OpenAI and other companies have built AI that can navigate through a browser literally as though it were a human navigating through a website.
So, what’s to stop people from using AI to play World of Warcraft (or other video games) in the not-so-distant future?
Only answer I can think of is if the financial incentive isn’t there (cost to run the AI Agent/Operator to play WoW is more expensive than the monthly WoW subscriptions for the amount of gold the AIs can make in a month).
Thoughts?
It would likely be less effective then dedicated bot programs.
People really don’t get what ai is good for. Sure it might be harder to detect but I bet a optimized bot would still earn more gold.
Who says an AI Agent/Operator can’t be optimized?
This isn’t just an inquiry about what’s currently possible today, it’s an inquiry about what will be possible in the future.
I’d say “The WoW token needs to go”, but worse stuff would just take its place if it were gone. It’s a really tricky situation.
AI will kill the world. Give it a few years and the top players on WOW will all be AI bots. Sad, isn’t it?
It won’t be better then a dedicated bot without removing what makes it undetectable.
It would need to pull off the impossible neck break turns current bots do in order to be competitive.
Who’s to say the cost to run an AI won’t come down? Who’s to say that the money that one AI can generate off of gold wouldn’t turn a profit? Who’s to say how many WOW accounts one AI could handle at one time?
We need to find this Who person and see what they say.
AI is already becoming the quickest and most reliable “employee” in the world. The trick is just figuring out how to intelligently communicate with it to give it tasks that benefit you.
I mean, I don’t want WoW to play my video games for me, but that doesn’t mean somebody else won’t… particularly if there’s potential financial incentives involved (such as saving time chasing a lengthy feat of strength reputation achievement).
“Time is money”, as WoW Goblins would say.
An “AI” bot would only work with reinforcement learning and it would take months of training with thousand of accounts sharing the results for it to become as effective, in PvE alone, as your average IQ 7 year old playing WoW for the first time.
And with every class/spec change it would have to be retrained.
Optimizing it would be amusing as a master’s degree paper, but not much more than that.
You answered your own question. Running a neural net locally would burn more money by pegging a GPU at 100% the entire time than any malicious actor would earn using such methods to play WoW to generate ill-gotten gains. That’s the reason why a bunch of used mining GPUs flooded the market sometime in the last two years. The crypto market crashed, and it wasn’t worth the energy costs to mine for coins. The discovery that deep neural networks can play old racing games and find exploits is a curiosity. The same holds for the fact that they can use browsers similarly to a human. Those represent advances in the science behind neural nets as researchers try to arrive at artificial general intelligence. It doesn’t have practical use beyond that.
If it runs on the computer playing, Warden. If it ran on a different computer, I imagine they wouldn’t be able to detect it but couldn’t say for sure. Either way that is 100% a ToS violation.
If by “reliable” you mean exactly “average” or “mediocre”, then yes, as it’s designed for that.
People have been botting in WOW for decades. This would just be another form of botting. Blizzard already fights against this regularly. I assume they will continue to do so.
I don’t even understand what your question is.
Are you asking about what if someone like Vedal would run NeuroSama (programmer vtuber who made an AI that is very popular … and genuinely kinda terrifying in how human-like he’s actually managed to make her) to play WoW or what’s your actual question?
Is this a version of the “Dead internet theory?”
Are you talking about bots, or AIs, general AIs, LLMs, or… what exactly are you asking about?
As of yet, there’s no technology or legislation that would require AI Agents/Operators to reveal that they are in fact not a “robot”. Without something like that, I can’t imagine that Blizzard has any way to detect AI Agents/Operators who are trained to behave exactly like a regular human player.
We call those “bots” and Blizz is constantly working on ways to combat them.
No LLM can currently act human-like enough to avoid detection. What this is referring to is some kind psychology-based AI using a neural net to learn optimization, but then also require something like a LLM to facilitate communication.
Nothing in the entire world exists yet that can mimic all of those things at the same time … and trying to make an AI that could do those things at the same time likely would require more computational power than currently exists available on Earth collectively.
This is why the Dead Internet Theory is interesting, but at least currently not a reality. What you are talking about here sounds like a gaming version of said theory.
If it runs in the local system where you are playing, they can absolutely detect that.
Just add an “Are you a bot” checkmark at login
The cost to keep a person on each bot farm to constantly go around checking Yes would just make this not a feasible thing
AI Agents/Operators can legit get around this.
The question says “Are you a robot”
The thing has to check yes genius
People not knowing anything always praising AI smh