Art v. Science

[k-j]

See Figure Eight's comment (21st on the thread I think): viewtopic.php?t=1823.

[k-j]

For me, great art and great science approach the same point of perfection, allying inspiration with order. It's only when the standard slips below greatness that one or the other starts to go, and we call it art or science. Great science = art and great art = science.

[Bombadil]

Art and philosophy take over when science can see no further. I see evidence of scientific method in art and I think art has pushed science into realms it could not have gone. Perhaps, it's that science has no imagination or inspiration without art, and art has no effective device in the absence of scientific method. Ah, I've gone and confused myself.

Cheers,

Keith

[figure eight]

I thought it might be easier to copy the main point into this thread:

Surely art attempts to capture the fundamental truths of science: the human form, nature, the mind & body's emotional responses.

I know I might have trouble arguing this side of the debate on a poetry forum but wondered what others thought.

Please don't get me wrong I love poetry, but I see it as being similar to mathematics. It's a system we've come up with to represent feelings, moods, etc that's understandable and makes sense to us. It's the same with maths; it's is our own system of rules that we try to impose on nature and science.

And very briefly to fight the other corner; science and maths can never really be used to truly understand great poetry. There are no computer programs capable of writing poetry to match that produced by humans.

[k-j]

There are no computer programs capable of writing poetry to match that produced by humans.

...yet.

20 years ago it was inconceivable to the (majority of the) chess community that a computer would ever match a grandmaster in a series. But it wasn't inconceivable to the computer science community, and as it turns out the so-called experts - the chess players - were wrong. Speaking chess, that seems like a good example of "art" / inspiration and "science" / logic meeting at the highest level.

[Bombadil]

I think for mock-jobs computers would do just fine; a set syllable count, a rhyme scheme and set theme. But free verse and a non-specific theme? The computer would have to feel. How do you program angst and discontent?

[figure eight]

Good point. There is something about chess that brings the two a little closer together.

This has reminded my of a poem I read a while back. I can't remember the name but it's about the way science and poetry are viewed by the world. I think it might've been translated from Russian? Anyway the first couple of lines go something like:

Looks like science is in honor.
Looks like poetry is not.

(Or it could be physics is in honor rather than science?)

I'm pretty sure i read it in the "Scanning the Centruy" book of poetry by penguin but can't be sure. Anyway, I quite like it, if I find out what it's called I'll edit this. I'm sure you'll have seen it anyway.

[figure eight]

Bombadil,

I've been at a computer most of today programming and I know that a computer can certainly make me feel angst and discontent. Maybe they do understand...

...OK I think i've been in front of this screen for too long.

[k-j]

I think for mock-jobs computers would do just fine; a set syllable count, a rhyme scheme and set theme. But free verse and a non-specific theme? The computer would have to feel. How do you program angst and discontent?

Free-verse isn't fundamentally different to any other kind of poetry, it's just more difficult to get right. As for a non-specific theme, I would argue that no poem is solely about a subject - it's about the poet's reaction to a subject. Every poem - every piece of writing - carries "feeling". So if you accept that a computer could match a human in writing a sonnet about, say, daffodils, then I don't see why it couldn't match up at writing free-verse about innocence, gladness, or any of the "feelings" generally connoted by daffodils.

But not yet. But before very long.

The other point to consider is the Turing test: http://cogsci.ucsd.edu/~asaygin/tt/ttest.html It's immaterial whether the poet - human, machine, whatever - "feels" the subject; all that matters is that the reader believes it. It's basically a con-trick. Perhaps that's one way of ditinguishing between art and science - art, like religion, can function without a consensus view as to its validity.

[Bombadil]

But not yet. But before very long.

I shall then withdraw from the world at large, live off the land and await the apocalypse.

-K.

[k-j]

I shall then withdraw from the world at large, live off the land and await the apocalypse.

-K.

I think you're noble and savage enough as it is Keith, without going all <i>Walden</i> on us.

[Bombadil]

Once my first book is published at large (if ever) or I hit the lotto...

Walden Pond here I come.

I think I'm a few chromosomes short of human, anyway.

[pseud]

So a discussion of art and science has come down to this, aye?

kj -

Chess - this is proper subject matter for a good argument.

