Darwinian Poetry: From Bad to Verse
For those who say design cannot take place through the process of selection, behold:
Darwinian Poetry.
Cull the prosaic or nonsensical snippets of text, reinforce the rest, and, slowly... genius? Guess we'll find out. Yes, the poems actually
have sex.
Darwinian Poetry
Welcome to Darwinian Poetry! The goal of this project is to see if non-negotiated collaboration can evolve interesting poetry using (un)natural selection.
Huh?
Ok, here's the idea: starting with a whole bunch (specifically 1,000) randomly generated groups of words (our "poems"), we are going to subject them to a form of natural selection, killing off the "bad" ones and breeding the "good" ones with each other. If enough generations go by, and if the gene pool is rich enough, we should eventually start to see interesting poems emerge.
The cool part is that YOU are the arbiter of what constitutes "good" and "bad" poetry. Once you start, you will be presented with two poems. In all likelihood they will both be abysmal pieces of nonsensical garbage. That's ok. All you have to do is read them both and pick the one you find more appealing, for whatever reason. Your decision might be based on a single word that you happen to like. It doesn't matter. Just pick whichever one strikes your fancy.
Once you choose a poem, your vote will be recorded and two more poems will appear. Keep doing this for as long as you like, and definitely come back frequently.
Over time the poems picked by you, and I hope by thousands of other people, will interbreed and more and more interesting poems will emerge. It could take a while. Weeks...months...I don't know. It all depends on how many people participate, and how often.
Keep coming back, for (I hope) the population will evolve steadily, so each day could bring increasingly interesting poems.
That's it. Just click on the "Get Started" link below to dive right in. Or click the "Get Report" link to find out what the current highest rated poem is, as well as to see other statistics.
THE HOW IT WORKS PAGE:
How it Works
"Many poems were butchered in the making of this site."
The Darwinian Poetry software relies primarily on a mechanism called "crossover", similar to the process that operates on chromosomes in biological evolution, except that here the basic genetic units are words rather than nucleic acids. When the program sees that there is room in the population for new poems (because some unfit poems were...um...culled from the herd) it randomly chooses two surviving poems to serve as parents. These two poems are then crossed over, producing two new offspring.
Here is an example to illustrate. These are two poems that I just grabbed off a test version of the site (color coded for convenience):
forest storefront semifinished decrees confirmed
scheming he congestive curdles refulgent
sceptered not of miffs syncretism
lose the but longer floor
the of but judgeship the
forty troweling him sufficing lysolecithin
of from when esurience they
rest timely wounded the perpend
If these two poems were chosen for breeding, the first thing the program would do is decide how many "snip" points to use. Currently this number ranges between one and five. Let's say 2 came up randomly. Now each poem gets randomly cut in two places. Note that this is different than biological crossover in that the cut points vary between the parents. Whereas real chromosomes need to maintain a constant length, our poems will evolve in length as well as content.
forest storefront SNIP! semifinished decrees confirmed
scheming he congestive curdles refulgent
sceptered not SNIP! of miffs syncretism
lose the but longer floor
the of but judgeship the
forty troweling him SNIP! sufficing lysolecithin
of from when esurience SNIP! they
rest timely wounded the perpend
Now the software performs the crossover operation resulting in two new poems:
forest storefront suffcing lysolecithin
of from when esurience of miffs syncretism
lose the but longer floor
the of but judgeship the
forty troweling him semifinished decrees confirmed
scheming he congestive curdles refulgent
sceptered not they
rest timely wounded the perpend
That's
--David
Darwinian Poetry
Code as Art:Poetry:Darwinian Poetry
Generations (Avg)4.502 Total Number of Poems6969 Top Ranked Poems#2496 where ghost sleuth with lingo
of the long with helicopt bodies
where eyes tore devilish covered
#4951 your victims come
#1486 when sometime the life loved
to be throne revoking of shield in blood
#4722 secretiveness her to sins even
send it and woe to
#4808 though eyes closed
helmets stood not
#2216 gone to signify when terribly
untidily of whom suffered him
the come befool in kissing onside he mere
fluidic of her fleetest yet
#861 of either to forgo conclusions
seen reordered hosts my to
tend of me his footprints
but infest lost people lies
#3054 when sometime the life loved
to be throne revoking of shield in blood
#4257 went here bisecting splendid lists
blood of quieting tressed
prince held by posers blood
slumber secretion drink in scene
#4578 into the the lightning divide which
bolstering through stricter
lies
Total Hits76146
Actually, I was trying to be Insightful, not Funny.
Or not at all.
Me lost me cookie at the disco.
Except this IS made by Humans. We're doing the selecting, ergo we're making it, just as we made Dogs.
So of course it's not the same as "Natural Selection," but it's still pretty evolution-like.
And at any rate, it is fastly approaching coherence, and I think it IS art. Before it got slashdotted, I saw grammar, and stuff sort of made coherent sense, in a moody surreal sort of way.
