Speaking Machines and Living Languages

Oct 17, 2025·
Matt Coler
Matt Coler
· 2 min read
University of Warsaw
Abstract
Artificial intelligence has revolutionized speech synthesis, yet this breakthrough excludes all but a handful of major languages. Through an interactive audio history spanning from Von Kempelen’s 1791 mechanical apparatus to today’s neural architectures, this talk traces how each era reimagined what speech fundamentally is: from physical resonance to encodable information to learnable patterns. Live demonstrations show how small regional languages can achieve natural synthesis, challenging assumptions about data requirements. But technical solutions raise deeper questions: What transforms when voice moves from embodied cultural practice to decontextualized acoustic reproduction? When lived experience becomes technological stimulus? These questions reveal why speech technology increasingly requires methodological contributions from social sciences and humanities: not as auxiliary concerns, but as perspectives on what breakthrough actually means in a field dominated by engineering.
Date
Oct 17, 2025 9:00 AM — 10:00 AM
Event
Location

Modern Languages Building, University of Warsaw

ul. Dobra 55, Warsaw, Mazovia

Invited keynote lecture at the University of Warsaw’s prestigious BREAKTHROUGHS conference, which showcases the latest research in humanities and social sciences.

Talk Overview

Through an interactive audio history, this presentation explores the evolution of speech synthesis technology from Von Kempelen’s 1791 mechanical speaking machine to contemporary neural architectures. The talk demonstrates how each technological era fundamentally reimagined the nature of speech itself:

  • 18th-19th Century: Speech as physical resonance and mechanical reproduction
  • 20th Century: Speech as encodable information and signal processing
  • 21st Century: Speech as learnable patterns in neural networks

Key Themes

Technical Accessibility

Live demonstrations illustrate how small regional and under-resourced languages can achieve natural-sounding synthesis, challenging conventional assumptions about the data requirements for quality speech technology.

Critical Questions

The talk raises fundamental questions about the transformation of speech from embodied cultural practice to decontextualized acoustic reproduction:

  • What is lost when voice becomes data?
  • How does technological reproduction affect linguistic and cultural identity?
  • What role should humanities and social sciences play in shaping speech technology?

Interdisciplinary Perspective

This presentation argues that speech technology development increasingly requires methodological contributions from social sciences and humanities—not as supplementary concerns, but as essential perspectives on what constitutes meaningful technological “breakthrough” in a field dominated by engineering approaches.

Conference Context

The BREAKTHROUGHS conference brings together leading researchers to present cutting-edge work at the intersection of humanities, social sciences, and technology. This keynote opens the second day of the conference, setting the stage for discussions about responsible research assessment and societal impact.

Venue

Location: AULA 0.410 (ground floor)
Building: Modern Languages Building
Address: ul. Dobra 55, Warsaw, Poland
Institution: University of Warsaw

Speech Technology Lesser-Resourced Languages Speech Synthesis Ethics in AI Multilingualism
Matt Coler
Authors
Matt Coler
Associate Professor of Speech Technology