So either I am wrong and Brann/Follower Dungeons are not an AI and its a deterministic code. OR, I am right and it’s an AI, but a badly trained one.

Either way, it cannot be improved upon any further. Due to the impossibility of training an AI to play cooperatively with a human being.

1 Like
306

Yeah, sure, easy shot to make. A wrong one I might add.

I just think we just look at if from a different perspective. You seem to refer to their use of AI as in how they computed the static script. May be, may not be. We don’t know, do we?

They’ve got decades of non-AI experience of creating dynamic creatures within a certain boundary. That those creatures now follow you around in a dungeon or Delve doesn’t instantly mean that AI was involved, does it? It can be, I don’t know.

What we do know is that there is no on-the-fly AI involved. That would also be way too expensive (resources), so I doubt we’ll ever see that.

Don’t you think there’s a reason why Brann has relatively limited abilities ánd relies on a (hunter) pet? Those self-sustainable pets have existed since the day WoW released. That logic existed long before AI came into the general view. That part they could practically re-use one-on-one. That’s most of his damage, right there.

In this context is makes no sense to call it “still an AI”. An artificial intelligence, artificial intelligence is just a concept. It would be at best the result of having used an AI model.

How can you even say this? To be very blunt; we currently do not really interact with those following creatures in any shape or form. They follow us, and attack targets. That’s about it. This concept has existed for multiple decades and is not anything AI enabled in any shape or form. Of course this can be wildly improved.

But there’s is indeed a problem; let’s say they used AI. What data did they feed the models? User data? The average user is not so great, possible, maybe, but unlikely. Employee data? Doubtful. Synthetic data? Possibly, but then the use of AI kind of reduces for this purpose, as it would take quite some work to generate meaningful data to feed the models.

No, I’m not convinced this is the result of AI. They may have used it, for sure. But I don’t see much reason to believe so, if they did it’d have been a pure waste of energy (power)…something the world has no excess of.

307

That is what AIs do. It is an algorithm to compute a static script.

But the end result is still a static script. That was my point from the very beginning.

You dont interact with the following creatures. But the following creatures interact with you.

And you need rules for that. Which rules ? That is what AI solves.

User data is either 1 dude playing alone. Or 5 dudes playing toguether. Not 4 computers and 1 dude.

The data you need dosent exist. The one that is available is are the current follower dungeons. Which if used to calibrate follower dungeons would inevitably diverge (sort of like inbreeding).

Impossible to make. You can make an AI run a dungeon alone. With 4 other AIs. But you cannot factor in the random input of a player. So you cant use this.

And FYI: It is perfectly possible today to create 5 AIs to run dungeons, and they could run M+ dungeons better than the pros. Like 4 or 5 key levels more than what a human could do. But incorporate 1 person into the party and everything would break down.

I do. They either trained an AI with synthetic data, or with faulty 5 player dungeon run data. Producing a static script, that you then use to run the AIs in Follower Dungeons. But since its used for a purpose it wasent trained for, that is why when you put together 1 person and 1 AI you get those weird results.

But I might be wrong. Maybe they do use a static script for that as you suggest. And in that case, an AI would not produce any better results. For the reasons stated above.

308

The more I read what you type, the more confused I get. Your are convinced and talk like everything is a fact, ending in “I might be wrong”.

Either you work for Blizzard and know what they did and what they do, or you don’t. And in the latter, you may want to check some sentences and tone them down a little. Because currently, it’s easier to follow a hormonal teenager on a random rampage.

309

Because I am 99% sure they used AI to create the static script. But there is always a 1%.

What matters is if I am wrong, would my point be invalid or not?

And my point would not change: It dosent matter if they use an AI or not… it still dosent change the fact that AI cannot make a better Brann. And of that I am 100% sure.

I will. Thanks for the suggestion.

Hormonal Teen… I feel young for once. :smiley:

The reason I get so emotional is because I had this conversation so many times. “Make an AI to solve X problem” has become the “go to” answer to everything. And it’s annoying to have to explain the same thing over and over again in my job.

I got a bit emotional. Like a teen. Sorry for that. So its better just to leave the conversation here.

310

Eh, currently Brann (and the followers too, though) has so many imperfections that I’m convinced that AI as well as a programmer-intern can improve that bad boi. Preventing him from fleeing through the closed door at Zekvir would be an improvement. Pretty sure that anything and anyone can do that.

I know this feeling. I’m actually mostly against the use of AI…and the hype some get from AI, including the desire to shove it into anything and everything. 99% of the time it’s an utter waste of energy (human as well as electrical).

