Many scientists are actively experimenting AI/ML methods to either replace the conventional methods or improving the existing data products to higher accuracy and resolution. However, most people complains that the experiments reported in research literature are very difficult to neither reproduce nor reuse. The source code and notebooks and associated data, models, and results are hard to find, access, interoperate, and reuse. Meanwhile, the trained models are often biased towards the majority and common patterns due to sampling strategy or natural distribution. These issues are significantly harming the usability and trustworthy of AI/ML in geoscientific application. This session aims to solicit community experiences, opinion, and vision to enhance the FAIRness and fairness of AI/ML.
View Recording
View Notes
