Thursday, 29 March 2012

Can a computer write poetry? How are science and art distinct?

Obviously these are two very different questions, and I'm certainly not going to suggest that computers can somehow do science but not art, but still, I hope by the end of this entry you will agree that I have justified connecting them like this.
I should give some background first though. The first question came up during a lunch break at work when one of my colleagues was talking about a computer programmed to write poetry. It worked by first being fed thousands of lines of actual human poetry, it was then able to replicate its own based on rules it had found (mostly I think simply which words it had seen paired together*). His argument was that, similarly to the way they have in chess, perhaps computers ability to search huge spaces efficiently would allow them to also create poetry better than humans one day. For reasons I still can't quite explain I had an immediate and powerful gut reaction against this idea. I couldn't say why at the time, but it seemed axiomatic to me that no computer could ever write poetry as well as a human. I've since spent a while searching for a reason why I felt so strongly about this and also trying to come up with a more reasoned argument. This entry will be all about the ideas I came across in searching for those answers and, ultimately, where they led.

The first thing I want to make absolutely clear is that I do not think we, as humans, are in some way 'more' than machines. I don't believe in (or think there is a good argument for) any basic essence or soul which will allow us to create art where computers never can. I think for almost every task, we will one day create robots which can do it better than us. I'm taking it as a given that we are essentially just biological machines.
However what I realised is that there is something involved in creating art which robots can't do (**). That is, when you make really great art, of any form, you are picking some corner of your brain and finding a way to show it to other people. This is why we enjoy these things, because when we look at a great painting we are looking at a map which guides us to those same areas in our own brain. It is about sharing the shape of our own heads with one another.
In this sense, I hope it's clear that computers can never do this. Not because they can't think better than us, but because they will never have those same pathways to discover for themselves. Of course, you can argue that they might be able to if we made them very similar, or identical, to us, with the same internal patterns, but it isn't clear to me why we would do that (and besides at that point the question probably shifts from 'can these robots create art?' to 'aren't these robots just humans themselves?). This doesn't mean they will never be able to create art. I'm sure there will be areas of our brains which are well enough mapped that they are able to work within them, but we, as humans, will always have access to more of our own minds (and from a more uniquely personal perspective).
[I think the other interesting thing about this argument, is that once robots become complex enough that they would be able to create art which is meaningless to humans, but which is beautiful to other robots.]

So then, how does this relate to the second question? Well having seen this perspective, it allowed me to consider creating in a different light, leading to comparisons between art and science in finding the value of the former.
Firstly, before I get into this, I want to admit openly that it is doing a brutal disservice to both thought and language to emphasise this distinction. I've found it helpful to consider for the purposes of clarifying my own artistic endeavours, but generally speaking I'm not at all a fan of dualism of this type.
Right, that disclaimer out of the way I think that once you frame art in the way I have here, it becomes easy to see a division between these two things. That is, science is the knowledge of external things (things that are dropped will fall, they will fall at some particular rate based on a set of verifiable equations), it is a set of rules which can be easily verified in the real world by anyone who tries. However art is a more internal knowledge, 'this Picasso is very moving' is a statement about the reality of what the painting does inside the human mind. Even if we were to hook a person up to study their brain patterns, what does “moving” mean, the statement says that the painting has a profound effect, but something profound inside my head may not translate to a big observable effect. This is especially true if we suppose that what the painting is doing is activating dormant possible links in our own brain, it is showing us new possibilities, but the only way to see those possibilities is to experience them first hand.
It's easy to see this second type of knowledge as invalid or useless, after all if it's unverifiable then what use is it? But there are a huge number of shared and repeated experiences in this area. Many people may be moved by the same song or same lines of poetry and artists may have ways of touching you which are repeated across many apparently different pieces.***