1 Like
311

IMO the best solution is to simply get rid of Brann.

WOW is not a game made for single player. And nothing short of killing the game and making WOW#2 will change that.

That was the mistake in my opinion.

NOTE for the solo players out there: My above statement is reserved for “difficult stuff”. Solo gameplay is perfectly possible at lower difficulties.

1 Like
312

Every mage tower challenge that had friendly NPCs was simply harder due to the NPCs being a burden.
Anyone who did the tank or healer challenges knows what I am talking about.
My experience with hard Zevkir was also affected by how often Bran would decide to be impaled by spikes and become unable to heal due to fainting.

2 Likes
313

Mage tower was a success because there was one map per spec. The challenge was catered to each spec. And that was fine.

But the moment you have only 1 dungeon for all specs… Not so fun as you say…

314

the challenges even then were easier on some classes over others

315

Could you pretty please take your hands of the content I enjoye?

Like how did we even got from we might need more specs that can tank to fight tank shortage to Delvs need to bee super easy content. And why is that even the case in your mind?

316

Why ? Delvers talking about the 16 Guided Crest cap in S2 are elbow deep into M+, the content I enjoy. Why should I not be able to do the same ?

Because people were suggesting hunter pets to tank. And that has a ton of issues (positioning, pulling, ect…). Someone suggested AI to fix that issue.

And the obvious answer to why that would not work is because the technology that controls Brann and Follower Dungeon NPCs absolutely would not be up to the task to tank in M+ keys.

Its all there. In the posts above you. Not so far. Its post #300 to #314.

If you have 1 instance. It can be a Delve. A dungeon. A Raid. Arena. Whatever you want. And you want to tune that so that you can be able to Solo that content with any spec you choose to do it with.

How would you balance out say… A BDK and a Warrior ? One has infinite self heal and the other one does not. But the instance is the same one for everyone. So mobs have to hit for X amount. The cant hit for X+10 in the case of tanks and just X in the case of a DPS.

But assuming you did. Lets say you do make “specially tuned Delves” for tanks. Lets compare 2 DDs then :

A Warrior with out self-heal. Or a Ret Paladin with free and infinite self-heal. Again. Its the same instance for everyone, so mobs have to hit for X. Cant hit X+10 for Rets and just X for Warriors.

And that is a simple case. Lets put in Ranged vs. Melee. Armour types (Plate vs. Cloth). Shield or no shield. Or ranged between themselves : Frost Mage (with infinite slow and kite ability) vs. an SP with nothing at all (not even a kick)… Pet classes and non-pet classes…

And then we can get into the burst of each class. CD availability. It it has BL or not… ect…

So it gets complicated really, really quick. And as you can see, it is mathematically impossible to guarantee the same quality experience to every single class in the game. You just cant.

Its already difficult in Raids with 20 players. Its borderline in M+ with 5 players… in Solo… its mathematically impossible.

The only way to guarantee a fun experience for everyone is (A) make it super easy. Or, (B) make 1 instance per spec. Which is what they did in Legion with the mage tower.

So in other words with a simple example :

If not every spec in the game has a kick, then…

  • You can remove kicks from the instance (no kicks at all) which would be option (A)…

OR…

  • You can make 1 instance with kicks for specs that have a kick. And another different instance for specs that do not. Which would be option (B)…

Its realistically the only 2 options that would work. Because relying on Brann to cover for what should be done by another player… well… you are a Delver. You should know if he is up to the challenge.

Think about this. Because Delves technically are not Solo. Mage Tower, the BFA nightmares and Torghast were Solo. Delves are a 2 man instance.

317

I don’t get it what does that have to do with m+ I thought in season 2 we are now just able to get crests from delvs

Oh that whole discussion. I didn’t read all of it I’m really not the guy that knows anything about ai.

I mean yes technically they aren’t solo that’s true but the are solo enough. I can do them without another player. Honestly it was a good idea to implement brann to cover for what certain classes lack.
It works quite well in my opinion and with him getting a tank spec next season he has everything he needs.
Sure he is not perfect but it’s better then dulling the content cause not every class has the same toolkit and I don’t think blizzy is willing to put in the work to fit every delv to every spec.

I have to say I did like mage tower but I think it was to hard when it came out but to easy to outgear later. Having different difficultys is a good thing.
I did not like nightmares cause I was simply not allowed to play them as much as I wanted and then I just stopped playing them all together.