Computer-generated poems? Interesting. A bit like computer-generated chess? ah...no. Is that really a good example? A computer is only as good as its programmers. All one has to do is study a bit of game theory, logic, and a few chess strategy books written on the collective games of Bobby Fischer or Gary Kasperov, and presto-chango! - Deep Blue! Programmed to imitate - quite different from actively pursuing a strategy and developing habits and a style.

This imitation works because there are only 16 pieces with set moves and strict logical rules to adhere to on a 64-square grid. Checkmate=success.

But poetry, now that's just a bit different.

That requires taking the 600,000+ English words and programming a computer to assemble them - programming each defininition (or words with multiple definitions) accordingly. Also you'd have to account for rhyme, meter, etc., depending on the kind of poem.

Right about then you'd start to have problems.

1. meaning: how does a computer associate one scenario/image/noun with another? Even if basic sentences could be formed so that nouns could be connected with verbs and such, you'd have to make sure the computer had guidelines for symbolism. And what of abstract nouns? Illogical phrases? Redundancy?

2. craftsmanship: take any rhyming poem, for example - "orange" and "month" would be invalid endings. It would have to be programmed to form sentences in a certain manner to avoid parlor syntax tricks, and then not use two combinations of words we'd consider "crap rhymes," like "cat" and "hat," or "selfishness," "elfishness."

3. originality: whereas robots men have come up with so far can only imitate, this poem-writing task would require innovation. Unless that innovation could be reached logically, it wouldn't happen. And of course, a computer is still only as good as its programmers.

4. success in the abstract: what makes a poem good? Humans can't even figure that out; we simply know what we like. So how can we program a computer to aim at something like that? Strategy guides to Shakespeare? There is no "checkmate" here.

I don't think a computer would write anything worth a serious critics time. Not even a sonnet or something highly structured. Computer-generated haikus I could see as possible; that's about it though. Sure a sonnet on daffodils would be nice, but it would be a Hallmark poem at its very best outcome. You'd get some cheaply disguised breakdown of the definition of daffodil - no association between daffodils and say, a veteran of WWII walking in a nursing home parking lot picking flowers for his bed-ridden wife.

...

The only way this might come about is with quantum computers - who knows what they might be able to do. It's already confirmed they would be able to break any code that any government on the planet uses for classified information, so, who really knows...

Or of course there's always the Douglas Adams approach. "Come back in 7 million years..."

But chess...come kj, you must be joking.

- Caleb

[k-j]

I'm not saying that chess and poetry are equivalent or even similar, except in one respect: until recently chess was considered to call as much upon inspiration as logic, and therefore to not be "programmable" at the highest level. This was the consensus among chess players. What happened, as you say, was that increased processing power (and lot of brainy programmers) more than compensated for the lack of "inspiration" in the machine. It turned out that chess isn't, as had been thought, equal parts intangible "art" and hard science. Deep Blue imitates, you say? Of course it does. So do Fischer, Kasparov, and every other chess player. All they can do is imitate. Their genius, like Deep Blue's genius, lies in the breadth and spontaneity of their imitation. Similarly no artist produces masterpieces in vacuo, fully-formed, distinct and whole. I think most people, except a few crazed Romantics, accept that all art is a tradition, building on (or rebelling against) its precursor. Likewise, all games of chess - and yet no two games of chess or works of art are identical.

So this suggests that there's no fundamental impediment to a computer's poetry passing for that of a human. There are significant technical hurdles, as you point out, but remember: all that matters is the Turing test - the believability of the product. The source is irrelevant.

A computer is only as good as its programmers.

This is palpably untrue. Deep Blue is a far better chess player than any of its programmers, just as any great artist's talent quickly surpasses that of his or her early tutors.

[Poetry] requires taking the 600,000+ English words and programming a computer to assemble them - programming each defininition (or words with multiple definitions) accordingly. Also you'd have to account for rhyme, meter, etc., depending on the kind of poem.

