Idea for very positive/profitable use case of AI that also has the potential to make society significantly worse if made too easily accessible: According to https://apple.news/At5WhOwu5QRSDVLFnUXKYJw (which in turn cites https://scholarship.law.duke.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=1486&context=dlj), “ediscovery can account for up to half of litigation budgets.” Apparently ediscovery is basically the practice of sifting through a tremendous amount of documents looking for evidence of anything that might be incriminating. An AI that can consistently find all potentially incriminating evidence among a trove of otherwise mundane documents would probably save a tremendous amount of time/money for lawyers, leaving them able to serve more clients or spend more quality time on existing clients. Given the shortage of defense lawyers in the US, this could do a lot of good. On the other hand, the same AI would basically serve as a “cancel machine.” Right now if you want to cancel someone for something they said in the past, you have to either manually sort through their past public statements (which is rarely done even for most politicians due to time expense), or to naively search for specific keywords, which—unless they’ve used some horrible slur—is rarely helpful. Such an AI would thus make social cancellation even easier than it is now, which would arguably have net negative impact. I could see an argument pro-cancelation though, so I place significant uncertainty on that. Ultimately, is this something that should or shouldn’t be purposefully developed?
Idea for very positive/profitable use case of AI that also has the potential to make society significantly worse if made too easily accessible: According to https://apple.news/At5WhOwu5QRSDVLFnUXKYJw (which in turn cites https://scholarship.law.duke.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=1486&context=dlj), “ediscovery can account for up to half of litigation budgets.” Apparently ediscovery is basically the practice of sifting through a tremendous amount of documents looking for evidence of anything that might be incriminating. An AI that can consistently find all potentially incriminating evidence among a trove of otherwise mundane documents would probably save a tremendous amount of time/money for lawyers, leaving them able to serve more clients or spend more quality time on existing clients. Given the shortage of defense lawyers in the US, this could do a lot of good. On the other hand, the same AI would basically serve as a “cancel machine.” Right now if you want to cancel someone for something they said in the past, you have to either manually sort through their past public statements (which is rarely done even for most politicians due to time expense), or to naively search for specific keywords, which—unless they’ve used some horrible slur—is rarely helpful. Such an AI would thus make social cancellation even easier than it is now, which would arguably have net negative impact. I could see an argument pro-cancelation though, so I place significant uncertainty on that. Ultimately, is this something that should or shouldn’t be purposefully developed?