Thorghast was not bad and probably the best solo content right after delves. I eas not the biggest fan of the powers tho. Sometimes you felt like you where a god and it didn’t really felt like it was my character progression that made me better but the luck of the draw.

Overall they tried different forms of solo content and I think delvs are on a good way. Surely we will never have the perfect content be it for solo or group content but right now I think we are on the right track (I can’t speak for m+)

318

So basically Delvers complain that they have to farm M+ to get Guided Crests to upgrade the Hero gear that drops in Delves. Which is a fair complaint. And that is why Guilded Crests will drop in Delves.

However, the whole discussion is that a Delve 11 is orders of magnitude easier than a M+8. Especially early on in the season. And that is why Blizzard decided to put a weekly cap on how many (max 16 I think it was) you can get in Delves.

And they are in a pickle. Because the whole point is to give Delvers the freedom to not do M+. But the price that would need to be payed is to force M+ dungeneers do Delves (a content they dont want to do). Its a catch 22 situation as you can see.

So Delvers are literally saying “screw those M+ swetloards and their crummy game mode”. Not all Delvers. But some.

Oh for sure. 100% correct.

Brann does do an “OK” job at replacing what should be done by a player. But he has limitations. He cannot fully replace a real player. Unless he is OP. Never mind replace a party of Tank/Healer/DD, the Holy Trinity. The most efficient group you can make in WoW.

And the logic about class balancing also applies to “2 man” content. This phase I said:

Is sort of like a formula. The smaller the group, the more difficult it gets. That is the reason why Tanks are barred from playing in Arenas for example. Arenas are a 2/3 man content. And DD/Tank balance there is impossible.

And I would argue that DD balance between themselves is also impossible. Hence the absolutely horrible state of balance PvP arenas are always in. And nothing Blizzard does seems to fix the issue.

So to achieve a proper balance, Brann needs to be OP and solo the Delve on his own. Which would be Option A (nerfing content) with extra steps. And if you want to add a bit of difficulty (in other words, players doing stuff) then Brann has to do less things. And we are back to square 1 with an impossible balance.

319

Ah Oke that sounds like a problem. I’m not a good person for that discussion tho.
I firmly believe if someone feels “forced” to do some content cause it’s the fastest way to achieve what they want they are already in a be as optimal as it gets mindset and at that point they aren’t part of the casual mass.

Right now Delvers are “forced” to do m+ and m+ players kinda say screw you guys.

If I had to decide and I am not biased even tho I am a delve player since I don’t care about gear.
I would say give Delvers those crests but put a limit on it but make it a generous limit. Something you really only hit when you farm selves every day for 3hours. Something like that.

The normal delver will never hit that cap and let’s be honest it does not matter how smal the cap is. The second there are crests to be get m+ players will do them cause they want to be better and better. That’s what playing that content does to you and I don’t mean that as a bad thing. I gues they don’t want delves to be to rewarding just so they don’t “need” to do content they don’t actually want to play

About brann.

I think the way he is right now is pretty much perfect I don’t think he is useless (even so his butt is always in the way to pick up stuff) but he is also not to strong. It feels like a 1.5 person experience right now.

When I take my wife with me who is a healer I never dip even the tiniest amount but with brann I still have to watch how much I pull and that I get those flasks. I love those flasks by the way I wish my wl could just throw is stone s everywhere

320

Delvers dont understand the real meaning of that. Its not “strew you guys”. The narrative is the following:

People that play Delves exclusively do so specifically to not play in a team. So they dont fully understand what the group dynamics are. Gearing in M+ is not about “min-max midset”.

Its about forming a squad of soldiers. But if one of them dosent have an arm and a leg… well… there are consequences. And said team-member will feel left out and/or bad.

So people will try to compensate that one way or another. And if that means spamming Delves they will. But it has nothing to do with min-maxing or competition.

Delvers on the other-hand, don’t have the issue of being the “bad team-player”. They are alone. It dosent matter what gear they have, and do not suffer the consequences of having bad gear.

Hence why M+ community feels like its unfair. And feel left out.

My opinion is that this story between Delvers and M+ is not new. For the past 8 years we had Raid/M+/PvP. Recently, PvP got divorced and we just got Raid/M+. And believe me they tried to solve the issue multiple times with different solutions.

But its a problem with out a solution. The only way out is the PvP divorce option. Which is what I advocate because its the only one that actually worked for a change.

Everything else is just a deja-vu.

Those flasks are OP. You take 1 or 2 and you are immortal. Its the only way to make Brann a fraction as good and useful as your wife is.