Recent innovations in AI language technology have captured the public’s imagination. A key concept in building Artificial Intelligence is the model. A model is an algorithm or layers of algorithms that interpret data and make decisions based on that data. Generative models can generate novel data instances or possibilities. Discriminative models discriminate between different kinds of data instances. Used together, generative and discriminative models can result in both novel and valid (rule-following) possibilities. This seems very similar to human imagination and creativity.
Listening to a podcast of Rick Rubin interviewing Brian Eno, generative AI wasn’t expressly mentioned, but I couldn’t help but reflect on how it is a familiar conceptual territory in Eno’s body of work over the past half century, which I would describe broadly as a generative systems-based approach to art and creative innovation.
It suddenly struck me that artists should periodically reevaluate the limitations of their existing models and employ techniques to overcome them. In psychology this has been called lateral thinking.
In an applied sense, many creative endeavors and collaborations are variously derivative and involve models (other artists’ works) that serve as a limited range of stylistic parameters (rules) to satisfy the assumed expectations of an audience. I’ve always found this practical but frustrating, especially in collaborative endeavors. For example, “hey, let’s get together make a polka record that sounds like Frankie Yankovic!” Or perhaps a more creative synthesis, “hey, can you play that guitar part more like a cross between Bach and Lee Ranaldo?” Many musical collaborations begin with a strict set of implied and even unspoken rules and models.
Looking to Webster’s dictionary, I found three definitions of “model” that I found useful in this sense:
- a pattern of something to be made
- a system of postulates, data, and inferences presented as a mathematical description of an entity or state of affairs
- an example for imitation or emulation
“Archetype” and “schema” may be similarly useful.
These “creative models” probably make better sense in the context of human culture and personal identity. Put simply, people tend to prefer music with specific characteristics (a style or genre). These populations are a target market demographic that often correspond to a subculture that the members consider a declaration of their personal identity. Artists seek to create a “product” that satisfies their needs by using existing successful products as models, in hopes of attracting that market. However, authenticity and originality are important to subcultures. Blatantly copying an existing product or persona can and has backfired. The artist feels they may be walking a fine line between a product that is a novel and pleasantly surprising take on the familiar and one that is unsettling, unorthodox, or challenging to the cultural sensibilities of the listener.
Throughout the twentieth century, stylistic boundaries shifted and expanded, especially during periods of high social anomie in the 1960s & 70s, during the golden age of the album-oriented music market driven by a large baby boom youth cohort. In the twenty-first century, some argue that the algorithms of streaming services have caused feedback-loops resulting in a return to bell-shaped distribution where more songs sound the same, others suggest that tastes have become more diverse and personalized to geographic region, culture and even subculture. The data suggests that while the universe of possible genres has become larger, people have become more ensconced in their sub-groupings and subcultures. Likely a byproduct of recommendation algorithms, this is similar to the polarizing patterns we see in political ideology. Middle-of-the-road pop has existed since the dawn of mass media, but now that’s increasingly supplanted by the “average interests of others like you.” This trend is non-innovative and predictable by design. It is comfort food for the ears.
For decades neuroscientists have pondered whether musical taste is hardwired into our brains or develops as a consequence of acculturation. A study from MIT and Brandeis University suggests that it’s predominantly cultural. Fascinating examples of this abound now on YouTube “reaction” videos where everyday people react to music from other cultures. One of my personal favorites features Pakistani tribes-people reacting to various western cultural artifacts including the video for Metallica’s One. This reminds us how relevant cultural relativism is to our attempts to minimize bias when evaluating distinct cultures both global and local.
A person’s affinity to a particular style or genre in popular music has a lot to do with both personal and cultural identity. Nolan Gasser, musician and musicologist explain in his book “Why You Like It: The Science and Culture of Musical Taste,” that our musical tastes help us form our individual identities. “Music becomes that stake in the ground — ‘this is who I am,’” says Gasser. “But at the same time, the music people listened to at an early age becomes their native home comfort music. When they grow up, that music will be part of who they are, tied in with memories and growing up. All of these powers are why music is so important to us.”
Drop Dead Sunset is a collaborative project I’m involved in where the collective intentionally works within a well-established set of rules and models a musical template that was established over fifty years ago by acts like The Jimi Hendrix Experience, Cream, Led Zeppelin and Black Sabbath. Thousands of other musical groups have used this model. As a result, the music is a derivative but reliable product for a specific and demographically waning subculture.
