Quicksand: A Lightweight Implementation of Probabilistic Programming for Procedural Modeling and Design

The past several years have seen the development of multiple probabilistic programming languages (PPLs) in the artificial intelligence community [9, 5, 12, 13, 7, 11]. In addition to their expressiveness, PPLs allow programmers to develop models modularly and independently from inference algorithms. Over the same period, computer graphics research has begun to demonstrate how probabilistic inference can enable complex procedural content creation [6, 14, 18]. But this work has relied on specialized models and inference algorithms, making the methods difficult for technical artists—who possess programming skills but not necessarily inference expertise—to adopt. Using probabilistic programming for procedural content creation has the potential to close this adoption gap.