This seems like this is going to be heavily counteracted by better engines, and AI generation.
I wonder how it'll play out though.
I think this has always been the case, though. Engines haven't just suddenly got better, they've been getting better and better for decades now. Some of those improvements give you features "out of the box" that you used to have to implement yourself. One of the reasons Unity became so popular with smaller developers is because it lets you focus on building your game - most of the tech is there, you've got an asset store for additional models, plugins, etc. so save you time but ultimately making a (good) game still takes time. Making a game is a very iterative process and a lot of the quality of a game these days is less to do with developing the engine and more to develop the mechanics of the game itself - the way your characters move, the responsiveness of the controls, the UI layout and so on. All of that stuff is hard to be given to you by an Engine, because it's specific to your game.
Exactly, we've been getting better engines, tools and educated game devs for the past decade too and it's what led to current situation. I don't think AI is going to help with anything, it will just result in more soulless cash grabs if it's used the same way ChatGPT has been lately.
Procedural terrain generation in Deep Rock Galactic is pretty cool. I could see also using it for textures and NPCs to make a game more varied for not much more work.
The problem with procgen for variety is that it's almost always a few procedural changes layered onto a finite, typically small, set of "types". You can see this in games like No Man's Sky, where there are technically billions of different animals that you might encounter on a planet, but a lot of them are pretty similar. Even in DRG with their terrain gen, they're building on room templates that you'll start to recognize the more you play.
It's kind of like those ad campaigns about how many millions of ways you can make a burger. Sure, a 1/4 lb cheeseburger with lettuce, tomato, onions, and ketchup on a sesame seed bun is technically different from a 1/4 lb cheeseburger with lettuce, tomato, onions, and mustard on a sesame seed bun, but they're both still burgers. You might hit onto some unique combinations (e.g. meat, cheese, and toast on the bottom, with no top bun -> patty melt) but you're ultimately still just seeing burgers everywhere, and the system that generated the burger isn't ever going to generate aloo gobi.
It's pretty good, agreed, but we've had procedural generation since before minecraft. It doesn't have anything to do with ML/AI afaik.
I think so too. The process of content creation will become more efficient. I hope it will allow companies to try new and weird things with less risk.
It'll at the very least make indie studios capable of insane things.
That also. I've been keeping an eye on this kind of technology for my one person projects.
It helped me to start making my first commercial game, i always looking for new AI tech to see how I can use it to make my games. I just need a better graphic card and i could generate 3D models with DreamFusion.
I believe that, to an extent, this has actually caused some of these problems we're seeing. When tools become easier to use, more is expected from the devs, particularly in the AAA space.
A tool is made that, in theory, helps you do 12 months worth of work in 6, so they make the game twice as big. However, in reality you still have to deal with various unforseen problems, especially those caused by overconfidence in those tools. The real-world time is actually 9 months, but they're still expected to make that huge game in 12.
Crunch ensues, which burns people out, which means less quality work and damage to health.
I think it's generally up to responsible indie devs to use such tools well and control the scope of their projects. Just because you can, doesn't mean you should.
This seems like this is going to be heavily counteracted by better engines, and AI generation.
I wonder how it'll play out though.
I think this has always been the case, though. Engines haven't just suddenly got better, they've been getting better and better for decades now. Some of those improvements give you features "out of the box" that you used to have to implement yourself. One of the reasons Unity became so popular with smaller developers is because it lets you focus on building your game - most of the tech is there, you've got an asset store for additional models, plugins, etc. so save you time but ultimately making a (good) game still takes time. Making a game is a very iterative process and a lot of the quality of a game these days is less to do with developing the engine and more to develop the mechanics of the game itself - the way your characters move, the responsiveness of the controls, the UI layout and so on. All of that stuff is hard to be given to you by an Engine, because it's specific to your game.
Exactly, we've been getting better engines, tools and educated game devs for the past decade too and it's what led to current situation. I don't think AI is going to help with anything, it will just result in more soulless cash grabs if it's used the same way ChatGPT has been lately.
Procedural terrain generation in Deep Rock Galactic is pretty cool. I could see also using it for textures and NPCs to make a game more varied for not much more work.
The problem with procgen for variety is that it's almost always a few procedural changes layered onto a finite, typically small, set of "types". You can see this in games like No Man's Sky, where there are technically billions of different animals that you might encounter on a planet, but a lot of them are pretty similar. Even in DRG with their terrain gen, they're building on room templates that you'll start to recognize the more you play.
It's kind of like those ad campaigns about how many millions of ways you can make a burger. Sure, a 1/4 lb cheeseburger with lettuce, tomato, onions, and ketchup on a sesame seed bun is technically different from a 1/4 lb cheeseburger with lettuce, tomato, onions, and mustard on a sesame seed bun, but they're both still burgers. You might hit onto some unique combinations (e.g. meat, cheese, and toast on the bottom, with no top bun -> patty melt) but you're ultimately still just seeing burgers everywhere, and the system that generated the burger isn't ever going to generate aloo gobi.
It's pretty good, agreed, but we've had procedural generation since before minecraft. It doesn't have anything to do with ML/AI afaik.
I think so too. The process of content creation will become more efficient. I hope it will allow companies to try new and weird things with less risk.
It'll at the very least make indie studios capable of insane things.
That also. I've been keeping an eye on this kind of technology for my one person projects.
It helped me to start making my first commercial game, i always looking for new AI tech to see how I can use it to make my games. I just need a better graphic card and i could generate 3D models with DreamFusion.
I believe that, to an extent, this has actually caused some of these problems we're seeing. When tools become easier to use, more is expected from the devs, particularly in the AAA space.
A tool is made that, in theory, helps you do 12 months worth of work in 6, so they make the game twice as big. However, in reality you still have to deal with various unforseen problems, especially those caused by overconfidence in those tools. The real-world time is actually 9 months, but they're still expected to make that huge game in 12.
Crunch ensues, which burns people out, which means less quality work and damage to health.
I think it's generally up to responsible indie devs to use such tools well and control the scope of their projects. Just because you can, doesn't mean you should.