In 1984, Lejaren Hiller, Slee Professor of Music at the State University of New York at Buffalo, was invited by the State Department to prepare a demonstration applying artificial intelligence to musical composition for the US pavilion at the 1985 International Exposition in Tsukuba, Japan. That the invitation was personally directed to Hiller struck me as odd, because although Hiller's composing programs had AI goals, Hiller himself had never used AI techniques. Indeed he hadn't even progressed to basic programming methods such as recursion or linked data structures.
I fancied at the time that the invitation had been prompted not just by Hiller's composing programs, but by the activities of people surrounding him. I had just then completed my degree under Hiller, but had already published six articles describing pieces of music generated using composing programs — five of which projects had employed actual AI techniques. The celebrated mathematician John Myhill, Hiller's close personal friend, had developed compositional processes using a system of “controlled indeterminacy”. Significant work was also being pursued by Kemal Ebcioğlu, to whom Myhill was acting as thesis adviser. Ebcioğlu had developed true AI applications in 1980 for species counterpoint and before 1986 for harmonizing chorales in the style of J.S. Bach.
So I fancied that I myself had been among the considerations drawing the invitation to Hiller. With hindsight I now believe differently. I think they went to Marvin Minsky for advice and Minsky suggested Hiller. Another factor was that even before they invited us, the pavilion commissioners had obtained a Kurzweil 250 synthesizer and needed some way of working it into the exhibit.
In any case when Hiller asked me to join the team, I faced a dilemma. I could hardly turn down the opportunity, but Hiller had seemed lax in the past about giving due credit, and I wanted nothing to do with his FORTRAN II code. My solution was to take full control of the programming — not just the coding but the structures and methods as well. Hiller's role was to be limited to that of composer. Hiller agreed to this at the outset. It turned out that I was right to protect myself in this manner but not due to any failings on Hiller's part. Hiller in fact acknowledged my contributions generously. Yet it is not in the nature of reporters and the media to take care about who really does what.
We faced a great deal of pressure getting things together. It did not help that Hiller began showing symptoms of what later proved to be Alzheimer's. These initially manifested as a tendency to enlist everyone he happened to meet into the team. Fortunately we were blessed when Heinrich Martens, director of computing services, recommended Bob Franki. Bob took the presentation graphics for himself and found capable people to handle the performance subsystem. Hiller's enlistees were given the opportunity to prove themselves, but several who did not contribute positively had to be shown the door. Bob also took the initiative in contacting vendors, successfully acquiring a small mixing panel, a power amplifier, a set of JBL speakers, a Sequential Circuits Drum Traks, and an Oberheim Xpander
A general description of the Expo '85 installation, including photographs of the equipment and images of Bob Franki's presentation graphics, was published in a 1985 Perspectives of New Music article which, for the record, was written by me. A paper presented at the 1985 International Computer Music Conference provides implementation details for the four composing programs. The remarks that follow supplement these two primary sources.
Results generated by the Expo '85 project were released in Hiller's 1986 Computer Music Retrospective, which was produced under the Wergo label. This recording is long out of print, but I managed to locate mp3's of the Expo '85 material at mp3hampster.net. For the record, the Wergo recording acknowledges who composed what fairly diligently, although Hiller's liner notes over-generously assigns sole credit for Mix or Match to me. (mp3hamster, by contrast, credits everything to Hiller.)
Hiller had two ideas. The first came to be known as Circus Piece, which consisted of a list of written-out cadences which the program was to piece together randomly. This obviously had nothing to do with artificial intelligence but it was not my role to question. I dutifully coded the program, substituting statistical feedback for random selection.
Circus Piece by Lejaren Hiller. Realized using the Kurzweil 250 synthesizer exclusively.
Hiller's second idea was a program to compose tunes in “Tin Pan Alley” style. This project, named Mix or Match, ultimately became the most suitable of all for an AI-in-music demonstration. However it did not start out that way. After several weeks of “rumination”, Hiller finally handed me a sheet of staff paper containing a set of rhythmic patterns. The remaining questions: how to arrange patterns into a form, how to choose pitches, and so forth, were left to me. For these creative contributions I demanded, and was granted, co-composership. My composing program generated pieces in several passes. The first pass decided which sections should contain original material, and which sections should imitate previous sections; it also determined the chord progression. The second pass generated the rhythm, selecting not simply durations but also note functions: cadence tone, chord tone, ornamental tone. The final pass selected pitches. Bob Franki's presentation graphics showed the computer actively backtracking as it considered alternative solutions.
