The results of this paper could be of interest to NeurIPS. However, the author(s) should try and address most of the concerns raised by the reviewers, specially a review of existing work on Kalman filtering with infinite variance.