About

Fabian Offert is an Assistant Professor for the History and Theory of the Digital Humanities, Director of the Center for the Humanities and Machine Learning, and Principal Investigator of the AI Forensics Project at the University of California, Santa Barbara. His research focuses on the epistemology, aesthetics, and politics of artificial intelligence, studying how machine learning models represent culture and what is at stake when they do. His most recent book, Vector Media, writes the first comprehensive history and theory of vector space as a space of universal commensurability in contemporary machine learning.

Research topics

• Computer Science
• Machine Learning
• Artificial Intelligence
• Humanities
• Data Mining
• Mathematics
• Psychology
• Art
• Philosophy
• Human–computer interaction
• Medicine
• Epistemology
• Data science
• Medical physics
• Statistics
• Cognitive science

Selected publications

• Synthesizing proteins on the graphics card: protein folding and the limits of critical AI studies

  AI & Society · 2026-02-19 · 1 citations

  articleOpen access1st authorCorresponding

  Abstract This paper investigates the application of the transformer architecture in protein folding, as exemplified by DeepMind’s AlphaFold project, and its implications for the understanding of so-called large language models. The prevailing discourse often assumes a ready-made analogy between proteins—encoded as sequences of amino acids—and natural language, which we term the language paradigm of computational (structural) biology. Instead of assuming this analogy as given, we critically evaluate it to assess the kind of knowledge-making afforded by the transformer architecture. We first trace the analogy’s emergence and historical development, carving out the influence of structural linguistics on structural biology beginning in the mid-twentieth century. We then examine three often overlooked preprocessing steps essential to the transformer architecture, including subword tokenization, word embedding, and positional encoding, to demonstrate its regime of representation based on continuous, high-dimensional vector spaces, which departs from the discrete nature of language. The successful deployment of transformers in protein folding, we argue, discloses what we consider a non-linguistic approach to token processing intrinsic to the architecture. We contend that through this non-linguistic processing, the transformer architecture carves out unique epistemological territory and produces a new class of knowledge, distinct from established domains. Our search for intelligent machines thus has to begin with the shape , rather than the place , of intelligence. Consequently, the emerging field of critical AI studies should take methodological inspiration from the history of science in its quest to conceptualize the contributions of artificial intelligence to knowledge-making, within and beyond the domain-specific sciences.

  PublisherOA PDFDOI

• Liberatory Collections and Ethical AI: Reimagining AI Development from Black Community Archives and Datasets

  2025-06-23 · 2 citations

  articleOpen access
  PublisherDOI

• A sign that spells

  Journal of Digital Social Research · 2024-12-31 · 3 citations

  articleOpen access1st authorCorresponding

  In this paper, we examine how generative artificial intelligence produces a new politics of visual culture. We focus on DALL·E and related machine learning models as an emergent approach to image-making that operates through the cultural technique of semantic compression. Semantic compression, we argue, is an inhuman and invisual technique, yet it is still caught in a paradox that is ironically all too human: the consistent reproduction of whiteness as a latent feature of dominant visual culture. We use Open AI’s failed efforts to “debias” their system as a critical opening to interrogate how DALL·E dissolves and reconstitutes politically and economically salient human concepts like race. This example vividly illustrates the stakes of the current moment of transformation, when so-called foundation models reconfigure the boundaries of visual culture and when “doing” anti-racism means deploying quick technical fixes to mitigate personal discomfort, or more importantly, potential commercial loss. We conclude by arguing that it simply does not suffice anymore to point out a lack – of data, of representation, of subjectivity – in machine learning systems when these systems are designed and understood to be complete representations of reality. The current shift towards foundation models, then, at the very least presents an opportunity to reflect on what is next, even if it is just a “new and better” kind of complicity.

  PublisherOA PDFDOI

• Synthesizing Proteins on the Graphics Card. Protein Folding and the Limits of Critical AI Studies

  arXiv (Cornell University) · 2024-05-16 · 1 citations

  preprintOpen access1st authorCorresponding

  This paper investigates the application of the transformer architecture in protein folding, as exemplified by DeepMind's AlphaFold project, and its implications for the understanding of so-called large language models. The prevailing discourse often assumes a ready-made analogy between proteins, encoded as sequences of amino acids, and natural language, which we term the language paradigm of computational (structural) biology. Instead of assuming this analogy as given, we critically evaluate it to assess the kind of knowledge-making afforded by the transformer architecture. We first trace the analogy's emergence and historical development, carving out the influence of structural linguistics on structural biology beginning in the mid-20th century. We then examine three often overlooked preprocessing steps essential to the transformer architecture, including subword tokenization, word embedding, and positional encoding, to demonstrate its regime of representation based on continuous, high-dimensional vector spaces, which departs from the discrete nature of language. The successful deployment of transformers in protein folding, we argue, discloses what we consider a non-linguistic approach to token processing intrinsic to the architecture. We contend that through this non-linguistic processing, the transformer architecture carves out unique epistemological territory and produces a new class of knowledge, distinct from established domains. We contend that our search for intelligent machines has to begin with the shape, rather than the place, of intelligence. Consequently, the emerging field of critical AI studies should take methodological inspiration from the history of science in its quest to conceptualize the contributions of artificial intelligence to knowledge-making, within and beyond the domain-specific sciences.

