As an aside, my worry with AI/ML is that we’re creating these giant, un-debuggable systems that can’t be fixed when specific problems arise. Every time I see complaints about YouTube/Facebook/<some other company> giving nonsense recommendations to people I think “How would they even fix that? Do they even know how their recommendation systems work?”.
One of the great things about software is that when something is broken, you can go right in there and fix it. With ML, I can imagine getting to a point where our only option is to take the black-box we’ve created to therapy and hope for the best.... :)
As an aside, my worry with AI/ML is that we’re creating these giant, un-debuggable systems that can’t be fixed when specific problems arise. Every time I see complaints about YouTube/Facebook/<some other company> giving nonsense recommendations to people I think “How would they even fix that? Do they even know how their recommendation systems work?”.
One of the great things about software is that when something is broken, you can go right in there and fix it. With ML, I can imagine getting to a point where our only option is to take the black-box we’ve created to therapy and hope for the best.... :)