Assigning semantic labels to points within a point cloud aids in both visual interpretation of the data and as a preprocessing step to other forms of analysis like building footprint extraction, hydrological modeling, and biomass estimation. Our talk will focus primarily on earth observation data and airborne lidar data sources in particular, where labels are commonly aligned with those classes specified in the ASPRS LAS specification (e.g., ground, vegetation, and building), but we are also beginning to explore the extension of these same methods to data generated by commodity, consumer-grade devices like iPhones. For many years, hand-tuned models have been developed for this segmentation task, building on reasonable assumptions about the data. For example, ground points should include those lowest elevation returns within a local window or building segments should typically be planar. Within the past decade, we have seen a surge in AI/ML powered models that are able in many cases of outperforming the prior methods, being able to learn novel features and adapt to the intrinsic variability of data. We will provide an overview of the open source ecosystem powering this trend, from benchmark datasets like US3D and DALES to machine learning frameworks (i.e., PyTorch and Tensorflow) and key libraries such as PDAL, Open3D, and PyG.