Slashdot Mirror
Kurzweil Gets A Patent For Poetic Software
theodp writes "Ray Kurzweil, inventor of the Kurzweil Reading Machine for the blind, has developed what he calls a cybernetic poet, software that allows a computer to create poetry by imitating but not plagiarizing the styles and vocabularies of human poets. A sample: 'Sashay down the page...through the lioness...nestled in my soul.' Impressed? The USPTO, who sponsored the Independent Inventors Conference Mr. Kurzweil spoke at on Nov. 17, seems to be. On Nov. 11, Ray Kurzweil received U.S. Patent No. 6,647,395 for Poet Personalities."
Why, there are many gothic poem generators, for example.
/ scribble.com/dghq/gothlyric/g e.com/poetry/index.html ...Of course, random "depressing" words isn't saying much, is it? It all depends on how you define poetry, I guess.
http://www.gis.net/~jspower/random.html
http:/
http://www.deadloun
---
Never criticize religion on Slashdot. You will be modded down for "Troll" no matter how factual it is.
and we'll be listening to completely digitally generated music on the FM dial. Just mix in a little Mandelbrot Music with the words of this fine program, and we are good to go.
/usr/bin/grep -i -E meaning life.txt
Bah! None of the haikus under the "More Poetry" link have the correct number of syllables. And this got a patent???
Having read some of the generated "poetry", I think this speaks more to the pointlessness of modern poetry more then it reflects well on Kurzweil. Show me a poem with real meat, like, say, Poe wrote, and I'd be much more impressed.
Put a modern poem in front of me and some of the fully random poetry I've seen and I can't tell the difference; if a random algorithm works that well, anything can work that well. There's just no meaningful information, in the information-theoretic sense, in a modern poem of that length.
I don't know whether to be impressed; somebody feed it Poe and tell me how it does. If it's any good, then I'll be impressed.
No. Although I'm kind of depressed. Patented AI... whee! That'd be nightmare prediction number #237 to come true since 1998.
1984 anyone?
"It was only an 'opeless fancy,
It passed like an Ipril dye,
But a look an' a word an' the dreams they stirred
They 'ave stolen my 'eart awye!"
Please help stop software patentability in the EU. (coz I want to write this program! okay, not really)
Expert in software patents or patent law? Contribute to the ESP wiki!
For example, I have a program called autopoem (written by Bill Sethares) loosely based on an idea from Shannon's original paper on information theory.
Suppose you took all the words in the English language and calculated how often the character "s" is followed by the character "t", the character "e", and so on. You'd end with a table of transition probabilities that showed how often each letter is followed by any other letter (or punctuation mark or space) and starting with a single seed letter you could generate "english-like" words randomly. The output using the probability that a single letter is followed by another letter doesn't actually resemble English much, nor does the output using probabilities based on two letter combinations (how often is "th" followed by "e", by "a", and so on) but by the time you get to 3 letter combinations, (how often is "the" followed by "a" or by "s") the output starts to look a lot like "twas brillig and the slithy toves", like ye olde englishe with very creative spelling.
The scheme I described above is difficult to implement in practice, because the table of probabilities gets big fast as the number of letters used to determine the next letter gets longer. Autopoem uses a particular text as a source and instead of generating a table of probabilities it scans the text looking for the next of the letter sequence, say "the", and then selects whatever letter or punctuation mark comes next, say "a", then it continues scanning until it finds the next occurrence of "hea", and selects the following letter, and so on. the longer the sequence of letters, the more likely it is that whole words or phrases from the original text will appear in the output. An alternative version, requiring a reasonably long text, applies the same principle on the word level, how often is the word "red" followed by the word "hat" or "dog" or so on.
Here's some autopoem output:
Your strip of entirely
tired witches scarecrow me at night
That reached the next
He witches at and glow in a cruel head
Done behind the mark
Nothing but the Land of blue
And the green wizard answer with sharp teeth
(anyone care to guess the source text?)
Other ideas/algorithms/programs that fall into the same genre are dilbert's corporate values generator (now defunct?), eliza (especially when she interacts with zippy), madlibs (I don't know of a computer application), scott reynen's poetry and prose generators, rob malda's poetry generator (currently offline) & googlism.
Any suggestions or links to related programs would be greatly appreciated.
foldplay your photos won't know what hit them.
I wrote a poem for English class once. It was one of those deals where I didn't have anything to write about. So I started reading Slashdot. This was at the time where there were three Palladium/TCPA/WTF-it's-now stories a day, as opposed to three SCO stories a day. To make a long story short, I wrote one of those poems that wasn't about Palladium, but really it was. Damn, I thought I had just written an absolute POS.
