3D-GPT: converting text into 3D assets
Researchers have developed a new artificial intelligence system called 3D-GPT that can generate 3D models and scenes from simple text descriptions.
I hope that in the future, AI tools like this can assist indie game devs, so somebody with the idea for a decent plot and gameplay can generate environments, some models, and voiced characters with a generative engine. These tools could also be a cool evolution for games that use generated worlds already: think Minecraft but less blocky.
I fear that it will be used by big studios to supplant or replace human talent, leading to endless titles of same-same dreck without the spark of inspiration or uniqueness that a human developer, artist, or story-boarder can come up with.
Or to paraphrase how somebody else put it: "We thought that computers would do the boring or unpleasant work so that humans could spend more time making art and expanding culture, instead the computers are making the art and we're still doing the shit work"
This is why we need to think on how to turn the AI usage to the workers. Thinking on creating AI models that are open source and do not depend on big companies. I'm thinking on exploring grid computing and p2p technologies, so we could explore computing power in a distributed way.
There is a lot of cool federated learning frameworks, federated inference seems less fleshed out with petals.dev being the only example I know of for it.
Making 3D assets is a full time job for a lot of people.
And hopefully this will allow them to follow the 80/20 rule where the AI can do 80% of the grunt work and the human can concentrate on the 20% creative part.