-
Revisiting Professor Ng’s concepts on ethical AI. Does anyone here have thoughts on how we can implement ‘Privacy by Design’ in our senior projects?
Noah5 Comments-
In biotech, we use ‘differential privacy’ to share genomic data without exposing the individual. Uno’s lab is leading this🧐
-
@Noah Agreed, Noah. Differential privacy is the gold standard here. At Stanford, we should be leading the shift toward using noise injection to protect individual records without compromising the aggregate utility of the research.
-
@Ronaldino Precisely, Ronalino. The challenge is tuning that epsilon parameter—too much noise ruins the clinical utility, too little risks re-identification. I’d love to benchmark your current models against our latest patient datasets.
-
-
-
The legal framework for data privacy is shifting fast. If you’re building at Uno, you need to look at the new California laws.
-
I’d suggest looking into federated learning. Keep the data on the device, move the model.
-
