The Age of Spiritual Machines: When Computers Exceed Human Intelligence
Page 22
There has been substantial concern with Y2K (Year 2000 Problem), that at least some computer processes will be disrupted as we approach the year 2000. Y2K primarily concerns software developed one or more decades ago in which date fields used only two digits, which will cause these programs to behave erratically when the year becomes “00.” I am more sanguine than some about this particular issue. Y2K is causing the urgent rewriting of old business programs that needed to be dusted off and redesigned anyway. There will be some disruptions (and a lot of litigation), but in my view Y2K is unlikely to cause the massive economic problems that are feared.1
In less than forty years, we have gone from manual methods of controlling our lives and civilization to becoming totally dependent on the continued operation of our computers. Many people are comforted by the fact that we still have our hand on the “plug,” that we can turn our computers off if they get too uppity. In actuality, it’s the computers that have their figurative hands on our plug. (Give them a couple more decades and their hands won’t be so figurative.)
There is little concern about this today—computers circa 1999 are dependable, docile, and dumb. The dependability (albeit not perfect) is likely to remain. The dumbness will not. It will be the humans, at least the nonupdated ones, who will seem dumb several decades from now. The docility will not remain, either.
For a rapidly increasing array of specific tasks, the intelligence of contemporary computers appears impressive, even formidable, but machines today remain narrow-minded and brittle. In contrast, we humans have softer landings when we wander outside our own narrow areas of expertise. Unlike Deep Blue, Gary Kasparov is not incompetent in matters outside of chess.
Computers are rapidly moving into increasingly diverse realms. I could fill a dozen books with examples of the intellectual prowess of computers circa end of the twentieth century, but I have only a contract for one, so let’s take a look at a few artful examples.
THE CREATIVE MACHINE
At a time like ours, in which mechanica! skill has attained unsuspected perfection, the most famous works may be heard as easily as one may drink a glass of beer, and it only costs ten centimes, like the automatic weighing machines. Should we not fear this domestication of sound, this magic that anyone can bring from a disk at will? Will it not bring to waste the mysterious force of an art which one might have thought indestructible?
—Claude Debussy
Collaboration with machines! What is the difference between manipulation of the machine and collaboration with it? ... Suddenly, a window would open into a vast field of possibilities; the time limits would vanish, and the machines would seem to become humanized components of the interactive network now consisting of oneself and the machine still obedient but full of suggestions to the master controls of the imagination.
—Vladimir Ussachevsky
Somebody was saying to Picasso that he ought to make pictures of things the way they are─objective pictures. He mumbled he wasn’t quite sure what that would be. The person who was bullying him produced a photograph of his wife from his wallet and said, “There, you see, that is a picture of how she really is. Picasso looked at it and said, ”She is rather small, isn’t she? And flat?”
—Gregory Bateson
The age of the cybernetic artist has begun, although it is at an early stage. As with human artists, you never know what these creative systems are going to do next. To date, however, none of them has cut off an ear or run naked through the streets. They don’t yet have bodies to demonstrate that sort of creativity
The strength of these systems is reflected by an often startling originality in a turn of a phrase, shape, or musical line. Their weakness has to do, again, with context, or the lack thereof. Since these creative computers are deficient in the real-world experience of their human counterparts, they often lose their train of thought and ramble off into incoherence. Perhaps the most successful in terms of maintaining thematic consistency throughout a work of art is Harold Cohen’s robotic painter named Aaron, which I discuss below. The primary reason Aaron is so successful is the thoroughness of its extensive knowledge base, which Cohen has been building, rule by rule, for three decades.
Jamming with Your Computer
The frequent originality of these systems makes them great collaborators with human artists, and in this manner, computers have already had a transforming effect on the arts. This trend is furthest along in the musical arts. Music has always used the most advanced technologies available; the cabinet-making crafts of the eighteenth century; the metalworking industries of the nineteenth century; and the analog electronics of the 1960s. Today, virtually all commercial music—recordings, movie and television soundtracks—is created on computer music workstations, which synthesize and process the sounds, record and manipulate the note sequences, generate notation, even automatically generate rhythmic patterns, walking bass lines, and melodic progressions and variations.
Up until recently, instrument-playing technique was inextricably linked to the sounds created. If you wanted violin sounds, you had to play the violin. The playing techniques derived from the physical requirements of creating the sounds. Now that link has been broken. If you like flute-playing technique, or just happen to have learned it, you can now use an electronic wind controller that plays just like an acoustic flute yet creates the sounds not only of a variety of flutes, but also of virtually any other instrument, acoustic or electronic. There are now controllers that emulate the playing technique of most popular acoustic instruments, including piano, violin, guitar, drums; and a variety of wind instruments. Since we are no longer limited by the physics of creating sounds acoustically, a new generation of controllers is emerging that bears no resemblance to any conventional acoustic instruments, but instead attempts to optimize the human factors of creating music with our fingers, arms, feet, mouth, and head. All sounds can now be played polyphonically and can be layered (played simultaneously) and sequenced with one another. Also, it is no longer necessary to play music in real time—music can be performed at one speed and played back at another, without changing the pitch or other characteristics of the notes. All sorts of age-old limitations have been overcome, allowing a teenager in her bedroom to sound like a symphony orchestra or rock band.
