Slashdot Mirror

← Back to Stories (view on slashdot.org)

Dartmouth Contests Showcase Computer-Generated Creativity

Posted by samzenpus on from the man-or-machine dept.

An anonymous reader writes: A series of contests at Dartmouth College will pit humans versus machines. Both will produce literature, poetry and music which will then be judged by humans who will try and determine which selections were computer made. "Historically, often when we have advances in artificial intelligence, people will always say, 'Well, a computer couldn't paint a sunset,' or 'a computer couldn't write a beautiful love sonnet,' but could they? That's the question," said Dan Rockmore, director of the Neukom Institute for Computational Science at Dartmouth.

4 of 50 comments (clear)

  1. Still misunderstands the Turing Test by John+Allsup · · Score: 4, Interesting

    A major feature of the Turing Test is that it is interactive: later lines of conversation are sent to the computer after earlier lines of conversation are known. A 'Turing Test' where someone is given a transcript and asked to decide who in the transcript is human or computer is a much weaker test. A more 'Turing Test' like test would be one where I give the computer/human a brief to draw a sketch in, say, an hour, and see the end result, then get to ask for a few more sketches. The kind of thing they are developing here is more like an algorithm to generate a single line of conversation given knowledge (pre algorithm design) of the previous lines in the conversation.

    Even so, computer generated art is something which should be explored, and then the art will be in finding new clever ways to use the computer as an artistic medium.

    --
    John_Chalisque
    1. Re:Still misunderstands the Turing Test by PhilHibbs · · Score: 3, Interesting

      It doesn't "misunderstand the turing test". It's a different test. No computer has yet passed the Turing Test, so doesn't it make sense to have other tests?

      Let's assume that the Turing Test is a good test for AI. It's debatable, but let's accept the premise. We don't have good AI yet, so what is the point in testing what we have against a test for good AI? Doesn't it make sense to aim for something with a lower bar, achieve that, and then tackle the tougher problem? When I was at school, we didn't set the high jump at olympic champion levels. We set it at a level that was a stretch for us but still achievable.

    2. Re:Still misunderstands the Turing Test by AK+Marc · · Score: 3, Interesting

      Let's assume that the Turing Test is a good test for AI. It's debatable, but let's accept the premise. We don't have good AI yet, so what is the point in testing what we have against a test for good AI?

      The same reason we tested inferior chess programs against grand masters. So we could learn the weaknesses, and improve upon them. So testing an AI improves the AI, like testing a chess program lead to improvements in the chess program.

      When I was at school, we didn't set the high jump at olympic champion levels.

      When I was at school, the pool was olympic length, and the high jump could be set at olympic heights, as well as lower ones. So you do what you can, and compare your failure to the desired levels. It's not just the pass-fail as given. But it lets you compare your failure to the ideal.

      Like lasting longer in chess against a grand master, or fooling more people in a Turing test (or lasting longer in the question sequence until the tester correctly identifies the AI).

      --
      Learn to love Alaska
  2. Re:both will produce "literature"? by mlts · · Score: 4, Interesting

    With the boilerplate novels and cookie-cutter movies being cranked out, I wonder if a computer would eventually be a better writer/artist than what we have now.

    One could even add music into the list as well, where a marketing person could click on an interface, randomly select what an album would have for songs, and the computer would create a band (name, color scheme), write the lyrics, compose the pieces, even pick people from YouTube who would become the band members. Or toss the physical band members, have an avatar like Miku and call it done. Pop music can just be relegated to a cronjob that fires off, weights lyrics on statistics gleaned from ad sites and news articles, makes the songs, mixes/masters the album, and spits out music for the music stores, no human effort needed in the creativity process.

    Similar with movies. The computer would grab weighting on what social topics are being thought about when the movie is created (so the movie has some impact), create some characters, follow a meta-script to generate the dialog, generate terrain and scenery render the scenes and CGI action, and out pops a blockbuster hit at the push of a button, no actors needed.