HOME OF THE CREATIVITY MACHINE
The Big Bang of Machine Intelligence!
The IEI Blog is now open for discussion of new business ventures and opportunities!
- For many decades efforts have been underway to write computer programs
that can generate original musical themes. Creativity
Machine Paradigm now allows computers to compose both novel and appealing
melodic concepts without any help from human intelligence.
Since musical tastes vary, these systems automatically tune their
acoustic output to the perceptual responses of an honest audience.
Details - For decades, IEI has been using its patented Creativity Machine architectures to generate fresh and appealing musical compositions. For the most part, such projects have served largely to demonstrate the vast flexibility and applicability of this new AI paradigm. It has also served as a social experiment, enabling us to better understand the psychological and sociological impact of an autonomous invention machine invading the turf, so to speak, of an established trade or craft.
Key to our observations is the following: If Creativity Machine generated music is played to an audience that doesn't have a clue as to melody's origin in silicon, the reaction is everything from "beautiful" to "intriguing" to "who composed that? It's great!" Alternately, if such a song is played with the caveat that it is purely computer generated, the reaction, especially from musicians, is that it does not manifest the emotion, feeling, or warmth of a real human being. Overall, the reaction has more to do with vested interests and pride rather than the intrinsic appeal of the melodic piece. Instead of appreciation and wonderment that a machine can generate such rhapsodic beauty from a collection of dumb switches (i.e., neurons) and interconnects, the response is typically that of intimidation.
Within IEI, music generation served as an important reduction to practice within our Creativity Machine patents, the first of which taught the use of all manner of feed forward and recurrent networks to generate candidate music, as critic networks selected the very best of the emergent works. The formula for generating this art was simple and elegant: One or more networks absorbed the "zen" (i.e., the myriad mathematical constraint relations) of what constituted palatable music. When such nets were seeded with random numbers to their inputs or internal architecture they would generate plausible yet novel musical sequences. One or more observing nets, trained by example, could then isolate and play the more promising of these machine-imagined tunes. In other words, the perturbed net, or imagitron, could automatically confine its musical search to those novel patterns it deemed potentially acceptable music. The other neural assemblies implicitly came to understand the pattern behind musical preferences. Therefore, if the latter were trained upon an individual's musical tastes, what emerged from the machine was, by definition, acceptable to them. If such critics were trained upon consensus, the resulting art was certain to appeal to the masses, unless of course the above prejudice was called into play.
The latest generation of Creativity Machines does not rely upon such pre-training. Instead, the system experiments with its environment (the listener) through any manner of sensor, whether it is a keyboard or advanced machine vision system, carrying out a series of experiments with the intended consumer and cumulatively bootstrapping the perceived quality of the music. In our opinions, the real value of this system is not the music itself, but the important lesson it provides, namely that Creativity Machine Paradigm is the methodology that can bootstrap a solution to anything, no matter how complex, whether it be an intricate hexapod robot developing mobility strategies, or solving multibillion dimensional economic or societal problems on humble computational platforms. Understanding that potential, the music generated through these fundamental neural architectures is even more powerful in its impact.
From 1975 to 1996, Creativity Machine generated music was largely a laboratory curiosity until the Dutch documentary program Noorderlicht showed interest in producing a show featuring this bleeding edge technology. To produce background music for this television show, a Creativity Machine went to work developing a score that included rather ethereal music emanating from the system as its neurons were randomly destroyed. In other words, it originated its own near-death experience music.
In 2006, Imagination Engines self-produced its documentary "In Its Image," the musical score composed largely by Creativity Machine Paradigm. Later that year, the newest generation of Creativity Machine, was able to bootstrap an entire musical album, called "Song of the Neurons," using only feedback from an IEI machine vision system that was able to gauge satisfaction of candidate melodies purely from the expression on a human listener's face. Among the myriad pioneering achievements of this project was the truly amazing fact that the resulting music was born from a collection of neurons bent upon pleasing a human participant. No other humans were in the loop writing code or performing statistical studies to decide what the most appropriate note was to follow another. It all happened rather spontaneously!
© 1997-2017, Imagination Engines, Inc. | Creativity Machine®, Imagination Engines®, Imagitron®, and DataBots® are registered trademarks of Imagination Engines, Inc.
1550 Wall Street, Ste. 300, St. Charles, MO 63303 • (636) 724-9000