  PublisherOA PDFDOI

• Reproducibility and explainability in digital humanities

  International Journal of Digital Humanities · 2024-01-03 · 1 citations

  articleOpen accessSenior author
  PublisherOA PDFDOI

• Computational Humanities

  Debates in the digital humanities · 2024-06-20 · 2 citations

  book

  Bringing together leading experts from across North America and Europe, _Computational Humanities_ redirects debates around computation and humanities digital scholarship from dualistic arguments to nuanced discourse centered around theories of knowledge and power. This volume is organized around four questions: Why or why not pursue computational humanities? How do we engage in computational humanities? What can we study using these methods? Who are the stakeholders? <br><br> Recent advances in technologies for image and sound processing have expanded computational approaches to cultural forms beyond text, and new forms of data, from listservs and code repositories to tweets and other social media content, have enlivened debates about what counts as digital humanities scholarship. Providing case studies of collaborations between humanities-centered and computation-centered researchers, this volume highlights both opportunities and frictions, showing that data and computation are as much about power, prestige, and precarity as they are about _p_\-values.

  PublisherDOI

• imgs.ai. A Deep Visual Search Engine for Digital Art History

  Zenodo (CERN European Organization for Nuclear Research) · 2023-06-30

  paratextOpen access1st authorCorresponding

  We present a Web application that facilitates the deep visual search of image collections using contemporary machine learning. We discuss image retrieval as a combined computer vision/human-computer interaction problem, and propose that the standardization of feature extraction is one of the main problems that digital art history faces today.

  PublisherDOI

• Art History and Artificial Intelligence: Opportunities and Challenges of Large-Scale Visual Models in the Digital Humanities

  Zenodo (CERN European Organization for Nuclear Research) · 2023-06-30

  paratextOpen access1st authorCorresponding

  A decade ago, Johanna Drucker asked: "Is There a 'Digital' Art History?" In this paper, we attempt to reconsider Drucker's question in the light of recent developments in computer vision by investigating the epistemic implications and methodological affordances of large-scale, transformer-based vision models.

  PublisherDOI

• Reproducibility and explainability in digital humanities

  International Journal of Digital Humanities · 2023 · 4 citations

  Senior authorCorresponding
  • Humanities
  • Computer Science
  • Humanities
  PublisherOA PDFDOI

• Forschungssoftware rezensieren - Konzeption, Durchführung und Umsetzung

  Zenodo (CERN European Organization for Nuclear Research) · 2023-03-10

  paratextOpen access

  "Wissenschaftliche Rezensionen von Forschungssoftware schätzen deren Beitrag zur Lösung einer Aufgabe im Forschungsprozess, ihre Anwendbarkeit und Zielgruppe, sowie ihre handwerkliche Qualität und Nachhaltigkeit ein. Softwarerezensionen sind daher ein wesentlicher Bestandteil einer Wissenschaftspraxis, die ihre Methoden offenlegt und kritisch reflektiert. Bislang aber erscheinen noch wenige Rezensionen. Der Workshop soll Interessierten einen Einstieg bieten und somit zur Verbreitung und Anerkennung dieses wichtigen wissenschaftlichen Formats beitragen. Gemeinschaftlich mit den Teilnehmer_innen möchten wir am Beispiel kleiner, überschaubarer Tools alle Arbeitsschritte einer Rezension von Forschungssoftware am praktischen Beispiel durchführen. Im Mittelpunkt steht das Kennenlernen und Anwenden von Aspekten, mit denen eine Software besprochen werden kann. Rezensionen von Software fragen ganz unterschiedliche Kompetenzen ab und sind daher besonders gut und effizient als Team zu bearbeiten. Im Idealfall entsteht während des Workshops genug Material, um unterstützt von den Workshopanbieter_innen ohne umfangreiche Nacharbeiten eine Rezension zur Veröffentlichung einzureichen." Ein Beitrag zur 9. Tagung des Verbands "Digital Humanities im deutschsprachigen Raum" - DHd 2023 Open Humanities Open Culture.

  PublisherDOI

Frequent coauthors

• Peter Bell

  4 shared

• Leonardo Impett

  University of Cambridge

  2 shared

• Thorsten Ries

  The University of Texas at Austin

  2 shared

• Barbara McGillivray

  King's College London

  2 shared

• Karina van Dalen-Oskam

  2 shared

• Peter Bell

  2 shared

• Thao Phan

  1 shared

• Manfred D. Laubichler

  Santa Fe Institute

  1 shared