Published on: April 11, 2025
The Himalayan School of Science & Technology organized a seminar titled “Ctrl+Z for AI: The Art and Science of Machine Unlearning” on April 11, 2025. Conducted by Dr. Shiv Preet, Assistant Professor, Department of Computer Science & Engineering, the seminar explored the emerging concept of machine unlearning, addressing the growing need for data rights and ethical AI practices. The session explained how traditional machine learning models retain all training data, making removal of specific data points costly and impractical. Dr. Preet introduced the SISA (Sharded, Isolated, Sliced, and Aggregated) Training framework as a promising solution to enable efficient unlearning. Students actively engaged with questions on real-world implementation, model performance, and ethical concerns surrounding data deletion. The seminar emphasized the importance of machine unlearning for compliance with privacy regulations such as GDPR, improving model transparency, and maintaining user trust. The session concluded with the thought-provoking message that in an era where AI remembers everything, teaching it to forget may be humanity’s most important breakthrough.
