Towards a History of Artificial Intelligence
Workshop at Columbia University
May 23-24, 2019
Brown Institute for Media Innovation Pulitzer Hall
No thorough professional history of artificial intelligence and machine learning from World War II to the present exists, despite the longstanding importance of the field to intellectual history, the history of science and technology, and its spectacular and explosive rise in quotidian systems worldwide in the last decade, for good, for evil—and, well, for—the jury’s still out.
No one scholar could write a history of artificial intelligence and machine learning with global range, technical rigor, and contextual richness. Working alongside the capacious scholarship of reflective technical practitioners, anthropologists, sociologists, and historians of computing who have offered important perspectives on AI's complex history, we seek to scaffold a more comprehensive account by bringing together a community of scholars from around the world with diverse methodological commitments and subjects of inquiry. With support from the Center for Science and Society and Dean of Social Sciences at Columbia University, the Leibniz Fund, and the Leverhulme Center for the Future of Intelligence (CFI) at the University of Cambridge, we have convened a group of historians to share their current research and plan worthwhile outputs on the major stages of AI’s development since the Second World War.
While seeking to revisit major narratives centered on the UK and US, we plan as much as possible to incorporate a global story of AI, which has often been told predominantly in an Anglo-American framework, and to draw together a broad range of methodological approaches, kinds of histories, and historians.
Thursday May 23 - Evening
4:00 PM Registration
4:45 PM Welcoming and opening remarks: Mark Hansen, director, Brown Institute for Media Innovation
5:00 PM Panel 1 - Whose Intelligence?
7:00 PM Dinner
Friday May 24 – All day
9:00 AM Panel 2 - Creative Reasoning
11:00 AM Panel 3 - Sites of Decision Making
2:00 PM Panel 4 - Computing Institutions and Ideology
4:00 PM Panel 5 - Knowledge Practices
6:00 PM Wrapping up session
Thursday May 23
Panel 1 Whose Intelligence?
Lucy Suchman has proposed that artificial intelligence “works as a powerful disclosing agent for assumptions about the human.” What behaviors or competencies are selected as the hallmarks of intelligence? Whose expertise is considered sufficient for the engineering of artificial intelligence? Whose labor and knowledge is seen as automatable, and whose is held apart as uniquely human? But we might go further to say that our histories of artificial intelligence similarly serve as ‘powerful disclosing agents’ for our own assumptions. Where do we look and what do we ask when we write the history of intelligence, whether artificial or human? What have our histories left out? These panelists invite us to consider alternate origins, multiple localities, competing visions, failed efforts, and gendered ideologies, as they recover and reconstruct what artificial intelligences reflect back about ourselves.
Chair: Stephanie Dick, University of Pennsylvania
Electric Eve: Modernism and Gendered AI, 1816-present
Allegra Fryxell, University of Cambridge
"If your idea really did work": Donald Hebb, Nathaniel Rochester, and the Other Origin of Neural Networks, 1945-1960
Yvan Prkachin, Harvard University
Who's Teaching, Who's Learning?: Gendered AI Lessons from Postwar American Educational Computing
Joy Rankin, Independent Scholar
The Spring of Artificial Intelligence in Its Global Winter: Korean Language and AI Researchers in the Late 1980s
Youjung Shin, Korea Advanced Institute of Science and Technology
“A programme which provides for the transformation of its own order”: Early machine learning and its use in articulating social questions
Aaron Plasek, Columbia University
*No recordings
Friday May 24
Opening Remarks - Meredith Whittaker, AI Now
Panel 2 Creative Reasoning
Reasoning and creativity are often characterized as opposite human faculties - the former associated with rule following and the latter with spontaneity. However, this panel demonstrates that, in the context of artificial intelligence research, both are moving targets and they have rich and surprising intersections. In some historical contexts, formalization is seen as the key to creativity. Elsewhere, techniques from art history have informed automation efforts. Artificial intelligence has in turn been put to work in the art world. This panel explores the character, limitations, and possibilities of formalization and its relationships with creativity, broadly conceived.
