Autonomous vehicles are continuously collecting data, and the system must decide which observations are sufficiently interesting enough to save for future use. In settings like this, one cannot use a static model to determine what constitutes an anomaly. For instance, when the car first encounters snow, this is likely an interesting event worth collecting data for. However, after the car has been driving in snow for an hour or more, the setting is no longer interesting enough to warrant more data to be collected. Hence, the notion of what is anomalous must be adaptive, and models that detect anomalies or distribution shift must be learning continually.