• Gsus4OP
    link
    fedilink
    arrow-up
    8
    arrow-down
    1
    ·
    edit-2
    3 months ago

    I don’t care which way the causation goes or whether it comes from the name choice and preferences of the parents (yea, bad title), what surprised me is that ML can predict a name based on a face at all, there should be no correlation in first place.

    • Windex007@lemmy.world
      link
      fedilink
      arrow-up
      19
      ·
      3 months ago

      I disagree. Age biases the statistical likelihood of a name. Features that hint in any ethnic direction. Also some visible features might bias towards urban vs rural upbringing which also would also lean names in another direction.

      • Gsus4OP
        link
        fedilink
        arrow-up
        5
        ·
        edit-2
        3 months ago

        Oh yea, baby name trends, ethnicity, I don’t know how heterogeneous the datasets they are using are when they mention the strength of the correlation, I was imagining it would distinguish a John from a Peter, but there could be a much lower correlation in that subset.