mknepprath’s avatarmknepprath’s Twitter Archive—№ 27,600

    1. …in reply to @mknepprath
      The AI apocalypse is a ways off yet
      1. …in reply to @mknepprath
        “An algorithm trained on a biased dataset found that answering yes to any question that begins with ‘Do you see…’ would result in 87 percent accuracy.”
        1. …in reply to @mknepprath
          1. …in reply to @mknepprath
            My fav point made in this book is that machine learning does not ~recommend~, it ~predicts~. For instance, when an algorithm trained on job resumes selects the “ideal” resume, it’s selecting based on what it predicts the humans that generated its training data would have selected