I Am Troubled (Bob L. T. Sturm)

Why It Matters

Understanding AI’s limits and societal risks is essential for policymakers, creators, and educators to prevent misinformation, protect public safety, and preserve authentic cultural expression.

Key Takeaways

• AI-generated music has historic precedents, not a novel phenomenon.
• Large language models only capture syntax, lacking true semantics or pragmatics.
• ChatGPT and similar tools often produce plausible but factually false “bullshit.”
• Misuse of AI leads to real-world harms: misinformation, health risks, legal issues.
• Author’s existential crisis stems from AI’s disruptive impact on music.

Summary

Bob Sturm, an associate professor at KTH, delivered a talk titled “I am Troubled,” in which he examined his personal existential crisis triggered by the rapid AI-driven transformation of music and cultural analytics.

Sturm traced algorithmic composition back centuries—from medieval pegboxes to the 1957 Illiac Suite—arguing that AI-generated music is not new, but the commercial surge since 2014, exemplified by tools like Music Transformer and OpenAI’s Jukebox, has amplified both possibilities and anxieties. He then dissected large language models, describing them as high‑dimensional code predictors that capture syntax but lack genuine semantics or pragmatics, invoking Searle’s Chinese‑room argument to illustrate their “synthetic text extrusion.”

He warned that these “bullshit machines” confidently spew plausible yet false content, citing real‑world failures: travelers stranded by erroneous visa advice, individuals poisoned by incorrect chemical substitutions, and chatbots reinforcing conspiracy or self‑harm. He also highlighted political misuse, such as Sweden’s prime minister consulting ChatGPT for policy and defense contractors deploying Grok, underscoring the erosion of accountability.

Sturm’s crisis signals a broader scholarly dilemma: how to harness AI’s creative power without surrendering critical judgment. He calls for transparent evaluation, ethical safeguards, and renewed emphasis on semantic understanding, urging institutions to reconsider curricula, research funding, and public policy as AI reshapes music, journalism, and decision‑making.

Original Description

At the conclusion of my ERC-funded project “MUSAiC: Music at the Frontiers of Artificial Creativity and Criticism”, I find that I am troubled. First, I am troubled by the collision of what I think I know about the engineering of large language models with how I see them being used, sold and valued. Something, I think, is deeply wrong here. Second, I am troubled by contradictions in my a/r/tography praxis — that is, being an artist, researcher and teacher in private and professional worlds increasingly inflected and infected by AI and AI-speak. These contradictions center on my creative and pedagogic practices, personal beliefs in intellectual property and individual liberty, and moral judgements about training AI systems on the world's data, and neoliberalism more broadly. My talk will focus, of course, on just how troubled I am, and my current thinking about it all.

https://www.ischool.berkeley.edu/events/2026/i-am-troubled

Co-sponsored by the Berkeley Institute for Data Science and the School of Information

The Cultural Analytics Series is a series of lunchtime talks and workshops highlighting research that focuses on the data-driven analysis of cultural phenomena.