Deep learning has made significant progress in various applications, but the widespread use of deep neural network models raises concerns about unfair biases and privacy protection. Machine unlearning is an emerging field that aims to remove the influence of specific training examples from a trained model. Retraining models can be computationally expensive, so an ideal unlearning algorithm would efficiently modify the already-trained model. To advance the field of machine unlearning, a Machine Unlearning Challenge has been organized with the goal of standardizing evaluation metrics and encouraging the development of efficient and ethical unlearning algorithms. The competition will consider scenarios such as protecting user privacy and correcting unfair biases.
https://ai.googleblog.com/2023/06/announcing-first-machine-unlearning.html