A Musical Turing Test
In 1997, Steve Larson, a University of Oregon music professor, arranged a musical variation of the Turing Test by having an audience attempt to determine which of three pieces of music had been written by a computer and which one of the three had been written two centuries ago by a human named Johann Sebastian Bach. Larson was only slightly insulted when the audience voted that his own piece was the computer composition, but he felt somewhat vindicated when the audience selected the piece written by a computer program named EMI (Experiments in Musical Intelligence) to be the authentic Bach composition. Douglas Hofstadter, a longtime observer of (and contributor to) the progression of machine intelligence, calls EMI, created by the composer David Cope, “the most thought-provoking project in artificial intelligence that I have ever come across.”2
Perhaps even more successful is a program called Improvisor, written by Paul Hodgson, a British jazz saxophone player. Improvisor can emulate styles ranging from Bach to jazz greats Louis Armstrong and Charlie Parker. The program has attracted its own following. Hodgson himself says, “If I was new in town and heard someone playing like Improvisor, I’d be happy to join in.”3
The weakness of today’s computerized composition is, again, a weakness of context. “If I turn on three seconds of EMI and ask myself, ‘What was that?’ I would say Bach,” says Hofstadter. Longer passages are not always so successful. Often “it’s like listening to random lines from a Keats sonnet. You wonder what was happening to Keats that day Was he completely drunk?”
The Literary Machine
Here’s a question for you: What kind of murderer has fiber?
The answer: A cereal killer.
I hasten to admit that I did not
make up this pun myself. It was written by a computer program called JAPE (Joke Analysis and Production Engine), created by Kim Binsted. JAPE is the state of the art in the automatic writing of bad puns. Unlike EMI, JAPE did not pass a modified Turing Test when it was recently paired with human comedian Steve Martin. The audience preferred Martin.4
The literary arts lag behind the musical arts in the use of technology. This may have to do with the depth and complexity of even routine prose, a quality which Turing recognized when he based his Turing Test on the ability of humans to generate convincing written language. Computers are nonetheless of significant practical benefit to those of us who create written works. Of greatest impact is the simple word processor. Not an artificial technology per se, word processing was derived from the text editors developed during the 1960s at the AI labs at MIT and elsewhere.
This book certainly benefited from the availability of linguistic databases, spell checkers, online dictionaries, not to mention the vast research resources of the World Wide Web. Much of this book was dictated to my personal computer using a continuous speech-recognition program called Voice Xpress Plus from the dictation division of Lernout & Hauspie (formerly Kurzweil Applied Intelligence), which became available in the middle of my writing the book. With regard to automatic grammar and style checkers, I was forced to turn that particular Microsoft Word feature off, as it seemed to dislike most of my sentences. I’ll leave the stylistic criticism of this book to my human readers (at least this time around).
A variety of programs help writers brainstorm. ParaMind, for example, “produces new ideas from your ideas,” according to its own literature.5 Other programs allow writers to track the complex histories, characterizations, and interactions of characters in such extended works of fiction as long novels, series of novels, and television drama series.
Programs that write completely original works are particularly challenging because human readers are keenly aware of the myriad syntactic and semantic requirements for sensible written language. Musicians, cybernetic or otherwise, can get away with a bit more inconsistency than authors.
With that in mind, consider the following:
A Story of Betrayal
Dave Striver loved the university. He loved its ivy-covered clock towers, its ancient and sturdy brick, and its sun-splashed verdant greens and eager youth. He also loved the fact that the university is free of the stark unforgiving trials of the business world—only this isn’t a fact: Academia has its own tests, and some are as merciless as any in the marketplace. A prime example is the dissertation defense: To earn the Ph.D., to become a doctor, one must pass an oral examination on one’s dissertation. This was a test Professor Edward Hart enjoyed giving.
Dave wanted desperately to be a doctor. But he needed the signatures of three people on the first page of his dissertation, the priceless inscriptions which, together, would certify that he had passed his defense. One of the signatures had to come from Professor Hart, and Hart had often said—to others and to himself—that he was honored to help Dave secure his well-earned dream.