So then what are these niches in our brains? We can see examples of them in the way that so many movie plots and pop songs tread such familiar ground, they are leading people along paths which they already know, taking them step by step along that road because there are some things (mostly of a wish fulfilment variety) which we never get tired of experiencing.
This is a bit of an aside, but I don't think this is necessarily as hateful as it seems. Obviously it's difficult to associate your standard romantic comedy with any concept of art, it is fairly cynically leading us along a known route without doing anything interesting along the way, but this doesn't have to be true. Take van Gogh for example, with most of his paintings he isn't showing us anything new, we've all seen a field at night, but he takes the time to point at the stars and say, 'look your brain can see them like this'. In this way even a tired romance story or hero's tale can be made refreshing and new if the artist, as they lead us along, takes the time to point out previously unseen landmarks or perspectives along the way.
Okay, aside over. I think the whole point about these niches is that they are knowledge, but knowledge about the shape of us as people. When I learn how calculus works, I build new circuits which allow me to do conversions and operations using these new rules, but when I see the Starry Night I am shown how close that image of the stars is to the part of me which produces exultations and wonderment. It isn't about me learning to love the sight of stars, it is about me seeing that the way I am wired I always loved it (of course, that's an oversimplification, but I hope it suffices to make my case).
I think there is a really good argument that we need this, that it is of great value to us as people. We spend so much time pottering around in our heads alone, that having someone else occasionally take us by the hand and say 'look, you missed this' is incredibly valuable, whatever form it takes. It doesn't just serve to show us how similar we are to one another, but it also teaches us about ourselves, extending our conceptual limits. We instinctively know this too, it's wonderful every time it happens, experiencing that burst of wonder at the freshness of it.

So, I should try to wrap up this uncharacteristically long entry. I hope I've shown that this perspective answers the first question rather neatly. That if we're talking about little pieces of the human mind, then while computers might stumble across some, we, as humans ourselves, will always have a huge advantage. Though I think it's important to say that this isn't saying that computers will always be limited to be less than us. Rather I think it's more accurate to say that WE, as humans, are limited, and that art is the expression of the bounds of that limitation.
As for the second question, well honestly I hope I haven't fully answered it, because I don't think there is much value to demarcating that dividing line more clearly. My intention was more to use the question to have a conversation about what the art side of the equation means, and what it ought to mean, at its best.
New perspectives are always interesting, partly because they enlighten us as to the skews of the old ones, but also because they force us to shift and accommodate them and, I often feel, it's by moving that we know we are alive.

*[Sadly my colleague wasn't able to provide a link to the paper in question, though it was written in Chinese (for replicating ancient Chinese poetry) so you're not missing out on much]

**[realised while editing that I quickly switched from computers writing poetry to robots creating art. I think that's because it's just a far more vivid image (while still describing essentially the same problem), so I've left it in]

***[note that, although what I'm talking about isn't necessarily emotional (we're talking about areas of our minds, little bits of potential experience), all of the words I'm being forced to describe experiencing them are emotional. I think this is because experiencing these things causes us to have an emotional reaction, to feel that charge of newness, and as a result these are the words which we've come to associate with them]


  1. Reading the section about using art to conceptualize different perspectives reminded me of this Stephen King quote from an interview:

    'Children see things from a different perspective. People respond to this perspective. It doesn’t really die. It atrophies and lies dormant. And I get paid to show people that different perspective. It’s like exercising a muscle, rather than letting it go slack.
    That’s why people pay writers and artists. That’s the only reason we are around. We’re excess baggage. I can’t even fix a pipe in my house when it freezes. I am a dickey bird on the back of civilization.
    I have no skill that improves the quality of life in a physical sense at all. The only thing I can do is say, “Look here, this is the way you didn’t look at it before. It’s just a cloud to you, but doesn’t it look like an elephant?” Somebody says, “Boy! It does look like an elephant!” And for that, people pay because they’ve lost all of it themselves.'

    I find it interesting to think about whether fully advanced robots, with their absence of biological emotions would still have a drive to create and consume art. The conclusion I've come to is yes; in that I believe "art" in its simplest form is the acquisition of, and creation of, new connections in data. It's a sort of bridge between the art/science duality in that they can practically be two sides to the same coin. Where science focuses on the objective, and traditional art closer to the subjective; with some room for overlap of course.

    1. Hi,
      Thanks for the quote. I really like hearing perspectives on this, particularly from creators themselves. I'm not sure I entirely agree with his definition, at least I think, as I now do about my own one, that it is only part of the truth, but it's definitely interesting to see that that is how he himself thinks about it.

      As to robots creating art. I definitely think they can and will at a certain level of advancement, the only question in my mind is how much crossover there will be between robot and human art (and what that crossover says about how close we are).
      Honestly I think it will likely be the same with aliens as well that, foregoing cultural influences, the areas where our art does and doesn't intersect would neatly define the differences between how our species understand the universe (there's a science fiction novel in there somewhere I'm sure).