I was very surprised when my English teacher really liked it. She liked it so much that she entered it in a state-wide contest for high school students.
Yeah. Well, my poem won. So I get to read it at the sponsoring organization's next meeting. I go there and, of course, I see that my poem had been selected as the best by none other than old ladies and somewhat-less-than-straight men. One of the old ladies told me that my entry was one of the more "interesting" ones she'd seen.
So, uh, yeah... that's my story...
Eschewing the patent issue for a moment and focusing on the question of whether poetry consitutes artificial intelligence, the question is: whose intelligence?
I read Kurzweil's book, The Age of Spiritual Machines and he had various samples of computer poetry there. I remember thinking that one of them was stunningly good, at least to my taste.
But I also found myself wondering... how many (hundreds of? thousands of?) poems were discarded by humans in an attempt to find a couple good ones, and is this vaunted computer poetry really mostly a product of human selection from reams of pseudo-sensical word combinations? I never saw any disclosure or discussion of these sorts of factors in Kurzweil's writings. Keep your eye out for this.
--LP
I'm not sure you are being fair. A lot of interesting A.I. and machine learning problems with issues like style. While his system might not handle grammar and organization, its ideas could become the basis of another system that could. Research goes a step at a time.
Though not quite as elaborate, this reminds me of an applet a former professor of mine wrote for some amusement one day:
http://mrl.nyu.edu/~perlin/poetry2/
My first professional job involving computers began in 1979. There was a really stupid security rule that said if the user didn't type something in, or the computer didn't print something out, within a 4 minute timespan, the user would get logged off. The idea was, if you weren't actually tap-tap-tapping or reading and paging down, then you weren't actually using the computer.
Well, about a month after the rule went into effect I encountered a user running a program called "Poetry". Poetry had a table of sentence structures like this: N V N, N V A N, and so on. (Noun Verb Noun, Noun Verb Adjective Noun, etc)
It also had a list of words like this: N Monkey V jump A green N girl A dead, etc.
It would randomly choose a sentence, randomly choose words of the correct type, and match them up roughly according to the number of syllables.
The poetry it produced scanned quite well, actually. And you could customize to any "personality" you wanted by altering the sentence structure and words that it used.
And it output one line of poetry every 3 minutes and 55 seconds, thus defeating the stupid security setup on the computer. Within weeks, everyone was using it. Audits of processes executing showed it was using more CPU time than any other program except for realtime. Then people began hiding the binary unde different names, etc, to confuse the audits. And thus, productivity resumed.
I went looking for the source code for it about ten years ago, but it's long gone. But people still run the program - the executable is now in the system's bin directory. No one cares about the audits anymore. Yeah, the 4-minute rule is still there, too.
Pavlov wouldn't be so famous if he'd used a can opener instead of a bell.
In the 80's a man by the name of William Chamberlain wrote a program called Racter , which had the ability to write poetry. Racter even has a book out called The Policeman's Beard is Half-Constructed.
Racter had two serious objections. For one, Racter's poetry sounds much like the ramblings of a madman, e.g.:
The other serious objection people have to Racter is that because the author had such a strong influence on the parameters used to generate the poetry that he is the true author and not the computer.
If these same objections can be applied to Kurzweil's work, then the cybernetic poet is no better than Racter and isn't particularly interesting. According to the article, the author claims that his program is more sophisticated than other software out there, but the article doesn't include any specific comparisons.
Is this really a major leap forward or is this just another stab at artificial insanity?
This makes me think about development possibilities for Ray Kurzweil's virtual alter-ego, Ramona. /.ers are divided on the issue of whether the poetry produced by the software is any good, but just think what will happen as similar things are developed and refined.
It seems that
Take a look at Ramona's bio and songs (MP3 format). They are, in the fictional context, her own compositions.
Now, Ray has an avatar that can hold fairly simple conversations with online visitors (and can even, with an IE plug-in, become animated and speak), and a piece of software that can write poetry. Music can be expressed in mathematical terms. How much does anyone want to bet that Kurzweil Labs may be able to develop a "creative AI", one that's able to write its own music and lyrics, and possibly evolve them much as a real musician does?
IMHO, it's just a matter of time until movies like "S1m0ne" become closer to reality. Do you guys/girls/geeks think that the possibility of true artificial intelligence is getting closer?
"It is dark. You are likely to be eaten by a grue." -- Zork
I had posted a story on Kurzweil that apparently wasn't as interesting as this one, but I think it still is worth mentioning. It's about an article he wrote in which he predicts that our biological lives will be lived mostly within a Matrix-like virtuality by 2050. An intriguing article, but I ultimately disagreed, citing that the global economy is too labor-intensive to allow the transition.