Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
"Advanced AI, smart enough to make Trinity obsolete" is an outright PR lie.
BearKnight
Member CommonPosts: 461
Coming from the software development world I know for a FACT that SOE is not employing people from MIT's AI department for their game. This would be the only way to produce AI even close to challenging enough to make the "Trinity" system obsolete. However, even MIT hasn't created AI advanced enough to make this claim.
Additionally, we had "difficult" AI back in 2004 during EQ2's closed beta. We started testing AI that would dynamically choose, on the fly, whom to attack based on class & actions they were taking. Mobs would randomly scream out "Kill the healer! It is keeping the weaklings alive!" (an undead skeleton within the instanced orc dungeon in Commonlands outside Freeport). I don't know what happened to it as mobs stopped doing that kind of stuff shortly before launch. Nothing was said about it, and launch went on without a further word on it.
The thing is, it made the game actually difficult and fun to play even with trinity. Tanks had to constantly watch their aggro or risk a healer or nuker being fried. It most certianly did NOT make Trinity obsolete.
ps: wtf with this whole "Trinity" name? CC (ie: control from enchanters, etc) has always been a critical role in Everquest. So why only pick Tank/Healer/DPS and call it Trinity???
Anyways, we've had basic-advanced AI since 2004, but no one has been using it. In the instances it was attempted it never made the game impossible, but from what I've been told by developers in the industry it made it that much harder to make new content for the game as they always had to consider how the AI would be used in that situation. This is the only reason I can guess as to why they stopped using it????
Either way, don't believe this nonsense about EQ:N. They're following the current "trend" of PR buzz phrases to catch people's eyes. They should just come out and say that they want to move away from selective roles like Tank/Healer/Dps/Control. After all, we all know dungeons in Everquest 1 cannot be done without that specific setup right guys? (In before I post about constantly doing DPS only dungeon runs etc in EQ1, or before someone else does)
Seriously, it really is getting frustrating reading about this "Advanced AI" when it doesn't exist. It's like the CEO of Spandex coming out to tell everyone we have conquered the effects of "Gravity" to a room that has at LEAST one Physicist/Engineer that knows for a fact we haven't yet.
/endRant
Comments
Blizzard did an interview, or wrote an article several years ago talking about the combat AI. I'm paraphrasing, but they basically said writing a good combat AI would be easy. The AI could have all the abilities that a player would have, it would be situationally aware and wouldn't get excited and make mistakes. It would use the ideal combination of abilities for any given situation. They need that kind of AI for testing.
They didn't use the more advanced AI because it wouldn't be fun for players. The AI was too good to be used against players. It would be good for testing though.
SOE is doing something completely different with the Story Bricks AI. That gnoll will attack because it is angry with you for killing his brothers. Or the gnoll will move to a less populated part of the world so it is easier to attack adventurers. The combat AI is going to be a version of that kind of behavior. It won't make perfect decisions, but if it's attacking someone, and another player heals that someone, it'll attack the healer instead of making the right decision at the beginning of the fight and attacking the healer right away. Something like that anyway. Or even better, mobs in groups will act like they are a group, repositioning themselves, healing each other and possibly running away.
The group AI has been done years ago in single player games. At least a single player game. Groups of mobs would have a leader and they would have the ability to do things as a group. If the leader was killed, the group would be disorganized unless another one of the mobs moved into the position of leader, and they would again act as an organized group.
"Intelligent" AI isn't smoke and mirrors, and it's not science fiction. It is well within the realm of what can be done. The only important part is that is has to be fun. Keep in mind that mobs don't exist to replace players, they are there to provide a challenge, and ultimately drop loot.
I can not remember winning or losing a single debate on the internet.
+1
ANET said something similar in an old 2005 interview
ANET intentionally dumbed down the henchman in GW1 so players would seek out other players in the game, instead of relying on NPCs
EQ2 fan sites
I actually believe the original 'Trinity' was Tank/Healer/CC because the rest of the group could be anybody. Any can DPS, but groups really needed those three to shine, and DPS was the least important role. At least that was my understanding of it. But MMO's kind of got dumbed down and CC was to complicated/hard so it has kind of been slowly removed from MMO's in general.
Now Healer is to hard/complicated so that has been on the list for removal, and next we are going with Tank.
Personally I enjoy a variety of different important roles, I used to love CC (and I miss doing it) and I love complexity of gameplay. And sorry for being mostly off topic.
I remember some of the AI in GW1 being rather frustrating. The damn mobs just wouldn't sit in the AE and die like good little lemmings, they had the audacity to move out of AE spells!