Assuming three definitions per word, that's 1.8m words, an increasingly insignificant number in terms of computing power. Rhyme and meter are eminently programmable, as are traditional forms. More important than any of this would be to provide the machine with a copious library - the entire poetic canon and a guide to the morphology of English for a start - so that it could read and learn. Give it a list of common cliches to avoid.

how does a computer associate one scenario/image/noun with another? Even if basic sentences could be formed so that nouns could be connected with verbs and such, you'd have to make sure the computer had guidelines for symbolism. And what of abstract nouns? Illogical phrases? Redundancy?

Computers are good at associating things with other things. It's not impossible to conceive of a vast Venn diagram, in which "knife" overlaps to a greater or lesser extent with thousands of other nouns - "food", "bread", "blood", adjectives - "sharp", "lethal", as well as common adjectives like colours, and so on. But this is tough. But then no-one said it would be easy.

take any rhyming poem, for example - "orange" and "month" would be invalid endings. It would have to be programmed to form sentences in a certain manner to avoid parlor syntax tricks, and then not use two combinations of words we'd consider "crap rhymes," like "cat" and "hat," or "selfishness," "elfishness."

This isn't so tough. Easy to come up with a list of crap- or non-rhymes and to program syntax. Already possible.

whereas robots men have come up with so far can only imitate, this poem-writing task would require innovation. Unless that innovation could be reached logically, it wouldn't happen. And of course, a computer is still only as good as its programmers.

What is this "innovation"? Is it the same innovation beloved of chess savants? The wild-eyed Bobby Fischer innovation that stumped Spassky and wowed the world in Reykjavik? The innovation allied by Kasparov to intense work and willpower? Why could it not be reached logically? What is the fundamental difference, if any, between innovation and logical deduction? To anyone not in the know, Deep Blue seems highly innovative. Its moves are proactive; it actively seeks to win games. Who cares how we get there? It's the result that matters.

success in the abstract: what makes a poem good? Humans can't even figure that out; we simply know what we like. So how can we program a computer to aim at something like that? Strategy guides to Shakespeare? There is no "checkmate" here.

If humans can't figure it out, then we shouldn't expect a machine to do so (a familliarity with Shakespeare never hurt any poet, though, man or machine). The only test of success for a machine would be the same as that for a person: send the results off to poetry mags pseudonymously and attempt to build up a critical reputation, maybe even get a book published. A latter-day George Eliot.

The only way this might come about is with quantum computers - who knows what they might be able to do. It's already confirmed they would be able to break any code that any government on the planet uses for classified information, so, who really knows...

Exactly. I didn't say it was possible today. But soon; before very long; in principle. At least, you can't write it off. I think humans are predisposed to write off any science greater than our own (my own, your own) as an art. Art - poetry - are very complicated, but I don't see any evidence to suggest that they're supernatural, that there's any reason why we can't analyse them scientifically. Five hundred years ago everything was an art! The leaves on the trees, a thunderstorm, a canyon, they were all the artistry of God. Pure mathematics is described as an art because so few people understand it. Even accounting is sometimes described as an art (or a black art). Art and science both strive to make sense of and communicate the universe (external or internal). If they're successful, there's really nothing between them.

[pseud]

Perhaps we need to back up: are you saying that a computer may someday be winning awards with its poems, or are you saying merely that computers may someday be able to write poems that would pass as human (regardless of quality)?

Deep Blue is a far better chess player than any of its programmers, just as any great artist's talent quickly surpasses that of his or her early tutors.

I meant Deep Blue is as good at chess as its programmers are at programming, not chess...and besides, I happen to think that writers like Shakespeare, James Joyce, Dante, Murasaki Shiburu actually developed new ways of word association and expanded their craft. Are you saying that, if we had these super computers before, they might have seen these word associations as simply another possibility?

Rhyme and meter are eminently programmable, as are traditional forms.

well, that's not so cut and dry. These would be as tough as symbolism - "Cat" and "hat" are only "crap rhymes" in certain contexts. In a nice limerick, it may not be so bad, nor in a very well-done interweaving sonnet.

So too with meter. "and if our" might be three unstressed syllables, three stressed, stressed on "and," stressed on "if," stressed on "our," all depending on context, and sometimes depending on how a human might choose to read it because there is more than one choice.

You never answered my question about illogical phrases or redundancy...wouldn't it be more likely humans saw the computer's poetry as dull? - seeing as it ultimately would be a breakdown of the definition of daffodil, or its position on a diagram; perhaps an ambitious computer-generated poem would be interweaving "daffodil" with what's already been said about daffodils by the greatest poets in history. When humans do that it's usually called cliche.

