"LegoGPT combines language models with engineering constraints to craft uniquely 3D LEGO designs from natural language, promoting both manual assembly and robotic execution."
"Developed by Carnegie Mellon, LegoGPT uses a dataset of over 47,000 LEGO models, ensuring generated designs meet both user prompts and stability requirements."
Researchers from Carnegie Mellon University have unveiled LegoGPT, a novel system capable of generating stabilizable and constructible LEGO structures directly from natural language inputs. By merging large language models with defined engineering constraints and leveraging a dataset of over 47,000 LEGO models, this project allows for the creation of unique designs. The models are physically validated and enhanced through the use of physics simulations, ensuring that results are not only imaginative but also structurally sound for manual assembly or robotics.
Read at InfoQ
Unable to calculate read time
Collection
[
|
...
]