Beethoven's Tenth: Ode to AI

Artificial intelligence and our creative future

In my previous post, I explored the deeply entwined history of chess and artificial intelligence, as illustrated through the evolution of computational theory and machine learning. In this post, I want to focus beyond the use of computation to solve optimal strategies for games and discuss how AI can help advance creative thinking.

I. Checkmate: Machines are the New Grandmasters

As I outlined previously, our cultural fascination with chess and computation began with the Turk, a mechanical chess-playing automaton created in the 18th century by Hungarian inventor Wolfgang von Kempelen. Even though the Turk was an ingenious hoax, secretly operated by skilled chess players hidden inside its apparatus, it sparked an enduring interest in the possibility of a mechanical intelligence that could vie with or even surpass human capabilities. Following the transformative impact of World War II on computing technology, this legacy saw chess become a key benchmark for twentieth-century AI research. Mid-century AI pioneers such as Claude Shannon outlined foundational strategies for programming chess algorithms, and early programs such as Turing’s rudimentary chess algorithm and Kotok-McCarthy’s system in the 1950s and 60s demonstrated the potential for computers to play chess at a basic level. But this would soon change.

Leveraging techniques such as tree search and evaluation functions to assess potential moves, IBM’s Deep Blue supercomputer defeated world champion Garry Kasparov in a landmark 1997 six-game match, showcasing the power of brute-force computation and advanced heuristics. This victory symbolized a shift, showing that AI could not only match the highest realms of human cognitive ability—Kasparov being the highest-rated chess player in history at that time—but could far surpass it in specific closed domains. Machine learning approaches have since revolutionized chess and other games. For instance, Google’s AlphaZero, released in 2017, uses reinforcement learning to master chess and other games without human input, achieving superhuman playing strength through self-teaching. Human grandmasters no longer have legitimate expectations of beating the strongest computerized opponents; ironically, it is now Stockfish and other chess engines that are helping to make humans stronger players through far-reaching analysis of their games.

II. From Code to Canvas: AI Joins the Creative Revolution

The entwined evolution of chess and AI nevertheless underscores an ongoing drive to push the boundaries of technology and cognition—one that took on new life with OpenAI’s public release of ChatGPT in 2022. The sophisticated language models underlying ChatGPT (and similar chatbots) “learn” by absorbing vast corpora of text, uncovering statistical relationships between words and phrases that they can then leverage to achieve remarkable humanlike capabilities. Almost immediately, an onslaught of AI-generated essays began raising complex plagiarism-detection challenges for teachers and professors. But ChatGPT also proved that it could outperform most human candidates on AP exams and the SAT, while also excelling at the LSAT and exams for the bar, medical school, and business school. Its capabilities include drafting corporate memos and legal contracts as well as solving complex coding challenges. Researchers have since extended AI systems to patterns in other domains, such as constructing images and generating video by anticipating sequences of pixels and frames, or generating music based on datasets of melodies, timbres and rhythms.

Generative AI is, as yet, in its infancy—and yet its advent has sparked much apprehension about its potential to replace human workers, especially in professions that involve tasks such as data analysis, legal contract work, or accountancy. But what about creative work? We have traditionally regarded creativity as one of our most inherently human attributes, integral to our sense of self and fundamentally resistant to disruptive transformation. But does generative AI now pose the most formidable challenge in history to human creativity? Does it render entire fields, such as writing, visual design, and composing music, susceptible to widespread automation?

Many scholars continue to believe that creativity transcends mere pattern recognition, involving the capacity to imagine and express a deeply human interplay of emotion, intuition, insight, and life experience. In such a worldview, artificial intelligence could never produce a work such as Beethoven’s Ninth Symphony or Joyce’s Finnegans Wake. Such a perspective dovetails with the view of Miguel Nicolelis, a Brazilian neuroscientist at Duke University, who asserts that artificial intelligence will never replicate the human brain: “You could have all the computer chips ever in the world and you won’t create a consciousness,” he has stated. He sees the brain’s dynamic, chaotic, and networked interactions among billions of neurons as irreducible to mechanistic or computational models, which seems to offer some reassurance in the face of the more grandiose predictions about AI’s future.

III. Unfinished No More: AI Completes Beethoven’s Tenth

But let’s consider again the instances of Beethoven and Joyce. After completing his Ninth Symphony in February 1824, Beethoven made some sparse sketches towards another symphony, although the work never progressed due to his health (he died in 1827, aged 56, after a prolonged illness). NPR even published a joke story about the discovery of “Beethoven’s Tenth Symphony” on April Fool’s Day in 2012. However, to mark the 250th anniversary of Beethoven’s birth, a team involving music historians, composers, and computer engineers worked to bring Beethoven’s Tenth to fruition by using AI to imagine what those sparse sketches might have become. A collaboration among Deutsche Telekom, the Mindshift, and Artrendex, with the support of academic experts from Harvard, Cambridge, and Rutgers, the project involved training the AI system on Beethoven’s entire catalog, including his symphonies, chamber music, and piano sonatas, allowing the system to identify his typical stylistic patterns, harmonic progressions, and structural inclinations. Then, human experts worked closely with the AI, interpreting its suggestions and guiding its compositional process in a way they judged to be consistent with Beethoven's artistic voice, with the goal of plausibly re-creating his Tenth Symphony.

