Hacker News Reader

> To be fair, it's already been torpedoed due to GDPR

Not really. I'm a deep learning researcher and my work has to be palatable to the GDPR. You can remove personally identifying information from training datasets with some effort (replacing faces with GAN-generated fake ones, ditto with license plates and telephone numbers, etc.) It's just another aspect of data cleaning and not that big a deal.

View on HN

Are there any good papers describing best practices for doing this? For example, how do you handle personally identifying information that needs to be in the training set? (e.g., names of actors, authors and other famous people) Do you have a mechanism in place to process "right to be forgotten" requests, e.g., by removing that person from the training set and then retraining the whole model, or do you put in some kind of filter on the output side?

The problem is that a lot of those databases will not exist to begin with. Also, data owners are much more protective of their data. 8 years ago I could just ask companies for anonymised data for academic purposes and they would give it or allow scraping. Now the answer is GDPR bla bla bla. On the other hand, I am not saying privacy efforts are bad.

Reply on news.ycombinator.com