Mind Matters News and Analysis on Natural and Artificial Intelligence
Old Trumpet Brick Wall
Old worn trumpet stands alone against a grungy pealing white brick wall outside a jazz club
Featured image: Old trumpet leaning against brick wall/ezume, Adobe Stock

Fan Tries Programming AI Jazz, Gets Lots and Lots of AI…

Jazz is spontaneous, but spontaneous noise is not jazz

It’s one thing to talk about AI creating music, it’s another to try making it happen. In 2017 @caryhk published a video at YouTube of his efforts to train an AI to generate jazz. One thing he succeeded at was exposing the myth of AI creativity.

Even @caryhk cringed at most of his results. Very few even sound like jazz of any kind, let alone good jazz. Most reminded me of kids pounding a keyboard.

Jazz is spontaneous, but spontaneous noise is not jazz. In his great book How to Listen to Jazz, (2016) jazz musician and critic Ted Gioia observes: “This bedrock layer of improvisation, almost beyond the scope of musicology, is the psychology or personality of the individual musician.” (Emphasis added)

That is the blindspot of AI creativity: There’s no one home. There’s no “personality” behind the “creation.”

Creativity arises from the soup of the moment, in music and in life (which is why Isaac Newton and Gottfried Leibniz simultaneously invented calculus when the circumstances were right). Gioia notes:

Strange to say, new art forms are similar to the plague or a virulent flu in how they spread. Art and disease proliferate via contagion, and similar conditions favor both. Densely packed populations, many individuals coming and going via land and waterways, an overheated mixture of people recently arrived from different locales, informal settings where they intermingle in close contact, a culture and environment that emphasize communal activities and get-togethers—these are nightmare conditions for anyone trying to stop an epidemic, but they are the same ingredients that can spur world-changing artistic revolutions.

What is created is not just a twist on an existing creation: New things draw on what exists but, in an ill-defined sense, they also extend them. Which is why we have cell phones and the Apollo astronauts did not, and why the Apollo astronauts had a moon rocket and Thomas Jefferson did not. Our creative work stands on what’s come before, just as what came before built its predecessors.

@caryhk’s video shows that AI creativity is not like that: AI creativity is always a remix of what it’s got. It may “learn” (that is, have encoded within its networks) that a musical 5th produces a certain type of sound and that other patterns “work well” when preceded by a 5th; that’s called parroting its training data.

This sort of trite repetition can fool hearers who are less familiar with a creative discipline such as music because they haven’t heard it all before, from better artists. But their lesser sophistication does not by itself elevate kitsch to art.

AI can produce unexpected output — for example, believable faces of people who never existed — but that does mean it’s creative. It is algorithmically assembling faces from millions of relevant examples. Creative work is fresh and unexpected, but being unexpected does not, by itself, make a work creative.

AI will succeed at generating faux artifacts such as paintings, sculptures, and jazz. But, on its own, it will not produce a Thelonious Monk-like advance to the art form. AI will get better at turning out pleasing jazz-like Muzak. But we should not let our untrained taste fool us into believing it is doing something more creative.


More from Brendan Dixon: on AI and the arts, especially jazz:

AI can’t do jazz because spontaneity is at jazz’s core. AI “artists”—in all the forms presently available — merely replay their programming.

Could AI authentically create anything?

AI creates kitsch, not art

and

The underwhelming creativity of AI

Featured image: Old trumpet leaning against brick wall/ezume, Adobe Stock


Brendan Dixon

Fellow, Walter Bradley Center for Natural & Artificial Intelligence
Brendan Dixon is a Software Architect with experience designing, creating, and managing projects of all sizes. His first foray into Artificial Intelligence was in the 1980s when he built an Expert System to assist in the diagnosis of software problems at IBM. Since then, he’s worked both as a Principal Engineer and Development Manager for industry leaders, such as Microsoft and Amazon, and numerous start-ups. While he spent most of that time other types of software, he’s remained engaged and interested in Artificial Intelligence.

Fan Tries Programming AI Jazz, Gets Lots and Lots of AI…