"But more importantly, Zelnick said that AI was "backward looking" and hence would not succeed creatively in making something as good as a Grand Theft Auto game. "Let's say there were no constraints [on AI]. Could we push a button tomorrow and create an equivalent to the 'Grand Theft Auto' marketing plan?" Zelnick said. "The answer is no. A, you can't do that yet, and B, I am of the view that you wouldn't end up with anything very good. You end up with something pretty derivative.""
"AI models, the executive said, worked on data sets of older information, and while that approach could help in other fields, it would struggle to create complex and immersive video game worlds. "Anything that involves backward-looking data, compute, and LLMs, AI is really good for it, and that applies to lots of stuff that we do at Take-Two," he said. "Anything that isn't attached to that, it's going to be really, really bad at.""
Take-Two maintains that artificial intelligence has a limited role in creative game development because models primarily rely on historical data and lack genuine creativity. AI cannot currently produce the complex, immersive open-world experiences exemplified by Grand Theft Auto, and any attempted AI-generated equivalents would risk being derivative and low-quality. Intellectual property constraints complicate the use of AI, since creations produced by AI may not be protectable. AI can be valuable for tasks that depend on backward-looking data, computation, and large language models, but tasks demanding originality, intricate world-building, and novel creative design remain poorly served by current AI.
Read at Gadgets 360
Unable to calculate read time
Collection
[
|
...
]