3 ms·

Relevant presentation by James Mickens: https://www.usenix.org/conference/usenixsecurity18/presentation/mickens https://www.usenix.org/conference/usenixsecurit

• mjburgess 5y ago
  The speaker starts, "we are computer scientists, we should be more sceptical". This is a little odd: computer scientists are mathematicians, not scientists: therein lies the issue. I think the more this is really absorbed, the clearer things become. * Why are ai models of intelligence logical and not dynamical? * Why are models rules of association, rather than interpretative frameworks? * Why is a statistical association deployed without empirical verification? * Why arent empirical (& verifiable) criteria given against predictive systems? * Against the mechanisms of these systems? * Why is ML an experimental discipline without experimental tests? * Why doesn't ML research employ tests of significance? * Why is the ability to experiment seen as incidental to intelligence? * Why is ML sold with a scientific technical vocabulary whose meaning is given metaphorically? * Why arent ML researchers aware these are metaphors? * Why aren't scientists of intelligence researcher reviewing and informing the claims of the artificial intelligence community? * Why do some of the most prominent "scientists" in the world advocate that understanding the world is a matter of mining datasets? * Why don't these scientists defend the need for actual science in producing accurate models of reality? * Why aren't computer scientists more sceptical?
  • ampdepolymerase 5y ago
    I am not sure if you are being deliberately obtuse or simply unfamiliar with how ML is designed and implemented. Almost every single point you mentioned does happen in practice. Most are limited by budget and scale, just like real world experiments.
    • mjburgess 5y ago
      I believe I'm very familiar. Where do you think I've missed something?