Learning to Trust AI

Our AI models are both capable and vulnerable. The advances in frameworks and computational technology offer tremendous opportunities to train complex models for various vertical applications. On the other hand, the quality of data, stability of performance metrics, and finicky training process also mean that these models may not perform as expected in the field or fall victim to various security attacks. In this talk, we discuss recent technology trends that can help bolster our confidence in AI models and showcase our current efforts to help bring trust into handling data and deploying models.