Bios robotlab writing robot I have seen a number of articles in recent weeks on the topic of whether a computer could write a book. One software company has software that will automatically write newspaper articles on certain topics. A college business professor is working on software that  has already written 200,000 nonfiction books, half of which he has for sale on Amazon.

I haven’t yet seen an example of what these software programs produce, but I’m extremely skeptical of their quality.

Let’s do an experiment. Following are two poems. Each is a xenia epigram, a poetic form originally found in Latin literature. One was written by poet Luke Wright for the BBC. The other was written by a computer after being given instructions about the poetic form.

Can you tell which is which?

Here they are:

To Truth, by ??????

To truth I offer this thanks,
when needing something like reality
When I’m writing and drawing blanks,
I almost settle using actuality.

I am in search of more,
trying to sing your praise!
It’s you I very much adore,
lacking in so many ways.

To Felicity, by ??????

Felicity, my dear, my thanks
the cheque you sent was great.
Tomorrow I’ll go to the bank
my rent’s already late.

And sorry for the shoddy rhyme
I’m tired, I’m not on it,
perhaps if you send more next time
I’ll scribble you a sonnet.

Which was written by the computer and which by the human? Leave your guess in the comments. Don’t look at the other comments until you’re ready to make your guess.

Can Robots Write Poetry? Or Fiction?

9 thoughts on “Can Robots Write Poetry? Or Fiction?

  • Tuesday, 30 October 2012 at 1:27 pm
    Permalink

    I’m guessing the first one was written by a computer, whereas the second was written by a human.

    I’m actually sort of impressed because the first stanza of the first poem almost seems like it’s making a profound tongue-in-cheek statement about how you can always fall back on Truth if you can’t think of something else to write. That would fit well coming from a writer for the BBC (Zing!). But then the poem devolves into something trite and unrelated. And throughout, it feels like it’s using words that don’t quite fit together, just to get the timing right.

    The second poem is coherent throughout, and pretty humorous. It doesn’t feel like it’s straining, and it feels fresh and creative.

    Computers may be getting to the point where they can churn out reports about events in the stock market or basketball games, but (unless I guessed wrong) they don’t grasp the finer points of poetry yet.

    Reply
  • Tuesday, 30 October 2012 at 2:27 pm
    Permalink

    I vote that the second one was written by Luke Wright. It sounds British to me! :)

    Reply
  • Wednesday, 31 October 2012 at 8:31 am
    Permalink

    My guess is Felicity was written by a human. It has wit and humor, which I daresay a computer cannot yet mimic. The first one, “To Truth” feels mechanical and soulless to me, and doesn’t really have anything meaningful to say.

    Reply
  • Wednesday, 31 October 2012 at 7:03 pm
    Permalink

    I’m hanging with my in-laws, and we all think it is the second due to its level of humor.

    R.

    Reply
  • Wednesday, 31 October 2012 at 10:45 pm
    Permalink

    I vote for the second, too. At first I wasn’t sure what to think — I was expecting the computer to be obviously clunky somehow, and both were technically excellent — but the second has more than technique. The touch of humor, the idiomatic speech and slant-rhyme in the second stanza…and the way it’s both figurative and literal is brilliant.

    If the second one is the computer’s poem, we’re all going to look nohow, aren’t we.

    Reply
    • Saturday, 3 November 2012 at 9:25 am
      Permalink

      Alright, I don’t think we’re going to get anyone else voting on this. It looks like the voting was unanimous, and everyone was correct! Felicity was indeed written by a human and Truth by a computer. Artificial intelligence still has a ways to go, it appears.

      Reply
  • Sunday, 10 November 2013 at 4:54 pm
    Permalink

    It was an interesting programme especially coming in the ‘Something Understood’ series. But the defenders of future robots (of which I am one) did not have enough time (or perhaps enough knowledge) to indicate how deep the challenges are. Humans are products of layers upon layers upon layers … of evolution, and I suspect that until we know far more about the intermediate layers (and the many discontinuous evolutionary changes — since they can’t all be smooth, continuous changes) we’ll only have machines with very limited and shallow capabilities. Moreover, it’s possible that digital computation is not rich enough and we need chemical computers with powerful mixtures of discrete and continuous information processing, along with parallelism at different levels of abstraction (parallel control of muscles involved in speaking, along with control of multiple linguistic and other cognitive competences, along with control of multiple types of motivation — to name a few examples).

    Following Alan Turing’s 1952 paper on chemical morphogenesis I use ‘Meta-morphogenesis’ as a label for the project addressing these problems that he might have pursued had he lived longer.

    In the meantime, don’t expect anything but isolated demonstrations of human-like abilities in robots/computers, for the foreseeable future. And don’t expect the current techniques of neuroscience to reveal what’s really going on in the most sophisticated known type of computer.

    Reply
  • Sunday, 10 November 2013 at 5:05 pm
    Permalink

    Sorry: in my comment I left out an introduction saying that I had heard the two poems in John McCarthy’s presentation in ‘Something Understood’ http://www.bbc.co.uk/programmes/b03gtnng , tonight.

    I agree with the comments above about the differences between the two poems — but deciding took a bit of thought.

    Reply
  • Monday, 18 November 2013 at 1:52 pm
    Permalink

    Aaron, thanks for your comments. I was a linguistics major in college and studied a smidgeon of cognitive science as well. Artificial intelligence definitely has some big challenges, far more than anyone dreamed in the early days of computing. I have never heard of the possibility of chemical computing, but it sounds really intriguing. I’ll have to look into Turing’s paper. Thanks for bringing it to my attention.

    Reply

Leave a Reply