I have to agree with the points made on existing AI. I think the reason companies don't use full-fledged AI for monster opponents is that it would be too much for the players to contend with.
Try playing Dragon Age, the Bioware RPG, on hard mode WITHOUT using the space-bar to pause the action. In many encounters the human is simply overwhelmed - people cannot hit the keys on a keyboard fast enough to match the AI.
Imagine a smart monster with appropriate AI in EQN:
- Let's say an ancient Liche who has survived ages. A party of adventurers shows up to claim his treasure. Usually this means fighting through predictable encounters of steadily increasing difficulty with the Liche's followers until you reach the inner sanctum, where said Liche obligingly makes his final stand.
Now imagine...
- the adventurers show up at the Liche lair to find - no one. Unbeknownst to them, the Liche knew they were coming and decided to prepare a surprise. As the adventurers head out, disappointed, he springs his trap as he and his followers emerge en masse from a hidden passage and steamroll the adventurers.
This would be way, way out of most MMO players comfort zones. People want to relax and bash keys while the bodies of the enemy (and the treasure loot) piles up. Not have to out-think Skynet...
IBM created Deep Blue nearly 20 years ago, which was advanced enough to beat the worlds best chess players, or at least play them to a draw. OP says MIT is the only institution with advanced computer AI in 2013.
Cool story, bro.
Advanced AI or even the ILLUSION of advanced AI seems easy to make, but it must be done with extreme care not to alienate and frustrate average player.
But what I would want to see from a "next-gen" MMORPG is not only a push away from LEVELS, but a radical push from HP system. It can be done on a large budget like this game seems to have, sadly it's not to be in there.
Imagine knowing your opponent's condition only by looking at the wear and tear of his clothing and armor and by his movement and cuts and bruises on his body. Sounds terribly difficult to implement and even hard on the player, but surely one of the perks of the future MMORPGs. Or only the ones that aim for the realism, of course.
NEW IDEAS that can refresh the STALE state of MMORPGs
Not sure if it was the same article I read but I did read an article that stated pretty much the same thing. that creating an AI that could beat a player was easy. Whether it was fun or not for the player was another thing.
I'm not sure where the OP is getting this "Needs MIT CSAIL level AI" in order to beat a player.
Godfred's Tomb Trailer: https://youtu.be/-nsXGddj_4w
Original Skyrim: https://www.nexusmods.com/skyrim/mods/109547
Serph toze kindly has started a walk-through. https://youtu.be/UIelCK-lldo
Regardless of what many posters might suggest, complexity in combat is not a core desire of most consumers. AI at a certain level introduces chaos into an encounter. To paraphrase: "no battle plan survives contact with the enemy". While this sounds enticing to some I can't imagine an MMO AI system where there won't be obvious 'tells' and logical sequence so that players can map out their time. They hate chaos.
The problem is the audience not the availability of technical expertise in combat AI. If you are willing to cater to a niche then many things are possible within the constraints of lesser investment/sales.
Some of Youtube's more watched clips revolve around every step in a dungeon run for every mmo. If you gave people the choice of the adventure or a "Press to Finish" button at the beginning of a dungeon run they will jump over you to press the button.
The same can be said for the options in roleplay. Most MMOs provide more than enough of a rough framework for roleplay but unless you handhold and specifically relegate assets people just can't grasp it.
rant over :P
Being a Program Mananger within the software development world I can tell you that you are full of shite.
Now, just because it CAN be done doesn't mean SOE is going to get it right.
Oh yeah, just because you may not be able to do it doesn't mean that it takes students at MIT to get it done.
Your argument is invalid since AI can be situational and divided into seperate action groups. Combat AI can be completely different from behavioural and/or enviromental. Your speculating that everything SOE mentioned about AI implies purely to combat.
As much as I'd love for SOE to deliver everything their selling I'm pretty sure I have a better chance of pulling a rainbow flavoured gnome out of my ass.
I am probably the minority on this; but I don't want "smart" AI, I want believable AI. What's the difference? "Smart" AI is player like. The Chess AI is a good example. It plays like a player might and is very good at what it does. We know, when playing against this Chess AI that we are playing vs a CPU and code.
When AI, such as mentioned with Anet's code and Guild Wars 1; we knew it was simply AI. After one of their expansion, they tried to push "Smart" AI back in a bit, and the lash back was pretty bad. The "One Size Fits All" AI issue came up and all NPCs and Mobs behaved the same way (expect the Henchmen which were left to be "stupid").
