Vernor Vinge’s mind-expanding novel A Fire Upon the Deep (1992) contains numerous postings on “the Net of a Million Lies.” But unlike the Usenet of our time, most of the senders have institutional names, like Khurvark University or Sandor Arbitration. I have assumed that these represent consensus cooked up by AIs from internal discussions.
Now Sean Kirby says that Columbia Newsblaster (a project of the Natural Language Processing Group at Columbia University) automatically “trolls hundreds of news sites, collects all the information from the articles, and summarizes them with links . . . . its computer-generated summaries are virtually free from any possible taint of bias.”
Oh? Newsblaster itself lists seventeen sources (Yahoo, ABCNews, CNN, Reuters, LA Times, CBS News, Canadian Broadcasting Corporation, Virtual New York, Washington Post, Wired, FOX News, NY Post, USA Today, Science Magazine, BBC News, Nature Magazine, Lycos) some of which repeat each other; Yahoo and Lycos get most of their stuff from Reuters. The summary will be as unbiased as the median of mainstream news; that beats reading just one paper, but is it revolutionary?
Unfortunately there’s no obvious link to information on how Newsblaster works.