3 ms·

while researching the blog post it crossed my mind that a technology capable of judging aesthetic value may also therefore be capable of recognising when it (or

while researching the blog post it crossed my mind that a technology capable of judging aesthetic value may also therefore be capable of recognising when it (or a sister technology) has generated poetry / literature that has aesthetic value. in other words they are complementary. i was only incidentally aware of the computational literary criticism movement, so obviously this will now inform my continuing research. thanks very much for that, and for the pointers. it's true what you say, there is a danger of isolation from both communities. i'm optimistic though. the traditional divide between science and humanities (with computation leaning towards the former) is, in many ways, a false dichotomy. i hope over time to demonstrate why this is so.

abathur 12y ago
When it comes to actually quantifying value things are still pretty crude as far as I've seen. I can't recall seeing much that is this aspirational yet. There's an interesting site/paper from 2013 (http://poetryassessor.com/poetry/ http://poetryassessor.com/poetry/) which attempts to rate a subset of poems on a amateur-to-professional scale. Concepts like "amateur" and "professional" and aesthetic value are slippery even for those of us who really enjoy poetry. It's worth realizing that aesthetic judgement can break down into a number of sub-tasks, most of which take for granted that we mean something like, "evaluate whether or not it is statistically likely that a blood-and-bones human will like this poem." I think that may be a trap. I think I'm far more interested in what poetry the computer finds interesting than in what poetry the computer thinks I will find interesting. The analogue in human experience is the difference between a poet writing poems they think their audience (be it end readers, or the slush readers at a given journal) will like, and a poet writing poems they themselves find interesting.
- thundercomb 12y ago
  Yes, I tend to agree with you on both counts … Quantitative and statistical approaches are unlikely to uncover poetic aesthetics by themselves. If economics has turned out to be hard to predict, after more than a century of quantitative efforts in that direction, I don’t hold out much hope for poetry or literature in general. I was thinking more along the lines of machine learning and language learning as starting points. On the second point, this is quite an exciting view to take in my opinion. The far end of the spectrum is that we begin to see AI as “subjects” in their own right. Not necessarily “flesh-and-bones” subjects who are human-like, but different entities with their own possibilities, and yet capable of interfacing with humans. Would we find their aesthetic tastes understandable? Could we learn to love it?
  - abathur 12y ago
    The reason my interest is in "what they find interesting" is because so much of poetry is surprise. We rarely praise a poem that uses the words we expect right where we expect them, and ended up right where we thought it would. (aside: check your twitter mentions, please?) It seems at least plausible to me that non-human agents could be Good For Poetry as long as we can clear a very specific hurdle. The hurdle is the fact that most people see art as something an artist--a human full of agency and intentionality--makes to communicate or share beauty. Obliquely, I think this definitional hurdle is a big part of why many struggle to find value in art produced from modernism on. I offered up this alternative definition of art (in the context of whether software can be art) in another thread a few weeks ago: Art is something--a sort of significance--we experience, and not really something we make. When we say we're "making art", we're making an object or giving a performance that we hope will induce others to have the experience of art. In reality, "making art" is an interpretative process; art, like love, is an experience--not an object. The point is that software most certainly induces people to have the art-experience, both intentionally and unintentionally. It's probably less common for source code to induce the art-experience (intentionally or not), but I suspect this too happens.
    - thundercomb 12y ago
      i like your definition of art as an experience; and the creation of art, then, as the communication of experience. there are many different types of experience, which accounts for the diversity of art.

abathur 12y ago

When it comes to actually quantifying value things are still pretty crude as far as I've seen. I can't recall seeing much that is this aspirational yet. There's an interesting site/paper from 2013 (http://poetryassessor.com/poetry/ http://poetryassessor.com/poetry/) which attempts to rate a subset of poems on a amateur-to-professional scale. Concepts like "amateur" and "professional" and aesthetic value are slippery even for those of us who really enjoy poetry. It's worth realizing that aesthetic judgement can break down into a number of sub-tasks, most of which take for granted that we mean something like, "evaluate whether or not it is statistically likely that a blood-and-bones human will like this poem." I think that may be a trap. I think I'm far more interested in what poetry the computer finds interesting than in what poetry the computer thinks I will find interesting. The analogue in human experience is the difference between a poet writing poems they think their audience (be it end readers, or the slush readers at a given journal) will like, and a poet writing poems they themselves find interesting.

thundercomb 12y ago
Yes, I tend to agree with you on both counts … Quantitative and statistical approaches are unlikely to uncover poetic aesthetics by themselves. If economics has turned out to be hard to predict, after more than a century of quantitative efforts in that direction, I don’t hold out much hope for poetry or literature in general. I was thinking more along the lines of machine learning and language learning as starting points. On the second point, this is quite an exciting view to take in my opinion. The far end of the spectrum is that we begin to see AI as “subjects” in their own right. Not necessarily “flesh-and-bones” subjects who are human-like, but different entities with their own possibilities, and yet capable of interfacing with humans. Would we find their aesthetic tastes understandable? Could we learn to love it?
- abathur 12y ago
  The reason my interest is in "what they find interesting" is because so much of poetry is surprise. We rarely praise a poem that uses the words we expect right where we expect them, and ended up right where we thought it would. (aside: check your twitter mentions, please?) It seems at least plausible to me that non-human agents could be Good For Poetry as long as we can clear a very specific hurdle. The hurdle is the fact that most people see art as something an artist--a human full of agency and intentionality--makes to communicate or share beauty. Obliquely, I think this definitional hurdle is a big part of why many struggle to find value in art produced from modernism on. I offered up this alternative definition of art (in the context of whether software can be art) in another thread a few weeks ago: Art is something--a sort of significance--we experience, and not really something we make. When we say we're "making art", we're making an object or giving a performance that we hope will induce others to have the experience of art. In reality, "making art" is an interpretative process; art, like love, is an experience--not an object. The point is that software most certainly induces people to have the art-experience, both intentionally and unintentionally. It's probably less common for source code to induce the art-experience (intentionally or not), but I suspect this too happens.
  - thundercomb 12y ago
    i like your definition of art as an experience; and the creation of art, then, as the communication of experience. there are many different types of experience, which accounts for the diversity of art.