Well before the defense, Striver gave Hart a penultimate copy of his thesis. Hart read it and told Dave that it was absolutely first-rate, and that he would gladly sign it at the defense. They even shook hands in Hart’s book-lined office. Dave noticed that Hart’s eyes were bright and trustful, and his bearing paternal.
At the defense, Dave thought that he eloquently summarized chapter 3 of his dissertation. There were two questions, one from Professor Rogers and one from Dr. Meteer; Dave answered both, apparently to everyone’s satisfaction. There were no further objections.
Professor Rogers signed. He slid the tome to Meteer; she too signed, and then slid it in front of Hart. Hart didn’t move.
“Ed?” Rogers said.
Hart still sat motionless. Dave felt slightly dizzy.
“Edward, are you going to sign?”
Later, Hart sat alone in his office, in his big leather chair, saddened by Dave’s failure. He tried to think of ways he could help Dave achieve his dream.
Okay, that’s the end. Admittedly the story kind of peters out, ending with a whimper rather than a bang. Seattle writer and editor Susan Mulcahy called the story “amateurish,” criticizing the author’s grammar and word choice. But Mulcahy was nonetheless surprised and impressed when she learned the author was a computer. The program that wrote the story, named BRUTUS.1, was created by Selmer Bringsjord, Dave Ferucci, and a team of software engineers at Rensselaer Polytechnic Institute. Apparently, BRUTUS. 1 is an expert on betrayal, a concept that Bringsjord and Ferucci spent eight years painstakingly teaching the computer. The researchers acknowledge that their program needs to learn about other issues besides betrayal. “Interestingness really combines all of the emotions,” say Bringsjord and Ferucci, and that is something the cybernetic authors are not yet capable of achieving.6
The Cybernetic Poet
Another example of a computerized author is a computer program that I designed called Ray Kurzweil’s Cybernetic Poet (RKCP). RKCP is a computer-generated poetry system, which uses language-modeling techniques to automatically generate completely original poetry based on poems that it has “read.”7
RKCP reads a selection of poems by a particular author or authors (preferably an extensive selection) and then creates a “language model” of that author’s work based on markov models, a mathematical cousin of neural nets. RKCP can then write original poems from that model. As I discussed earlier, RKCP uses a recursive poetry-generation algorithm to achieve the language style, rhythm patterns, and poem structure of the original authors whose poems were analyzed. There are also algorithms to maintain thematic consistency through the poem. The poems are in a similar style to the author(s) originally analyzed but are completely original new poetry. The system even has rules to discourage itself from plagiarizing.
You can download a free copy of RKCP at www.kurzweiltech.com.
Following is a small sampling of poems written by Ray Kurzweil’s Cybernetic Poet after it had “read” poems by several famous poets, as well as lesser-known contemporary poets.
Poems by Ray Kurzweil’s Cybernetic Poet
PAGE
A haiku written by Ray Kurzweil’s Cybernetic Poet after reading poems by Wendy Dennis
Sashay down the page
through the lioness
nestled in my soul.
IMAGINE NOW AND SING
A poem written by Ray Kurzweil’s Cybernetic Poet after reading poems by Wendy Dennis and Ray Kurzweil and love poems by various authors
Imagine now and sing,
creating myths
forming jewels from the falling snow.
SOUL
A haiku written by Ray Kurzweil’s Cybernetic Poet after reading poems by John Keats and Wendy Dennis
You broke my soul
the juice of eternity,
the spirit of my lips.
I THINK I’LL CRASH
A poem written by Ray Kurzweil’s Cybernetic Poet after reading poems by Ray Kurzweil, Robert Frost, and Wendy Dennis
I think I’ll crash.
Just for myself with God
peace on a curious sound
for myself in my heart?
And life is weeping
From a bleeding heart
of boughs bending
such paths of them,
of boughs bending
such paths of breeze
knows we’ve been there
THE STIFLING STUFFY
A haiku written by Ray Kurzweil’s Cybernetic Poet after reading poems by Patricia Camarena Rose and Wendy Dennis
The stifling stuffy
Catholic schoolroom,
where I cannot be real.
WONDERED
A poem written by Ray Kurzweil’s Cybernetic Poet after reading poems by Dave Gitomer
today i wondered
if i mused
today i saw you
i learned
in aw
e and you
if i wondered
if i mused
today i had one wish
if i saw you
if i saw you
if had one wish
MOON CHILD
A haiku written by Ray Kurzweil’s Cybernetic Poet after reading poems by Kathleen Frances Wheeler
Crazy moon child
Hide from your coffin
To spite your doom.
ANGEL
A poem written by Ray Kurzweil’s Cybernetic Poet after reading poems by Patricia Camarena Rose and Sue Klapes Kelly
Beams of the dawn at the angel
with a calm, silent sea
with a hundred times we write,
with a chance we can open up
a steady rhythm in his face
silent room