Now I'm wondering about those poets who break the rules. Some of Donne's greatest stuff, for example, breaks meter for effect.

And then there is this loaded section:

What is the fundamental difference, if any, between innovation and logical deduction? To anyone not in the know, Deep Blue seems highly innovative. Its moves are proactive; it actively seeks to win games. Who cares how we get there? It's the result that matters.

But these moves are proactive towards an end. What end could you possibly aim for in poetry? There is no checkmate, when talking about the greats there is no point in which one says "this is better than this, 100%, accross the board to everyone." Some like Milton, others like Poe. Sure there are guidelines for good poetry because there are some things that generally work better than others...

I guess, if it comes down to passing off computer poetry as human, that might be a heck of a lot easier than making computer poetry as good as what critics generally consider "the best" or "some of the best." The tragic (ironic) thing about it is, you could puts words in a random number generator and come up with some Pollock/Kandinsky type poetry that might "touch" some people more than all this hoopla about form.

Interesting stuff, I'm not saying you're wrong necessarily but it's still not as easy as you're portraying...

[k-j]

Perhaps we need to back up: are you saying that a computer may someday be winning awards with its poems, or are you saying merely that computers may someday be able to write poems that would pass as human (regardless of quality)?

Nope, like I said, the criterion is that it has be poetry which is competent enough to garner some critical acclaim pseudonymously. This is the same standard to which we hold ourselves. I don't need a Nobel prize but I need more than just teen angst or "experimental" claptrap (sorry Arco).

writers like Shakespeare, James Joyce, Dante, Murasaki Shiburu actually developed new ways of word association and expanded their craft. Are you saying that, if we had these super computers before, they might have seen these word associations as simply another possibility?

I think so. If we refine the program sufficiently, the computer will be able to recognise which word-associations are entirely novel yet might just work.

You never answered my question about illogical phrases or redundancy...wouldn't it be more likely humans saw the computer's poetry as dull? - seeing as it ultimately would be a breakdown of the definition of daffodil, or its position on a diagram; perhaps an ambitious computer-generated poem would be interweaving "daffodil" with what's already been said about daffodils by the greatest poets in history.

Not at all. I don't see why computer poetry would have to be diagnostic or purely analytical in tone. It's rare that you'll see a (good) poem solely about daffodils and how yellow they are, anyway. As you say, most effective poetry will refer to daffies where, for example, there's "a veteran of WWII walking in a nursing home parking lot picking flowers for his bed-ridden wife". The subject here is the veteran, not the flowers, and I think it's reasonable to imagine a computer arriving at this image in a poem about a veteran.

Now I'm wondering about those poets who break the rules. Some of Donne's greatest stuff, for example, breaks meter for effect

.

Again, I think you could build a computer which was able to vary its application of the core "rules", just as drum machines are built to be ever-so-slightly off - intangibly off, in a way you can't detect except in the overall "feel" of the rhythm - just as humans are.

But these moves are proactive towards an end. What end could you possibly aim for in poetry? There is no checkmate, when talking about the greats there is no point in which one says "this is better than this, 100%, accross the board to everyone." Some like Milton, others like Poe. Sure there are guidelines for good poetry because there are some things that generally work better than others...

You're right - of course, taste isn't absolute. No poem, no matter how produced will be universally accepted. The end towards which we're aiming isn't the perfect poem, it's just a good poem; a competent poem, which doesn't sound like it was written by a machine.

For me the beauty of art lies solely in its effect. The Wizard of Oz has always been a hero of mine. I also believe that poetry really is just words. It's just arranging words on a page: immensely complicated when you analyse it, but not alchemy, not a mystery, not some higher brain-process. I'm not saying it's easy; just that it's possible, just like countless other things once thought impossible.

[pseud]

"royt then. We'll call it a draw..."

[Minstrel]

' The subtleties of nature are greater than the subtleties of argument '

Francis Bacon.

[k-j]

Danny Bacon?

[pseud]

Kevin Bacon?

[k-j]

Bacon Soda. OK enough.

[pseud]

Eggs and...