turkishronin
Arcane
We already have AI voice and AI dialogue writer. How could RPGs utilize them? Can we have organic C&C?
We already have AI voice and AI dialogue writer. How could RPGs utilize them? Can we have organic C&C?
"We'll see." (TM)StableDifusion is walking into what I think will become a massive copyright litigation hell because it is using copyrighted material for training and the material it produces in many cases are derivative and not transformative as claimed, worse, you can even type the name of the artist you want the Ai to use, this will be turned illegal, 100%.
I doubt it. Plenty of material is available free of copyright, or under some anything-goes license. Or will be made available at a negligible price. Or will be made available by artists who support AI. etc. etc.Without copyrighted content to train the model on, the quality of the results will drop sharply. It will be still useful to quickly generate some generic assets but you will still need artists to make unique content that the AI wont be able to scrape from copyright free image libraries. Make no mistake, that will be already extremely useful.
This is false.ChatGPT cant generate unique text, it can generate text it can grab samples from somewhere [...]
This is correct, but so what? If you want a model that's essentially a multi-purpose chat assistant, then you train that model accordingly. If you want it to purely generate code snippets without explanation, you probably outperform ChatGPT at that job. If you generate a story-telling AI: same.Also, ChatGPT is generic aimed to do all sorts of things on a mediocre level, it cant be fully customized to your needs [...]
Japan made a law that dictates AI can freely use copyrighted material if it's for purpose of training it I think other nations will also follow suit.StableDifusion is walking into what I think will become a massive copyright litigation hell because it is using copyrighted material for training and the material it produces in many cases are derivative and not transformative as claimed, worse, you can even type the name of the artist you want the Ai to use, this will be turned illegal, 100%.
I'm no longer convinced there is a significant correlation between effort spent and quality.They (and all other games) will need even less effort to produce, and therefore become even shittier than they already are.
The times when more efficiency meant higher quality or lower prices for the customer have been over for a long time now.
I'm no longer convinced there is a significant correlation between effort spent and quality.They (and all other games) will need even less effort to produce, and therefore become even shittier than they already are.
The times when more efficiency meant higher quality or lower prices for the customer have been over for a long time now.
You are talking of buying professionally made Ai by big bucks company that may exist on an unknown future and run through the internet with your game depending on their service to continue running right? Because if it is implied that you are going to train a neuron net and produce a model that generates more than garbage on a normal PC or even a small server farm, especially now that silicon is reaching the maximum it can do and you cant expect big leaps in miniaturization anymore, that is big optimism there.This is correct, but so what? If you want a model that's essentially a multi-purpose chat assistant, then you train that model accordingly. If you want it to purely generate code snippets without explanation, you probably outperform ChatGPT at that job. If you generate a story-telling AI: same.
I'm not sure I understand you correctly. You don't need huge servers for inference. Rather, training is the costly part, but you don't need to train your own story AI, much like most people do not train their own version of Stable Diffusion.You are talking of buying professionally made Ai by big bucks company that may exist on an unknown future and run through the internet with your game depending on their service to continue running right? Because if it is implied that you are going to train a neuron net and produce a model that generates more than garbage on a normal PC or even a small server farm, especially now that silicon is reaching the maximum it can do and you cant expect big leaps in miniaturization anymore, that is big optimism there.This is correct, but so what? If you want a model that's essentially a multi-purpose chat assistant, then you train that model accordingly. If you want it to purely generate code snippets without explanation, you probably outperform ChatGPT at that job. If you generate a story-telling AI: same.
I could see an RPG randomly generate NPC portraits for each new game. Could be a cool alternate realities sort of theme or just funny.Well for one thing, [near] future RPGs will not lack for NPC portraits, and could likely even have AI generated UV textures for in-game face models. The same could go for material damage on items like swords and shields, and stones in the wall; using procedural normal mapped pitting, and cracks to indicate damage as it happens in gameplay.
AI voices will probably facilitate overdubbing changes of human voice actors, instead of additional recording sessions. It could also (obviously) be used for the entire voiced script; with any changes becoming trivial to alter.