2 ms·

Au contraire, much of machine learning (learned?) software is simply the data used to train it. I think we can all agree that generic image recognition is benig

Au contraire, much of machine learning (learned?) software is simply the data used to train it. I think we can all agree that generic image recognition is benign and should be permitted. How about a database of publicity photos from famous movie actors and actresses? Would it be OK to use that database to identify which celebrity I saw getting out of a limousine? And if that's permitted, then the regulators are tasked with determining how famous someone must be or how much the purpose of the image must have been for publicity. These kinds of questions are best suited for a case-by-case interpretation in the courts using current and slowly evolving privacy standards, not some broad new catch-all. Asserting that good laws are "perfectly possible" is a little tragic given our long history of poorly designed laws. Not to mention mismanaged and corrupt regulatory agencies.

Silhouette 11y ago
It's certainly true that we have problems both with our legislators not understanding technology and with their apparent willingness to help special interest groups more than their electorate at times. However, those are orthogonal issues to the ethical principles and the technical aspects here. The same level of technology that can now be used to compromise privacy could also be used to help protect it, and we can surely do a reasonable job of distinguishing between the kind of inadvertent background imagery that comes from living in a world with other people and the kind of systematic, targeted collection and use of data that leads to some of the more serious privacy risks with the new technologies today.