Older people may lose interest in new popular music in the same way that they lose interest in popular fashion trends. Individuals may carry on parts of the identity they established within a social group at a certain stage of life, listening only to a narrow range of music that conforms to those norms of acceptability, or they may explore more sophisticated options like classical, intercultural, intelligent electronic music (IDM), fusion and jazz, or they may lose interest in music altogether, deeming it relatively unimportant.
Imitation is a common and practical pathway in learning. A budding visual artist may copy, trace or paint by numbers to gain experience in technique, pattern and form. Computer programmers copy and modify code snippets or entire libraries from various sources as a starting point. Musicians learn cover songs. Bands mimic the style of an existing band they are influenced by.
Beyond these starting points, mastery may begin to emerge, where rote imitation evolves into intuitive and original composition. To paraphrase Daniel Dennett and Daniel Kahneman — intuition is knowing or reaching a conclusion quickly (fast thinking) without being able to immediately backtrace how you got there (which would require slower thinking and analysis). As the Jedi would say, you unlearn what you have learned and act on instinct. Mastery is usually not achieved until years of cumulative experience allows rapid heuristic patterns to emerge. In other words, a highly experienced person may know a solution without being able to quickly “unpack” how they know they know it.
For this reason, it’s often the case that advanced musicians with years of experience may find the limitations of popular music genres (sound fashions) uninspiring in the same way a twelve year old loses interest in toys. A great example of this was the stellar “studio band” at Motown Records in that company’s heyday who were paid a flat salary and appeared on hundreds of hit recordings. Bubble gum pop was their day job. At night, many of them would unwind by playing jazz in Detroit clubs. Jazz itself is a set of models, but inherent in those models is a strong emphasis on lateral thinking through improvisation and risk taking.
The constraining “language” parameters of genre and style are less relevant if identity, ego and social validation are no longer motivating factors. The ideal of a “Band” is more often a collaboration of artists within a set of agreed possibilities by the collective. The participants can’t do too much of this or that or it will deviate from the norms of the collective agreement and target audience – a conspiracy of limitations. This thinking can be anti-innovative. Many collaborations end in “creative differences” and many successful acts falter because they veer too far from their successful recipe, or they retread the same formula after fickle popular tastes and trends change. Very few successful music artists have continued to innovate while retaining their core audience and sales figures.
Good outcomes can arise out of collaboration when differing perspectives and conflict lead to an unexpected and novel synthesis within liberal models. Bad outcomes can arise when that tension leads to lukewarm, conventional and safe compromises within conservative models. If we think of a compromise as an average of inputs, the diversity of inputs matter considerably.
What techniques can an artist employ to reduce the influence of overused models? Listening to a wide range of music from various cultural traditions or even non-musical sources can provide more diversity of inputs. Some of the most interesting innovations in music over the past century were a result of cross-cultural interpretation or collaboration between artists from distinct musical backgrounds and cultures. In 1975 Brian Eno and Peter Schmidt developed a system of oblique strategies to encourage lateral thinking. A wide range of possible approaches can be employed:
- free association
- combining unlikely or seemingly incompatible models
- intentionally using the “wrong” tools or techniques
- forcing unnecessary or arbitrary constraints: like playing an instrument incorrectly or with a reduced set of options
- Decomposing an existing process or media and recomposing it
- Reflecting on whether the work is overly derivative of a previous work (of another artist or your own)
Edward de Bono, the psychologist who coined the concept of lateral thinking, defines four broad types of tools:
- Idea-generating tools break current thinking patterns—routine patterns, and the status quo
- Focus tools broaden where to search for new ideas
- Harvest tools ensure more value is received from idea generating output.
- Treatment tools that promote consideration of real-world constraints, resources, and support.
While many highly successful artists claim that their success was a mere byproduct of their instincts, it’s unlikely they were successful without consciously or subconsciously applying popular models — it’s more likely they took them for granted as implicit. Perhaps all this boils down to a simple prescription: if the acceptable models of popular music are too constraining, and popularity is not a significant criterion in your definition of success, then just stop trying to make music that you think people will like and it will increase your chances of creating something original and innovative.