Mix or Match by Lejaren Hiller and Charles Ames. The tunes presented were the ones Bob Franki selected to animate with note-selection graphics. Realized using both the Kurzweil 250 synthesizer and the Sequential Circuits Drum Traks.
The third project was my own Transitions, which explores 31-tone tuning. Transitions employed AI techniques to discover a scale whose intervals emphasized ratios involving factors of 7. It draws on principles from Gestalt psychology to generate a top-down hierarchy of origins and goals.
Transitions by Charles Ames. Realized using the Oberheim Xpander.
The fourth and final project, Toy Harmonium was conceived by John Myhill at my behest. Already overstretched by three other composing programs, I recruited Leonard Manzara, a fellow composition student, for the job. The program actualized Myhill's ideas about controlled indeterminacy, applying the principle of gradual interpolation not just to rhythm, but also to pitch.
Toy Harmonium by John Myhill. Realized using the Kurzweil 250 synthesizer exclusively using just the Grand Piano sound.
Bob Franki, Bob Coggeshall, and I came to Tsukuba about a month in advance of the opening to set up the exhibit on site and work out the final kinks. Exterior and interior photos can be found at www.worldsfairphotos.com, which site also lists the stations visitors would encounter as they progressed through the pavilion. As the exterior photo shows, the pavilion itself was essentially a large tent The interior photo shows the “Systems Demonstration” platform containing our installation.
The theme of the pavilion was US innovations in Artificial Intelligence. Of the stations listed on www.worldsfairphotos.com, I have limited memory of “The Road to Artificial Intelligence”, “Networking”, “Free Flow of Information” or “Reasoning Workshop”. I well remember the US “Nobel Laureates” station — a pavilion deputy commissioner (there were 7) told me that this station was originally supposed to honor key US AI innovators — however, the organizers couldn't agree on whom to feature, so that idea was dropped.
Other installations sharing the “Systems Demonstration” platform included a computer-drawing demonstration by Harold Cohen, a voice-recognition program that did paint-by-numbers (visitors pointed at a region with a mouse and voiced the color they wished filled in), and three Symbolics LISP machines that didn't do much at all — www.worldsfairphotos.com has it wrong, the LISP Symbolics had no role in the other demonstrations described here. My favorite demonstration was a robot arm coupled with a video camera, that assembled drink coasters from components which were randomly arranged in a tray. It was not on the platform, rather over against the wall. When this apparatus was first tested, it attracted several onlookers who watched bewildered as the arm accurately identified components, picked them up, and dropped them on the ground. (The problem was quickly resolved.)
Following the the “Systems Demonstration” area was a small theater where visitors could watch a short feature about a pretty little blond girl who interacted with a talking computer. This computer could also project 3-D graphics into mid air. The perceived message ended up being less about future computers being intelligent and more about future computers completely lacking initiative and therefore being suitable companions for small children in pristine spaces. The Japanese subtitles — reportedly prepared by President Reagan's “personal” translator — were so nearly illiterate that the pavilion staff intentionally positioned them straddling the edge of the screen, so no one could read them.
Other fair highlights included a 14-story Sony Jumbo Tron, the Wasubot Musical Robot, and the Mitsubishi pavilion's robot circus. This last exhibit featured an entire troupe of interacting robots who tossed around spinning metal tops and, in one case, guided a spinning top down the edge of a samurai sword. Mitsubishi totally put the US to shame.
Reviews of the US pavilion were scathing, especially in light of offerings elsewhere at Expo '85; see for example Thomas Marill's review in the AI Magazine.
The opening of Expo '85 was a frenzied event with crowds of dignitaries and reporters. Hiller's attempts to parlay the invitation into a southeast Asian tour having fallen apart, he was not present at the opening. (He did attend events later during the summer, when he met the Emperor.) Without Hiller around, I got filmed demonstrating Mix or Match for a spot on the CBS Evening News. My brother Tom happened to be watching from Vermont when my face came on. “Well, it's not Quincy Jones, but you have to start somewhere!”, the narrator quipped. Tom called my mother, who viewed the spot three hours later in California. These two members of my own family are the only two witnesses, known to me, of my moment of glory.
Among the dignitaries was the Crown Prince's second son, representing his grandfather. Warned by the embassy staff of his approach, I took station on the platform. He came by with his retinue, heard a Mix or Match example, gave me a bored nod, and moved on. I think perhaps he took my standing on the platform as an insult. Suddenly I was surrounded by Korean reporters snapping pictures of me shaking hands with their Minister of Culture.
| © Charles Ames | Page created: 2013-07-28 | Last updated: 2017-03-12 |