• A personal recommendation system which constantly trains locally on data from our day-today activities and offers the model on demand to the services seems like an straightforward way to address this need-gap.

    Our smartphones are capable enough for this purpose, As they are already being used for training models but the barrier would be to access our activities on other platforms due to the closed nature of apps, smartphone OS and zero interoperability. Web may not be as constrained, So I don't see a a major constraint in developing an add-on which tracks activities on demand and stores the data locally.

    But large platforms do provide some way to export our bulk data for legal compliance and perhaps this might be an easier way to fetch our data for training.
  • Voted!
    Need karma! Please check submission guidelines.
    Why pay twice?