People that loved this celebrated the AI, saying it was more "player" like. Mobs shouldn't stand in AoEs! was the cheer... only, players being adaptable learned that AI that avoided AoEs and pretended to Kite players were soon out smarted by clever use of AoEs and near no damage weapons. Example: Need to protect the healer? Cast an AoE on him/her or give the Healer Henchman a single Blind/pbAoE and the AI ran screaming away. They pretty much created a system where AoEs were free Force Fields. Ranged mob giving you a hard time? Wand it for 1 damage and it would run away "kiting"...
The worst side effect of AI like this to me, is the broken Immersion factor. Animals, Weird Creatures, Humanoid, etc ALL behaved the same way in every encounter. It became extremely boring and eventually I left... which, at that point I was one of the last in my Guild's Alliance. I heard they addressed that and fixed it sometime later, but I was long gone.
Believable AI to me, is a much preferred setup. An animal, for example should not know what a heal spell is, nor should it know and understand what a mage is. It might be able to determine a threat from someone holding fire or a long spear vs a dagger and a water skin. But beyond simply understanding, it shouldn't. The same animal should not be able to dodge around to avoid a warrior nor should it roll and skit away from a fire ball. It should, however flee and not want to return if said Fireball exploded near it.
Bandits, for another example should not attack a player head on nor attack on an even/fair fight. Bandits should instead wait, hidden, for that perfect victim to wander along. Back stab and go for the easy kill. They should also have believable tactics for being Bandits, such as dodging, rolling, flanking and know enough when the fight is not in their favor and flee.
I'm pretty sure they can make "Cheating" AI that will give us a run for our money. Magically know who has what skill/spells equipped and go for that person as a "if player has X then attack" (simple if and then code); but can they make AI believable? That's what I want to know...
I think you are right, but lets assume they have the best and brightest in AI right now. Say everyone quit what they were working on at (ai)2 and build EQNext.... for shits and giggles.
What would that AI do?
If possible, it would overwhelm the players with strength of numbers..
If possible, it would set a trap, sneak up with greater numbers and slaughter the players...
If caught in evenish combat, it would assess the biggest threat to its life first, incapacitate it, and then slaughter the players...
etc...
If they make it that smart, we all might have some fun.
But...like I said, I think you are right... It's all just BS.
As a software dev, I also know that advanced AI has existed for years, without machines to run it. All in how you define what advanced is.
Weather it be continuous or discrete, AI requires heavy processing and decision making. Dynamic or static ... learning vs. pre-computed. The best AI is oriented toward its purpose. I do not see any reason why SOE can not build a system that is specifically oriented around their purpose, just as Left 4 Dead did in their 2nd game. They wrote a fantastic crowd solver for their zombies based on a dynamically branching AI.
This is no approach a software dev should take, as our job is oriented around problem solving. Some amazing programs and code have been written by individuals who aren't in the crowd you suggest. Even then, these prestigious colleges many times are forced to exclude smart people due to lack of funds, location, or culture. If you really are a software dev, I don't see why you are so critical of SoE.
EQ1 had more advanced AI than any other mmo by a mile.
Just a few examples of how things used to work back when games didn't cater to children and the mentally handicapped:
If you aggroed a mob, every mob within like a certain radius based on level would assist. There was no mobs chained to other mobs, so you could use tactics to actually split them. It was very common for someone to accidently pull dozens of mobs at a time because they would assist each other.
In EQ1, mobs would actually cast target based AE spells on you if you were standing near other party members (they'd never cast those spells otherwise). They would run up to you and point blank ae as well.
If you were high on aggro list and you sat down, mobs would run up to you and get a free max damage round off on you, often 1 shotting you.
If you had high aggro as say, a wizard, and went within melee range, a mob would often switch to attack you simply because of proximity.
If you were low health, regardless of anyone elses threat, the mob would go for you and attempt to finish you off and unless the mob was rooted or mezed, no amount of taunting would stop it. Going below 20% on a boss meant you were the next to die if you weren't healed immediately.
Mobs would flee at low life if they were alone. If they reached friendly mobs aggressive to you, they'd turn around. If the mob was a caster and you pulled him away from his spawn area, he would teleport (gate) back at low health.
Mobs would roam around buffing each other. They would heal each other as well.
In EQ1, there was no ability that forced aggro on a tank without substantial threat. The taunt ability only worked permitted you already established considerable threat, and if someone else only slightly out-aggro'd you. If, say, a caster nuked or debuffed a mob early, taunt was worthless.
The best threat tanks were hybrid tanks, but they were the weaker of the tanks. The warrior which was the superior tank class, had to rely on weapons that procced spells or magic items with expendable charges. There was no easy way of generating threat as a warrior in EQ.