Chair: Matt Jones, Columbia University
Face as a “Mathematical Problem”: History of Facial Detection from Chinese Figural Art to 19th-Century European Criminology and Psychiatry
Yue Zhao, NYU Steinhardt
Creativity for the Information Age: AI and Soviet Computer-Based Education
Katya Babinsteva, University of Pennsylvania
Paraconsistent Society: Mathematical Logic and Social Paradox
Rodrigo Ochigame, MIT
Reinventing the Wheel (again): The forgotten histories of AI and the Arts
Sofian Audry, University of Maine
Group Discussion
All Panelists
Panel 3 Sites of Decision Making
Whether concerning individuals, bureaucracies, or societies, the development and use of models as decision-making technologies is always a political act. This panel examines a variety of historical cases in which the practices of modeling and simulation by different historical actors constrained and informed political and technical possibility across a range of concerns, including finance, labor, bureaucracy, and even human capacity itself.
Chair: Jonnie Penn, University of Cambridge
Financial Heuristics: Investor Judgment in Artificial Intelligence Research, 1950-1963
Devin Kennedy, Harvard University
Are Neural Networks Neoclassical? Utility, Loss, and Cost from Wald to TensorFlow
Michael Castelle, University of Warwick
A Brain Model for the Perception of the Outside World
Rudolf Seising, The Research Institute for the History of Science and Technology Deutsches Museum, Munich, Germany
Computing as Organizational Behavior: Rules, Judgment, and the Automation of Bureaucratic Decision-Making, 1958–1962
Daniel Volmar, Harvard University
“In the Nervous System of the Beast”: Mapping Resistance in the Artificial Intelligence Industry
Sarah West, AI Now Institute
Group Discussion
All Panelists
Panel 4: Computing Institutions and Ideology
This panel explores how various national and transnational AI and standards-building projects have been mobilized as solutions to political questions of ideology, trust, and justice. These histories of calculating institutions reveal important gaps in oft-repeated progressive narratives of the development of AI and simple hierarchical conceptions of power. These historical cases exemplify how technologies of trust have enabled institutions to know themselves, articulate worldviews, assert authority, and imagine the future.
Chair: Laine Nooney, NYU Steinhardt
Public, as in Nongovernmental: Negotiating Trust in the New Cryptography, 1972-1984
Gili Vidan, Harvard University
The Standard Head
Stephanie Dick, University of Pennsylvania
Was the Fifth Generation a Failure? It Depends on Who you Ask—Fuchi, Feigenbaum, and the Japanese Fifth Generation Computer Systems Project, 1978-1995
Colin Garvey, Rensselaer Polytechnic Institute
Panel 5: Knowledge Practices
Historical epistemology and attention to practice have long been central concerns for historians of science, but have attracted much less energy in the history of computing. This panel bridges the two fields by exploring ways of knowing and forms of practice in artificial intelligence. From competitions that serve to benchmark AI accomplishments, to the fashioning of artificial intelligence systems as scientific instruments, these panelists explore what artificial intelligence practitioners actually do. What tools and epistemic frameworks do they craft? What and how do they know?
Chair: Theodora Dryer, UCSD
The Machine and the Molecule: Chemistry in the History of AI
Evan Hepler-Smith, Boston College
Three Open Problems for Historians of AI
Momin Malik, Harvard University
Nondeterminate Lines: Pathfinding through Artificial Intelligence
Andrew Johnston, North Carolina State University
The rule of game has changed: ImageNet Challenges before and after Convolutional Neural Network
Yoehan Oh, Rensselaer Polytechnic Institute
The History of the Graphics Processing Unit in Contemporary AI
Yaqub Chaudhary, Cambridge Muslim College
Organisers: Stephanie Dick, History and Sociology of Science, Pennsylvania, Matthew L. Jones, History, Columbia; Big Data and Science Studies Cluster, Center for Science and Society, Jonnie Penn, HPS, Cambridge, Harvard Berkman-Klein and Leverhulme Centre for the Future of Intelligence, Aaron Plasek, History, Columbia University