"Never attribute to malice that which can be adequately explained by stupidity." -- Hanlon's Razor
While cells typically reproduce by splitting in half and copying the contents to each half, this is a sloppy method for making complex organisms, and doesn't produce the genetic variety one might like.
Sexual Reproduction is the proccess by which a cell splits WITHOUT copying its contents. The half cell then meets another half cell, which fuse to form a new cell with a brand new genetic code which contains a 50-50 mix of each former cell.
This method is used to perfection in multicellular organisms. Each half-cell is prepared especially to do its job, and when they unite, the cell begins to rapidly divide and reproduce using the traditional Asexual Reproduction method, but while still remaining in contact with its "brother cells."
Slight chemichal variations causes each cell to utilize its genetic code differently, and the cells become specialized. The cells eventually form a gigantic single organism formed of millions and millions of cells working together in tandem.
"Sex" is the proccess by which these organisms send the "half-DNA cells" (or gametes) to each other. In humans, (to pick a species at random) the proccess involves two humans of opposite "gender" (each gender produces a different form of gamete, and the opposite kinds have to come together for the proccess to work) line up their bodies. After the male's body is stimulated, the male's gamete enters the female's body, where the cell, equipped with a flagella, approaches the female's gamete. After the cells unite, the newly formed organism grows in the female for approximately 24 Megaseconds (9 "months"), after which the organism exits the female's body.
Typically, for approximately 560 Megaseconds (18 "years"), the organism is often under the gaurdianship of the parent organisms.
"Never attribute to malice that which can be adequately explained by stupidity." -- Hanlon's Razor
You can find examples of Cope/EMI compositions in MIDI and PDF format here.
Cope has written extensively about EMI in Computers and Musical Style and Experiments in Musical Intelligence. In the second volume, he includes a "mini" version of EMI called Sara, which is written in LISP and will run on a Mac. You can also find the source here.
Sara works by reassembling works of a composer to form new works. Basically, Cope "distills" a composer's work by simplifying the texture, changing the key, and doing other things that make it more amenable to recomposition.
Then the work is fed into Sara, which analyzes each bar based on harmonic function, melodic function, and so on. These analyzed chunks are then stored in a database.
To "recompose" a work, Sara picks a composition to use as the base. It then replaces each bar in the composition with a functionally equivalent bar from the database, based on harmonic function, melodic direction, and so on. The result is a composition which follows the same general contours of the original work, but has a different melody, texture, and often changed harmony - yet still follows the same stylistic rules of the composer.
EMI is significantly more powerful than Sara. At it's core is a rules-based composition engine, which can generate proper - and perhaps a bit bland - compositions in many styles following music compositional rules. For example, it can even generate a 'proper' four part fugue. EMI's pattern matcher is more sophisticated than Sara's, and EMI is much more subtle in how it weaves a composer's work into it's own. It's even difficult for Cope to tell where the material comes from.
EMI was written primarily to help Cope through a writer's block, and in The Algorithmic Composer he details Alice (ALgorithmically Integrated Composing Environment), yet another incarnation of EMI, which functions as a composer's assistance (included with the text).
Cope is an excellent author, and he makes much of his work understandable to people without a degree in Music Composition or Artificial Intelligence. He's is quite willing to acknowledge and discuss the shortcomings of his programs. In a field where some people consider using fractals as "composition" because the results resemble music, Cope has managed to create something that not only "resembles" music - it's fooled a lot of experts, too.
That's quite a feat.
Initially, the snippets remind me of unedited "l=a=n=g=u=a=g=e" poetry from the late '80s, but I suspect they'll be verging towards formal and stylistic standards like R.Frost or ee cummings, since that's what people got in school (and usually remember). I don't have faith that this will wind up with anything like the avant-garde direction that the newness of the generation technique suggests is possible.
There's a good tradition of last century's poets experimenting with generation techniques. Bryan Gyson and William Burroughs played with cutups, and someone's even automated the process with TextBlender Pro (disclaimer: haven't tried this one). I had a gas with this idea, and once had a month off so sequestered myself with a typewriter (yeah I'm getting old) and source texts by Buckminster Fuller, Nietzche, Attar, and some histories of WW2, in order to generate some centos for fun and non-profit (never published, needless to say).
William Carlos Williams claimed that poetry is a word machine:
- To make two bold statements: There's nothing sentimental about a machine, and: A poem is a small (or large) machine made out of words. When I say there's nothing sentimental about a poem, I mean that there can be no part that is redundant.
Anyway, the Darwinian P. reports indicate that the process has a long way to go. So what will literary critics (before their descent into hell) claim about the validity and category of these poems? Is it just one more disintegration of the canon that comes with the post-post-modern post mortem? Will the poems stand the test of seven layers of meaning? O machine, wax!Prose may carry a load of ill-defined matter like a ship. But poetry is a machine which drives it, pruned to a perfect economy. As in all machines, its movement is intrinsic, undulant, a physical more than a literary character. From: Williams's introduction to The Wedge, in Selected Essays of William Carlos Williams
Damn those pesky terrorists