Those are just a few things that come to mind regarding aggro in classic Everquest. Its no wonder people hate so hard on threat management systems when all modern MMOs provide their tank classes with surefire ways of keeping aggro with very little risk involved.
How would it be difficult to make ai more advanced comparatively to npc's in a game.
Its just more sub routines, and movement of the npc. Its not ai that will evolve into the matrix.
Its just more commands like a player has movement commands, and attack/utility commands, and use of terrain and line of sight.
The only tricky part is if they teach the a.i. to time moving behind objects and avoid line of sight.
Anyways, your skepticism is only trying to argue something as possible if its easy enough when MIT hasn't done it yet. Its not easy because it needs a lot of testing. And its not what you think.
Its easier to just add it, so it works: but the fine tuning so it feels like a pvp encounter or a raid encounter with the trinity is what makes it more complicated.
Either you are not a programmer or you are reading too much into what a.i. is. Its just a better npc. But if its smart enough to act like a player in a game its basically a.i. isn't it?
I think you are possibly trolling with the wording of a.i.. And you are trying to make a point about them calling it a.i. instead of something you think fits better.
Thats a very round about way to get your point across if thats the case.
Write bad things that are done to you in sand, but write the good things that happen to you on a piece of marble
OP your summation wouldn't have anything to do with trinity being your preferred style would it? You offered no technical evidence as to why this isn't possible, only injections that support your and only your view.
I'll give you the benefit of doubt and I'll assume you're writing that up now.
For every minute you are angry , you lose 60 seconds of happiness."-Emerson
This is different from which MMO? Which single player game? Here is a clue it is not AI it is scripted and based on hard number thresholds. The NPC has no choice, X has more threat than Y, nevermind he is kiting me.
Let's all come back to this red herring topic once we see more AI in action.
OP, don't bother. I made a similar thread, people here don't understand a single thing required when creating what they Storybricks are creating. I'm extremely skeptical, I will not say it can not be done anymore but I'm not going to be dragged into believing the PR at face value like a lot of people here are.
I've watched Storybricks's AI video, so far it's exteremly basic and consists of using "tags" to determine situational outcomes in the mobs behavior. In theory the "tags" could create the AI they are talking about but that is in theory, until I see footage of this in action I say it can't be done.
"if u forcefully insert foriegn objects into my? body, i will die"
As a not software developer but someone who has the ability to read a magazine or an internet article, I know that computer games don't use Artificial Intelligence. What is used in games is code that basically boils down to series of if-then statements, whereas true AI would be able to sit next to you and decide that it wants to play an MMO, then learn how to play it itself. True AI = Skynet, that develops it's own motivations and then acts on those motives. The orc in whatever game you're playing ain't Skynet.
But now that I've shaken my e-peen at you, to get back to the original topic, he's right. Making the mobs act more like players in PVP scenarios just boils down to removing the taunt mechanic. If you have a game where PVE uses taunt and PVP doesn't then no one is going to bother with killing tanks in PVP. If the game puts the taunt mechanic into PVP, and you basically have to kill at least some of the tank classes first, then you'll kill them, and all of the sudden the PVE AI is just as smart as the players.
It's all very simple and has nothing to do with "difficulty of AI". As a previous poster pointed out, EQ1 had really good PVE mob action simulation, and you all cried like little girls about trains and corpse runs and flocked to WoW when it came out because you could solo 3 monsters at once.
Who said they believe it? Where did they say it? Stop arguing with people who do not exist.
Just a bit of advice: Not believing you or the OP has nothing to do with believing SOE.
For every minute you are angry , you lose 60 seconds of happiness."-Emerson
Yes, you are right, they are all the same. Oh wait, no there not. Having played every major and most minor MMOs in the last 15 years, nothing was even vaguely similar to AI in EQ.
What are you talking about. Look at this thread. Look at the amount of people writing. "Complicated AI has been around for ages, therefore it shouldn't be a problem for SOE"
"if u forcefully insert foriegn objects into my? body, i will die"
Basically, stuff like this.
Is player heavily armored? /cast armor pierce ability
Is player using melee weapon against me? /cast disarm
Is player casting healing spell? /cast interrupt
Is player casting fire spells? /cast fire absorb shield
Are players within 3 meters of each other? /cast aoe
Is player under 20% health? /cast execute type ability
Really simple stuff like that could make your average encounter much more difficult, for the average player. Some games have already had mobs that do things like this for years. No super-advanced, state of the art developed by MIT for military applications type of AI needed.