FROM COOKIES TO MACHINE LEARNING

Cookies are a small piece of text installed by a server into one’s machine. The server can also be different from that visited (so-called third party cookies) for example, like how advertisements work – one can be on a different website and yet receive advertisements from a website visited previously. The cookies allow various features including authentication, shopping basket or personalisation of a website. However, in the last years, they have been criticised because cookies can be extremely intrusive and track a user’s activities inside a website and across the internet.

Recently, one of the monopolist of cookie technology, Alphabet Inc., the owner of Google, has recently declared that they are planning to phase out cookies, especially their paramount analytic cookies, which enable detailed statistic of a website. The new technology they want to implement is a completely different concept and it is called “Federated Learning of Cohorts”. This new interface should allow the same features of cookies applying machine learning algorithms to analyse users’ data. This kind of algorithms pose many problems: not just data protection concerns, but the outcomes could be problematic. As the algorithms needs to grind clean datasets, they can easily suffer from false positive or false negative results, when datasets are corrupted, incomplete or simply incorrect. It will be interesting to see how Alphabet Inc will implement this technology and what would be the outcome, especially in relation to the forthcoming e-privacy regulation.