page image

CHAPTER I - All that Jazz

Thirty years ago, John Clarkeson, who was then Chief Executive Officer of the Boston Consulting Group, wrote an essay that began by stating that “most of our organizations today derive from a model (such as the Ford assembly line) whose original purpose was to control creativity,” a model that was becoming less and less viable in a “rapidly changing environment.” The challenge in such an environment was to get employees to cross professional boundaries and learn to work together in new ways. Rather than seeing themselves as controlling an elaborate machine, corporate leaders needed to take on the role of orchestrating a group of talented but disparate musicians.

An obvious model for such a role was the conductor of a symphony orchestra, but Clarkeson noted that the “flaw in this analogy” was that “no one gives a CEO the music he should play.” A better model, one drawn directly from the American experience, was the leader of a jazz group. To illustrate his point, he described the leadership style of Duke Ellington, not in terms of his musical gifts, but his ability to work collaboratively with and inspire others:
His players were good but not without equal. He knew their quirks, their gifts, their problems, and he encouraged them to learn to do things they didn’t think they could do.… They developed through their membership in the group, and they learned from each other. Most of all, their capacity for innovation grew as they built on their cumulative experience.… The results were astonishing.
Clarkeson concluded by predicting that in the future, successful leaders “will be in the flow, not remote … [and] talented people will be attracted by the ability…to learn from other knowledgeable people and by the opportunity to create and grow.” This view of the evolving role of leaders and leadership has been a major theme of the Aspen Institute Roundtable on Institutional Innovation for the past decade.

At the Roundtable, Mickey McManus, Research Fellow at Autodesk, offered a new perspective on this theme by suggesting that we are now in “the Jazz Age of Cognition.” He began by noting the scale of the challenges that humans are facing: billions more people are coming, most of them concentrated in urban areas where they will need food, shelter and more. To accommodate this growth, we will need to construct some 8,500 new buildings every day—the equivalent of adding an entire New York City of buildings every month—even as much of the existing infrastructure is falling apart and in need of repair. If we are going to meet these demands in a sustainable way, we need to get much better at producing things: nearly one-third of the waste in the world is created by construction, while building materials and construction account for 11 percent of global greenhouse emissions, and building operations generate another 28 percent of GHGs; some 70 percent of spare parts created for automobiles are never used.

Even as the urgency of finding new ways of working is increasing, technology is providing both new insights about the processes that underlie creativity and new means for encouraging innovation. For example, researchers have been studying how the brain works when it is engaged in creative activity. In 2008, a group of researchers at the National Institutes of Health and Johns Hopkins University School of Medicine used an fMRI machine to study the mental activity of professional jazz pianists as they played. When the musicians were engaged in improvising (but not when they were playing previously memorized compositions), there was a marked decrease in activity in the dorsolateral prefrontal cortex, the area of the brain associated with planning and self-censorship. What this research suggests is that a key to fostering creativity is the ability to, at least temporarily, reduce one’s normal inhibitions in order to explore a path that is new and untested.

Perhaps even more significant is the emergence of new digital technologies that can enable new forms of collaborative innovation. With the advent of AI techniques such as deep learning, machines now have the ability to begin with a set of inputs (data) and then apply algorithms that are capable of generating novel connections, including connections that humans would be unlikely to make on their own—in other words, to engage in what looks like creative activity.

McManus suggested that these new tools seem to have “a weird sense of agency” that is distinctly different from older, simpler tools like a hammer: they have the ability to “surprise and confound us” with unexpected results that can push us to question and perhaps go beyond our unconscious assumptions and limitations. Computers are rapidly learning to drive cars as well or perhaps better than humans. It has become so common to see Waymo vehicles (built by Alphabet, Google’s parent) driving around the streets of Silicon Valley that it has become almost routine. And almost every day, we hear of new discoveries coming from the combination of deep learning and big data. In healthcare, AI systems have demonstrated the ability to read certain types of x-rays better than radiologists and to come up with ideas for promising drugs that humans had failed to find. In journalism, AI is being used to help reporters find important trends hidden in large datasets and to spot fake news stories.

A distinctive characteristic of AI systems, which can be either a weakness or a strength, is that they do not have common sense, the ability to draw conclusions and make decisions from practical matters. This can be a flaw that generates impossible or impractical solutions to a given problem. But it can also be a means of escaping conventional boundaries to find truly novel solutions, just as inspired jazz musicians can push beyond the conventional boundaries of music. What AI systems excel at is identifying patterns that are difficult or impossible for humans to see.

A skilled practitioner who understands how algorithms function can use them to work in new ways. A recent example is Autodesk’s Generative Design software that functions in a distinctly different way than traditional CAD/CAM software. Rather than being a tool that aids in visualizing the ideas of a human designer, this software takes a set of high level design goals for a particular products and a set of parameters (such as a type of material or cost or weight) and then generates a large set of alternatives by rapidly evaluating all possible solutions and identifying those that satisfy the initial criteria. It then invites the user to pick what the 20th century designer Raymond Loewy called the “most advanced yet acceptable” (MAYA) solution.

Much as a jazz musician improvises by responding in real time to musical ideas from fellow musicians, so a designer can learn to interact with novel ideas being generated by an AI program and create something new. The difference between the two types of collaborations is that musicians are the same species. Generative Design, on the other hand, is a process that involves collaboration between a digital tool that has vast pattern recognition capabilities and a human who brings creative thinking, common sense, and a range of subjective feelings. This new tool has already been incorporated into some of Autodesk’s more traditional design software packages and is being used to find new, innovative solutions in fields ranging from architecture and aerospace engineering to consumer products and automotive design.

FIGURE 1: Example of Generative Design

One part, a seat belt latch (right), created through a Generative Design process, that replaces an assembly of nine different parts (left): Lighter, stronger and capable of being manufactured via 3D printing, the new part is almost impossible for a human to design unaided. Source: Mickey Manus presentation for the 2019 Aspen Institute Roundtable on Institutional Innovation.

McManus concluded on a cautionary note: even though artificial intelligence may seem to be super intelligent, algorithms, like many other tools, are created by humans and therefore subject to human error or limitations. As we explore the potential of these new tools, we need to pay attention to initiatives such as aiweirdness.com, that uses humor to illustrate “how machine learning algorithms can get things wrong,” or the Algorithmic Justice League that uses storytelling to highlight the ways in which algorithms can encode human biases—especially gender or racial bias—in systems that purport to be objective.

 
 
Title Goes Here
Close [X]