AJ Chronicles: The Excellence Problem and Why it Matters

This week we collected 113 stories on ArtsJournal. Here’s what I learned:

Next month, the French-Canadian harpsichordist Jean Rondeau will perform the Goldberg Variations three different ways in a single concert: solo keyboard in the traditional manner; arranged for strings, flute, and continuo; and a third approach he hasn’t yet revealed. In an interview with Bachtrack this week, he was asked: when does Bach cease to be Bach? It’s an interesting provocation, and also, I think, a useful way into a question about what is the essential essence of an artistic work and what it can or should do.

More than an abstract or academic question, though, it has relevance because of the currently raging debates about creativity and AI.

The AI creativity debate has landed on the question of “excellence” as a central pillar of resistance. The argument runs something like: AI cannot produce genuinely excellent work because it lacks the training, the suffering, the embodied human experience that excellence requires. Therefore, human creativity is the apex of what art is or expresses and deserves protection. I have sympathy for the conclusion, but I think the argument is the wrong. Or rather, it’s making different claims and calling them one.

Is it possible that what most of these arguments are actually defending is credentialism — the idea that excellence requires certain credentials, lineage, preparation? Or provenance — the idea that human origin confers value? Both are real claims worth making, but they’re not excellence claims. Excellence is about what the work achieves. A piece of music either moves you, challenges you, realizes what it set out to do, or it doesn’t. The credential of the maker is a separate question. The origin of the maker is a separate question still. And the context in which you experience the work is yet another.

I don’t mean to be pedantic, but I think these distinctions really matter if we’re really going to figure out the place of AI in creativity. Four stories this week suggest layers to this debate:

Stanford Thompson, the musicians and entrepreneur who runs Equity Arc and trained at the Curtis Institute, published an essay this week that explores this idea from a different direction. He describes sitting behind the audition table at Curtis and starting to notice something. He wasn’t just hearing musicianship, he was hearing “access layered over time expressing itself as readiness” — years of private instruction, quality instruments, coaches who knew the path. The student who cleared the bar looks excellent. The students who didn’t, often weren’t less talented, they were less prepared. We’ve been calling preparation merit and we’ve been calling credentials excellence. These are not the same thing, and this conflation has costs.

The historical performance movement, the decades-long project of recovering Baroque and Classical music performed on period instruments, finds itself asking a version of the same question. Nearly half a century on, practitioners are grappling with whether historically informed performance has calcified into its own orthodoxy. The revolutionary challenge of half a century ago to received standards has over time become the received standard. This is what happens when provenance (“authentic” instruments, “authentic” practice) substitutes for the excellence question.

At the Metropolitan Opera, director Yuval Sharon has staged Tristan und Isolde with a new ending: rather than dying, Isolde gives birth to a child during the opera’s final minutes. Much of the critical objection amounts to: Wagner didn’t intend this. That’s a provenance argument. Whether the staging is excellent — whether it produces something true and powerful and worth two hours of your attention — is an entirely different question, of course, and one the reviews are genuinely divided on. These are not the same debates, even though they’re being expressed in the same sentences.

Then there’s the Korean band BTS, which returned this week from mandatory military service to find the world had maybe move on. Their new album is, as Slate put it, maybe not that great. Their credentials haven’t changed; the provenance is intact. But the excellence question — does the music achieve what it sets out to do, in the moment it appears, and is it still fun, came back with a different answer.

This is actually a good moment to get precise about all three terms — provenance, credentials and excellence — because AI is increasingly forcing the question. Provenance matters: human lineage carries real value, and we should be honest about why. Credentialism matters: serious preparation produces things that underprepared shortcuts don’t.

But neither is a substitute for the excellence question itself and conflating them in the AI debate doesn’t solve anything. If an AI-generated creation achieves what it (or the human behind it) sets out to achieve — it moves people, it challenges assumptions, it does the work it was supposed to — that doesn’t mean it isn’t excellent because it was made by AI. Rather, I think the response is that excellence isn’t the only thing we value. Provenance is also something we value. Context is important. Human experience too. We need to make the case for why these matter and how. That’s a more difficult, more honest, but ultimately more clarifying argument.

An organically-farmed tomato might look less appealing than an industrially-grown version with its glowing color and perfect shape. And in fact the Franken-mato might be better sometimes; organic vegetables can be a crapshoot. Both are tomatoes, but it’s the hidden qualities that make the case.

Also Worth Your Attention

When attendance becomes the measure of excellence, programming follows the metric. Denmark has restructured its museum subsidy model around visitor counts: institutions must now meet a minimum of 10,000 annual visitors to retain government support. This sounds like accountability, but it’s also a quiet redefinition. The institution that draws more people is the institution worth funding. Attendance captures popularity, not depth of engagement, educational reach, or the civic value of a collection. What Denmark is answering without quite saying so is whether museums exist to serve existing audiences or to expand them. That’s not a neutral question, and visitor counts don’t answer it neutrally.

The AI attribution scandal is also an access story. The twin disclosures this week — the New York Times dropping a freelancer for using AI in a review and Hachette pulling a novel under suspicion of AI authorship — framed the AI attribution question as a problem of authenticity and disclosure. Both are legitimate, but there’s perhaps a more troublesome question underneath. If AI assistance is invisible and increasingly ubiquitous, we should assume it will be used. And the writers most likely to be caught are perhaps those with the least economic ability to spend more hours on a piece: freelancers, not those staffers paid by institutions. Enforcement may end up measuring something other than what it claims to measure, which is, again, Stanford Thompson’s audition table problem, in a different room.

---

Editor’s Note: These weekly essays are meant to connect stories from the week to larger trends and ideas across the arts world. To see all the stories on which these essays are drawn from, subscribe to ArtsJournal’s free daily and weekly newsletters. To support our work, sign up at Patreon or subscribe to our Substack newsletter.