Miðstöð stafrænna hugvísinda og lista: Tom Tiest

Tom Tiest, University College PXL-Music, heldur fyrirlestur í fyrirlestraröð Miðstöðvar stafrænna hugvísinda og lista. Fyrirlesturinn nefnist Erring on the Side of Beauty: Making Music from Mistakes in Art Education

Fyrirlesturinn verður haldinn í stofu 206 í Odda 10. nóvember kl. 12:00 til 13.00. Verið öll velkomin og takið endilega hádegismat með ykkur!

Í fyrirlestraröð MSHL kynna sérfræðingar úr ýmsum greinum stafrænna hugvísinda og lista rannsóknir sínar í stuttum og aðgengilegum fyrirlestrum.

Fylgist með fyrirlestrinum í streymi á YouTube-rás MSHL.

Smellið hér til að kynna ykkur dagskrá fyrirlestraraðarinnar.

Um fyrirlesturinn

This research explores how Large Language Models (LLMs) and generative artificial intelligence (GenAI) tools can be employed as pedagogical and creative partners within education. Central to the investigation is a re-evaluation of so‑called “errors,” which are not approached as mistakes to be corrected, but as inspiring or meaningful learning moments emerging from contingency, instability, and unpredictability. Rather than aiming for flawless musical results, the research prioritises process-oriented work, experimentation, and reflection as core values of musical learning. The project followed a clearly delineated experimental learning trajectory. In its initial phase, commercial online 'Text-to-Music Generators’ such as AIVA, Suno, and Udio were explored within an educational context. While these tools seemed to offer sonic and stylistic consistency, they ultimately, because of multiple factors discussed later on, proved less suitable for in‑depth, learning‑oriented use. Apart from AIVA, their largely ‘closed systems’ in free versions provide limited insight into underlying processes and offer little room for intervention, adaptation, or structural experimentation. Moreover, their strong emphasis on optimised end products leaves little space for deviation, serendipity, and discovery, all of which are essential components of educational learning processes. As an alternative, the research shifted towards general-purpose LLMs capable of generating and modifying HTML files and the Web Audio API. This transition enabled a more open and flexible learning environment, in which pupils and students could actively construct, analyse, alter, and reconfigure musical systems. Within this context, LLMs functioned not as automatic producers of finished compositions but as co-designers in the creation of browser-based instruments, sequencers, and experimental digital learning environments comparable to simplified DAW-like interfaces. Importantly, creativity within this educational framework extended beyond prompt formulation. Once these digital instruments and tools had emerged, learners could then engage creatively with their instruments. Their musical learning practices were shaped by the limitations and unexpected behaviours of the AI‑generated systems. Music-making thus became an exploratory dialogue between human intention, musical imagination, and machine-generated outcomes. Over an extended period of experimentation involving multiple AI platforms, the research emphasises a continuous negotiation between the path offered by the generative AI and the users’ openness. In this new context, AI is neither positioned as an authority nor as a neutral tool, but as a fallible conversational partner within the learning process.