So yeah, I want to discuss or point out why I think Valve needs to fix Anti-Cheat issues. They have VAC but apparently its doing jackshit, be it Counter Strike 2 (any previous iterations) or something like Hunt: Showdown the prevalence of cheating players is non deniable. For me personally it has come to a point that I am not enjoying playing those games anymore, although they are great games by itself. But the amount of occurrences being killed or playing against cheaters is at a height, where I don’t see the point anymore.
- Why I think Valve is the only company able to something against cheaters?
Because they have the tools with VAC already aiming to prevent cheaters. Valve has got the resources to actually invest into something more profound which could be used for any game where anti-cheat protection needs to be implemented. And lastly Valve is the company which is interested in furthering the ability to gaming on Linux, the anti-cheat solution needs to work on both operating systems. Only Valve has the motivation and means to achieve that with their knowledge and resources. What do you guys think about the topic? Is the fight against cheaters hopeless? Do you think some other entity should provide anti-cheat protection, why? I skimmed over “anti cheat in linux kernel” posts in the net, but I have very little knowledge about the topic, what is your stance on it?
Edited: Mixed EAC with VAC. EAC seems to be part of Epic Company. Both of these tools seem unable to prevent cheating like mentioned above.
“At some point”?
Fuck, I can do that now. Training ai is trivial. It’s getting it to do well that’s hard.
But if you can mix in other things like regular object recognition it gets much easier
You could train a network with three nodes over 15 minutes with a dataset consisting of two screenshots of victory royales, and say you “did it today”.
But really you’d need a model that takes in the same inputs as a human (audio and display signal) then outputs the same inputs that a human would make (keyboard/mouse). This is orders of magnitude more complex than what even LLMs do.
There’s no easily accessible dataset large and detailed enough for training a model on any current competitive games, not to mention modern games are some of the most complex things humans interact with at all. Driving a car is a joke in comparison (in terms of complexity).
Looking it up, the most advanced game-learning AI is currently MuZero and it’s been learning old Atari games.
I disagree, replay files on something like starcraft work great for training. In fact, Starcraft2 was targeted after Go for AI gaming benchmarks, and it’s replay functionality is probably a reason why.
But yes, to set it up as a separate machine that is sending nothing but the same user inputs as a person to a computer running the game both completely bypasses and invalidates any of the anti-cheat malware that gets installed, and becomes more and more undetectable to server side code.
insert .gif
yes, that people do for fun, as hobbies. With great success. And they are at the point where they have their own leaderboards and benchmark games. The ethics of it are the biggest reason why it hasn’t been applied to competitive games.
I did say current competitive games. I have no doubt it will happen, but I don’t think your claim that it’s doable right now if only all the right pieces were put together, is sound.
It’ll happen, but not tomorrow.