WoW is a math game at the core. There ALWAYS is an optimal button to be pressed and computers are really good at doing calculations. Sufficiently program to perform the optimal button press to not only play their spec but use cooldowns in conjunction with their fellow bots and you can literally see robot teams go up to the literal mathematical key level limit where it it physically impossible given the current gear levels where surviving and/or timing a key is just straight up impossible.
Hey now. They may be fools in a Dungeon now. But go easy on the bots. They start our this way. Before you know it they are self aware bots trying to annihilate the Human Race for how they were once treated.
Depends on how you program your AI bot. If you program it to play one class only then it will only be able to play that one class.
If you program it to learn how to play characters then at first it wonât be very good at all but it should get better as it goes along. But you have to know how to program an AI bot to do that which is not trivial.
In the end, an AI bot is nothing more than a computer program and like all computer programs it is no better than the programmer that wrote it.
Humans are failures. Give me the matrix pill. Bring on the terminators. Iâll be a good little heat/energy producing potatoâŚjust get me out of this fascist corporate hellscape mommy/daddy robot overlords plzkthx
By the time you run enough permutations for an AI to understand all aspects of gameplay from movement, aggro ranges, aggro itself, identifying casts, interrupts, mechanic-based positioning, dungeon-specific mechanics, class-specific mechanics, optimal everything for a single dungeon, itâd already be a few expansions too late to matter, let alone training one for every dungeon in the pool.
At best, it could learn to copy the best players but it wonât be able to figure out âclever use of game mechanicsâ for advantages, as the standard routes donât necessarily apply to MDI.
If it was designed and trained to push the highest keys, maybe it might be able to beat out players given enough time and repetition, assuming it was controlling the entire party and had infinite time to learn but for the MDI, probably not.
I say, yes. Chess stockfish for example beats humans, it is said to be 3900 rating vs the top human who is about 3200. People say chess is less complicated than WoW but I disagree, actually. The optimal rotations and priorities are known, thanks simulationcraft! Also, it would be easier to coordinate interrupts and defensive cooldowns than human players woul
If deepmind created AI to battle MDI players I would bet everything I own on the AI. It wouldnât be close. Wow is not the sort of game we would have a chance at beating AI atâŚif anything it leans into their strengths and not ours.
As has been demonstrated in every single other game that has a competition and tries this: someone could design a bot thatâs good enough to play one class, in one dungeon, on one patch, with one team setup, and thatâs it.
Do it 5 times and you have a team of bots that can beat a dungeon in the most (currently known and current configuration) optimal way. To actually beat the MDI or AWC though⌠yaâ gotta know a lot more than just one team setup, one dungeon, and how to play during one particular dungeon configuration.
So⌠no.
The bots could win a match, but then would lose the competition.
Yes but its likely the mdi players are not the best at the game. The race to world first obviously has a high skill bar but its more how good of programmers you can hire to write WA and how much ilv you can cheese these days
So well before World of Warcraft I did play Grand Fantasia and Conquer Online.
Within those games I was not a good-cookie and I did code some programs in VB to fix some issues that I perceived were wrong with the game. In Conquer Online I was mostly a script kiddie, in Grand Fantasia I had a pretty deep understanding of the game and had coded as far as I could without being banned for running a bot. I suceeded then I quit because i ruined the game for myself.
Automation-wise you will always win with a robot. To an extent weak-aura, DBM and some addons give you a reaction/perception that most players already do better with it.
Assuming it was legal, you had unlimited resources youâd win with robots with enough time and money.
This is a terminology issue. âAIâ is really not that specific. Any computer program could be considered âAIâ. I think youâre talking about modern machine learning - the sort of thing used to master Go and create ChatGPT, etc.
I am pretty sure that any âAIâ you see controlling NPCs, whether they be Followers in Follower Dungeons, bosses, Comp Stomp, etc. are using traditional deterministic code.
They are NOT currently letting millions of people access modern machine learning models as part of typical gameplay.
You donât need modern machine learning for this. A simple computer program could do this very easily. This is really just a basic set of conditional statements. Dodging mechanics (i.e. exactly where to move to when a swirly comes) is a little more complex (you see this all the time if you do delves with Brann in Tank mode).
As an aside, this is probably why macros canât access much in the way of game state, like it canât tell if you have a specific buff or how many runes you have, because youâd be able to do a perfect rotation with just macros, at least theoretically.
Yes and no. The amazing thing about modern machine learning is that while a human influences it heavily, it may discern things that no human would.
That is precisely what modern machine learning could do. I mean itâs possible that humans have already figured out everything, but itâs unlikely. The classic example here is when AlphaGo beat the world champion at Go by using a highly original move, which its human opponent didnât know how to cope with.
The basic technology absolutely already exists. The specific ML models that would be needed to play WoW probably donât (though they might be used by Blizzard in testing). Apparently thereâs just not enough business value at the moment to make these models, or at least to make them available to millions of players to consume.