Artificial intelligence lacks the creativity needed to develop something new in video game design, noted ID Software co-founder and Doom co-creator John Romero in an interview with Ars Technica.
He expressed skepticism about the impact of generative AI on the gaming industry.
“For us, it’s important to be original. Generative neural networks rely on existing databases, while we try to create something entirely new,” Romero noted.
Game studios employ neural networks for various tasks such as character enhancement or chat moderation. The programmer believes that the still-developing AI is not yet capable of designing and assisting in development at a human level.
“Creating entirely new things is not what generative AI does yet. I know they’re working on it. But when creating an original game, you use your design sense, work with people, and come up with things that didn’t exist before. Generative AI limits you if you try to use it that way,” said Romero.
Among the advantages of generative AI, the developer cited the field of information retrieval, noting that a neural network is a “super-Google” for getting answers to questions. He pointed out the system of model training as a drawback.
“When AI is used to create images or creative concepts, it’s problematic because it doesn’t rely on your own data,” Romero added, emphasizing the importance of ethical use of neural networks and compensating sources used for training.
Earlier, DeepMind released the Genie model, which creates games based on text prompts.
Back in March, the startup CognitionAI introduced an AI capable of generating complex code for websites or games from prompts.
In the same month, Google’s AI learned to play video games.