After two years of work, the symphony was performed on stage for the first time in Bonn, Germany, in 2021. How closely it represents the symphony that Beethoven would have written is open to fierce debate—and yet the project represents an ambitious co-creative endeavor across centuries that would never have been possible in the era before AI. The story of creating Beethoven’s Tenth is recounted in Hannes M. Schalle’s documentary Beethoven X: The AI Project.

IV. Passing the Pen: Joyce’s Bid for James Stephens to Finish Finnegans Wake

The Irish novelist James Joyce, the most innovative prose stylist of the 20th century, faced many challenges during his composition of Finnegans Wake, his labyrinthine experimental final novel written over 17 years from 1923 to 1939. Struggling with his worsening eyesight, a condition requiring multiple surgeries, and with the novel itself, a kaleidoscopic tapestry of languages, allusions, and experimental narrative techniques, Joyce began to doubt whether he could sustain the relentless effort to weave its densely layered prose. This led him to consider an unprecedented handover, which his biographer Richard Ellman later described as “one of the strangest ideas in literary history”—he considered delegating the completion of the work to another Irish writer, James Stephens.

A fellow Dubliner, Stephens was known for his works of Irish folklore and for original novels including The Crock of Gold (1912). He shared Joyce’s fascination with myth, language, and the cadences of Irish speech, which seems to have convinced Joyce that he could master the polyphonic voices, multilingual wordplay, and vast trove of allusions that make up Finnegans Wake. Superstition seems to have played a role, too: Joyce and Stephens had been born in the same hospital, one week apart, and Stephens shared both the first name of Joyce himself and that of Joyce’s fictional alter-ego Stephen Dedalus. Ever attuned to synchronicities, patterns, and coincidences, Joyce found these details compelling, interpreting them as a possible omen that Stephens was destined to play a role in Finnegans Wake. Joyce even considered attributing authorship to “JJ&S,” both “Jameses Joyce & Stephens” and a pun on the Dublin whiskey distiller John Jameson & Son. Even though Joyce’s collaboration with Stephens never materialized—partly because Stephens, while flattered by the idea, hesitated to take on a project of such daunting complexity—Joyce’s contemplation of entrusting Finnegans Wake to Stephens offers a peculiar historical lens through which to explore the potential handover of a creative work to artificial intelligence.

The episode reveals Joyce’s willingness to reimagine authorship as a potentially shared or transferable act, suggesting that even a work as singular and idiosyncratic as Finnegans Wake could, under the right circumstances, be extended or completed by another. Joyce also worked on the novel during a period of seismic technological shift in communications technology, marked most notably by the rise of commercial radio broadcasting, which began in the early 1920s and quickly became hugely popular. By the early 1930s, most households in the United States owned at least one radio receiver, transforming ease of access to entertainment and information. Always alive to how technological change could alter the transmission of culture (he had opened Dublin’s first cinema in 1909) Joyce seemingly came to envision his last novel as a self-perpetuating communication machine—a “harmonic condenser enginium”—that both contained and transmitted the entire world, a supercharged and universal broadcasting service.

This concept of delegating creative work to another—whether a human or a machine—finds a provocative contemporary analogue in the rise of generative AI systems, which now offer a means of continuing or augmenting creative endeavors. If Joyce were working today, might he consider AI as a collaborator in the creation of Finnegans Wake? Generative AI could be trained on Joyce’s drafts to emulate some of the dense intertextual style and linguistic inventiveness of the work. AI could also theoretically analyze the structures and motifs already present in Finnegans Wake to propose extensions or refinements in a manner conceptually similar to how it helped imagine Beethoven’s Tenth Symphony.

But this hypothetical handover also raises profound questions. Could AI truly emulate the Joycean project of telling a grand universal story deliberately obscured by the fall into night, and, after the fall of the Tower of Babel, by multilingual layers of punning wordplay? Or would its output be a hollow simulacrum, technically impressive but devoid of the genius that shaped Joyce’s vision? While AI excels at identifying patterns and generating plausible imitations, it lacks the human capacity to synthesize subjective experience, grapple with ambiguity, and produce art that resonates with the fullness of human emotion and intellect. That is the dystopian future that some fear—a world flooded with work that lacks authenticity or soul and yet makes it difficult for human artists to be heard (and, more bluntly, paid) above the fray.

V. Conclusion: Is AI the New Muse in Creative Collaboration?

The debate around AI and creativity is fraught with wild speculation, ranging from the idea that AI is an unprecedented boon to humankind to the notion that AI will impoverish, enslave, and ultimately kill us all. None of this feverish speculation is likely to transpire. When it comes to creative work, I believe the potential of AI lies not in replacing human creators but in augmenting their processes. AI systems can assist artists and writers by generating drafts, suggesting stylistic variations, or uncovering novel connections within existing frameworks, but the creator needs to maintain a key role in carefully mediating the end product, ensuring that the resulting work retains an authentic human touch. As such, it might be more useful to imagine AI less as an autonomous creator than as part of a continuum of tools—like typewriters, editing software, or music compositional tools—that have already expanded the capabilities of writers and other artists.

The integration of AI into creative processes can certainly open new avenues for collaborative creativity. But while machines may assist, augment, or even complete creative work, we must remember that the essence of art remains intrinsically tied to our uniquely human capacity to imagine, reflect, and feel. In a future post, I will look more closely at what such a future might entail for working writers, and especially screenwriters, who decide to harness AI as a creative tool.