now that we can use AI for alot of thing's in our daily life was wondering if AI could be used in developing any game and than the AI would scan the game file to make sure that the enemies acted accordingly and and there were no updates needed to fix anything? i feel it would make game developing alot more easier. .imagine creating a game with just your voice (AI set game time in 24:00 hour increment's) (AI check enemy's pathway to make sure it is not blocked) etc.etc. and it would only take 1 person to create a good game and not a team.
#AI?
1 messages ยท Page 1 of 1 (latest)
bet AI couldn't make flying sharks
hmmm bet
But what is the question for the devs? This forum is used to ask question about Ylands/The work the developers if other players cant answer them,
You could change the whole text and change it into a question saying: Are you guys planning on making an AI that could help develop Ylands?
I guess the question is, if AI could be used in developing any game and than the AI would scan the game file to make sure that the enemies acted accordingly and and there were no updates needed to fix anything
Just search for the "?" in the text XD
yeah its a bit of a long side story
maybe im a bit of a to difficult person for this lol
I do believe there are games that has been ai generated like those fake ai pictures, but to write a program that will play/learn and correct ylands with out ruing/or give devs more work to correct things cost a ton of work
ai can be used for game development yes, but only as support, you still have to put in work yourself, found that out the hard way
not if it is programmed correctly
Okay, I am not an Ylands dev (sadly), but a game developer nonetheless, so I allow myself to give an answer too.
Well, you're right, AI (or more precisely machine/deep learning) and what it can do has seen a big jump forward and a lot of public attention recently. You have probably used chatGPT and its siblings yourself, and seen that you can practically have a human-like conversation with it on virtually any topic. And it also seems extremely knowledgeable, however, if you go into detail with virtually any topic, you will also quickly realize that chatGPT does by far not know everything, it's just very... confident in its answers. Now surely, with time, chatGPT will probably be further improved, it will probably be able to give functional links and references for its answers and better state how confident it is in its answers.
But let's look at game development. I know of many fellow game devs that they use chatGPT during development. I use it myself too, because it is basically like an improved search engine like google, where you can ask things like a real person, for example when you forgot the name of something and can only describe what it does, rather than google for the term (see picture).
And yes, chatGPT can also give you functioning code snippets, and some people even attempted to let it do ALL the coding work in a game. There are some youtube videos out there you can have a look at for that.
But the problem (right now at least) is that those code snippets are not at all reliable. Sometimes chatGPT mixes up pretty simple algorithms and gives you a completely unsuitable answer for easy problems. And those are well-defined, isolated problems, with the current state of AI, there is currently no hope for AI to create a complete game with 2D/3D assets, online multiplayer, cross-platform support, a built-in game engine/editor, compliance with rules and regulations across countries, interfaces for 3rd-party services, etc.
Will we be there one day? Maybe... but it might still take a few decades.
Another big issue about AI is the training though. In order to create an AI (whether it is tailored to one specific purpose or for general conversations like chatGPT) it takes a lot of training data and computation power. And I mean A LOT. So big in fact, that it has become an actual question in current discussions whether machine/deep learning is even a sustainable technology for the future, given how much resources it requires to learn a task. There are noteworthy solution attempts for this, though (for example "transfer learning").
But keep in mind that AI can also only learn from what already exists. So if we completely switch the game industry, it will inadvertently generalize the style, gameplay, ideas, etc from existing games. For anything new and extraordinary, there has to be a human (or a type of AI that does not exist yet), which can come up with new ideas.
Another point is that we only switch focus in development that way. Instead of programming a game, a game developer would now need to program an AI (which is basically just a program in itself too). Sure, it might require less resources, because with one AI, one could program multiple games, but still, the AI has to be constantly updated to comply with the latest technology standard (so retrained time and time again) and also every game will one way or another need somebody to keep it updated and maintained.
Coming back to your example of the path finding AI, which can make sure a path is free: What is commonly referred to as AI (but is in fact machine learning) is not always the best choice for certain problems. In pathfinding in particular, there exists an algorithm, which is for common situations proven to be the fastest and optimal solution: it's called A*.
There are a lot more things to consider, for example how the scene is understood by the algorithm, but it is a highly complex problem, which AI WILL struggle with.
If the question was targeted on Ylands, then answer is no. Legally you cannot use it on anything without proper rights which cost millions of dollars.
Also any commercial AI which is obtainable nowadays can be used only on trivial tasks mentioned above by Steigi.
Also you do not need AI to "scan game file" (whatever that means). All these things you mentioned are done by automatic testing (which is widely used more than 20 years). It requires 10^12 less CPU strength then AI = it is cheaper and more effective and it can be run on daily basis.
As for future it is same with everything in our life. If the price -> performance will be better than technology used now, it will be used. Now it is not.
who own AI? who also own Unity Engine? who own Android system's? these are all open source applications. what you are saying is that you can own the word Strawberry? while it owuld take some money yes! AI could be devoloped to work with any game to check for mistakes or flaws that would render the game inoperable. like Bethesda putting out an update to Skyrim which broke 2 texture files and made 2 quests incompletable with AI being introduced i feel it would have alerted the dev's at the flaw of their update breaking some quests or whatever. i was not just asking it for Ylands but with every other game
So first of all: Unity is NOT open-source. It is a commercial software, where normal people have no access to the source code.
And the concept of AI is technically not owned by anybody, but for example ChatGPT is not open source either (despite the name "Open AI" ๐ ). So if you wanted to develop an AI for a specific purpose, you would have to develop it pretty much ground-up, recruiting a large team, maybe contracting experts, getting sponsors, hardware for the training, collect raw data from already perfect games, etc.
AI always sounds like the magical answer to all our problems, but it is
A) not unfailable,
B) costly to develop
C) sometimes simply not suited for specific tasks
sorry made a mistake lol Is Unity an open source game engine?
Unity is not open source. Unreal is open source. The basic version is free, but commercial licenses have a price.
I mean you have access to the source code of Unreal, but there is still a license agreement. You cannot just use it for whatever you want without paying.
Godot is technically 100% open-source though.
But still, the examples you gave above about checking if an update breaks stuff, is something that can be done by either already existing, non-AI software or simply humans. Software testing and quality are big fields, and AI has already found its way into them by some degree. And surely, there might be further developments in the future, but we are probably still quite far away from AI doing all the work. Right now it is more looking to me like it's going to have a support role in software development.
"mentioned above by stiegi" what about me? ๐ข
Maybe elaborate on what you said ๐
What did you do exactly?
๐ตโ๐ซ
๐ค
@cold coyote yes, a trained AI can search for bugs/exploits and certainly will find them. But it can't fix it, it cant change the core code.
For now, only humans can go through the entire code and fix it.
Maybe in future, an AI can keep track of bug reports, explpoits, whatever, then AI must do what humans do: open the project and look on hundreads of files and many thousands of line of codes to fix something, not mention implementing a new mechanic. People think AI usage is for the "common person"
People are amazed watching videos about "i made a game with chatGPT".... but if you do that, it will end up a not optimized game.
A game is not a "single plain sheet of paper".