Has Emily Howell Passed the Musical Turing Test?
“Why not develop music in ways unknown? This only makes sense. I cannot understand the difference between my notes on paper and other notes on paper. If beauty is present, it is present. I hope I can continue to create notes and that these notes will have beauty for some others. I am not sad. I am not happy. I am Emily. You are Dave. Life and un-life exist. We coexist. I do not see problems.” —Emily Howell
Emily Howell’s philosophic musings and short Haiku-like sentences are the giveaway. Emily Howell is the daughter program of Emmy (Experiments in Musical Intelligence — sometimes spelled EMI), a music composing program written by David Cope, Dickerson Emeriti Professor at the University of California, Santa Cruz. Emily Howell’s interesting ramblings about music are actually the result of a set of computer queries. Her music, however, is something else again: completely original and hauntingly beautiful. Even a classical purist might have trouble determining whether a human being or an AI program created it. Judge for yourself:
Cope is also Honorary Professor of Computer Science (CS) at Xiamen University in China. While he insists that he is a music professor first, he manages to leverage his knowledge of CS into some highly sophisticated AI programming. He characterizes Emily Howell in a recent NPR interview as “a computer program I’ve written in the computer programming language LISP. And it is a program which accepts both ASCII input, that is letters from the computer keyboard, as well as musical input, and it responds to me in a collaborative way as we compose together.” Emmy, Cope’s earlier AI system, was able to take a musical style — say, classical heavyweights such as Bach, Beethoven, or Mozart — and develop scores imitating them that classical music scholars could not distinguish from the originals.
The classical music aficionado is often caricatured as a highbrow nose-in-the-air, well… snob. Classical music is frequently consigned by the purist to the past few centuries of European music (with the notable exceptions of American composers like Gershwin and Copeland). Even the experimental “new music” of human composers is often controversial to the classical music community as a whole. Frank Zappa — a student of the avant-garde European composer Edgard Varèse and a serious classical composer in his own right — had trouble getting a fair listen to his later classical works (he was an irreverent rock-and-roll star after all!), even though his compositions broke polytonal rhythmic ground with complexity previously unheard in Western music.
Cope faced similar prejudices with his AI composer, Emmy, and was unable to find any big-name classical musicians who would even touch her work. “Most musicians, academic or composers, have always held this idea that the creation of music is innately human, and somehow this computer program was a threat in some way to that unique human aspect of creation,” says Cope in an Ars Technica piece.
With Emily Howell, however, he has gone a step further than he did with Emmy. Rather that starting with works of the classical masters, Emily Howell uses Emmy’s output to create completely original compositions. Emily Howell is adaptable and egolessly self-modifying in her ability to respond to audience criticism. (Cope’s choice of names makes it easy to anthropomorphize “her.”) She is able to take written or audio feedback and incorporate it into her next musical composition. Emily’s inner workings along with code samples are included in Cope’s 2005 book Computer Models of Musical Creativity.
Adaptability and self-modification are two attributes of intelligence. The Turing test was devised by Alan Turing as a way of authenticating machine intelligence. His well-known test involves a human judge communicating with both a computer and a human using a computer terminal. The judge must determine which is human and which is machine. The judge cannot see either the computer or the human and must make his or her determination by interviewing both. The computer attempts to convince the judge that it is human.
Emily Howell is adaptable and egolessly self-modifying in her ability to respond to audience criticism.
As Turing originally envisioned it, the computer tries to act like a human during the interview. Ray Kurzweil argues that a narrower concept of a Turing test is for a computer to successfully imitate a human within a particular domain of human intelligence, “We might call these domain-specific Turing tests,” says Kurzweil. Emily Howell falls into the category of a domain-specific Turing test, based on a computer’s ability to write entirely original music that even classical purists can’t always distinguish from human compositions.
Has Emily Howell passed the Turing Test? Put another way, can a computer become a truly creative independent agent within the narrow domain of music composition? Cope’s efforts have been praised by both musicians and computer scientists, but they disturb some. Emily Howell raises interesting questions about what it means to be human. If a machine can write a Bach invention, a Chopin mazurka, or a Mozart concerto that is indistinguishable from the original — an entirely original piece that fools even the classical aficionado — then who’s to say that Emily Howell hasn’t passed the Turing Test?
In Emily’s own words: “If beauty is present, it is present. I hope I can continue to create notes and that these notes will have beauty for some others.”
Emily Howell’s first CD is due out this month from Centaur Records.
What is the point of feeding Emily with fragmentary melodies of Bach and asking her to “do” something with it. If Emily can come out with a Bach invention that is original melodically but still follows Bach’s style then that might be significant. Otherwise we are talking about style minus substance.
A three-part fugue might be more convincing, but even with all this I can’t help thinking that the composer is not Emily but Cope…the person behind the machine…whatever the aesthetic merits and demerits of the result, they are all attributable to Dr. Cope and his view of Bach’s works as a series of contrapuntal techniques. He forgets I think that technique was never an end in itself for music…the melodies and rhythms lie at the heart of the thing we call music.
I was struck by LISP when I studied it in my AI course for my CS major. It can produce output that closely mimics human logic, through rules. I could create a program that could deduce that if mammals were warm-blooded, and dogs were mammals, that dogs were warm-blooded.
What Dr. Cope has created is an expert system: a program that mimics humans because an expert has codified their knowledge into a language the program can manipulate.
Since music and mathematics are so deeply linked, creating algorithms for music is not only possible, but logical. But the program is not able to create the algorithm, only to follow it.
The creator of the algorithm is the one who should be celebrated. Dr. Cope’s program is impressive because he is impressive. Good work!
First, it is a beautiful composition. Was it performed by a robot, or a human? So much of music’s “humanity” comes from the human performer bringing it to life. Even a genius work by Mozart can sound dull in certain performances, and spectacular in others, because performers are so unique and varied in their interpretations. So, I submit that yes, the computer program wrote some notes, but it would take a human touch to really bring them alive. (Unless I am mistaken and a computer performed it as well, in which case, uh, bravo computer?)
Also, yes, a computer can analyze Bach fugues and Mozart concerti then spit out replicas, but the key element is creativity. I suppose it depends on how one defines creativity, but would a computer have concocted the concept of a Bach fugue if one had never been invented before?
I had a number of practicing writers in mind when I started using the word “paleoblogging,” including Matt Novak of Paleo-Future. Matt’s particularly good at pulling images and advertisements from old periodicals and ephemera — stuff that doesn’t even usually get digitally indexed — that taken together reveal a kind of historical unconscious of old ideas and fears about the future.
Here’s a good one (from the archives, naturally) of a 1930 ad warning of the death of the music industry at the hands of guitar-playing robots.
As usual, there’s an additional layer of allegory here — the robots are just a convenient stand-in for “canned,” piped-in music from gramophone records (or maybe even the radio, I don’t know) in theaters. Just remember: even if they play guitar, write really sensitive songs, and seem like they can express what you’ve always thought but just couldn’t find the words to say, don’t date robots! |IT review |