A CBI Tomash Virtual Lecture on the History of Machine Learning
Join CBI's 2020 - 2021 Tomash Fellow Aaron Mendon-Plasek, History Department, Columbia University presenting his paper "How 1950s ideas about creativity in machine learning continue to inform social and political possibility today".
This virtual event is free and open to the public. Registration closes at 12 CST on Tuesday, May 4.
Registration is now closed.
Abstract: Accounts of early machine learning often center on late 1950s work well-known to artificial intelligence researchers such as Samuel's checkers-playing program and Rosenblatt's perceptron. However, as Newell observed retrospectively in 1983, by 1955 machine learning researchers had already splintered off from what "became the AI community.” If early machine learning wasn’t artificial intelligence, what was it? This talk provides an answer.
Drawing on a study of thousands of published technical articles and research in more than 12 archives, the objectives of this talk are twofold. First, I show how problem-framing strategies of early machine learning valorized a notion of creativity as the “production of completely new ideas not deducible from known data.” In contrast to 1950s “artificial intelligence” projects employing limitations of scope as a necessary evil, machine learning embraced such limitations as facilitating a novel kind of mechanized decision-making (1) that could make human-like judgments given messy, unexpected real-world data, and (2) that could redefine “significance” for a given task apart from the programmer’s understanding of the task to be performed.
Second, by tracing a constellation of efforts ranging from individual commercial labs to transnational research networks spanning the globe, I identify two existential crises that emerged for researchers attempting to implement this specific, idealized form of creativity on digital computers in the 1950s and 1960s: namely, (1) the inability to compare different machine learning systems that ostensibly did the same task, and (2) the inability to articulate what made machine learning qua pattern recognition a unique discipline. The intellectual (and political) solution devised by machine learning researchers in the 1960s to address these crises offered new modes of scientific identity for individual researchers and enlarged what constituted a legitimate description of the physical and social world.
These historical forms of being and knowing continue to frame possibility in ongoing debates regarding the relationships between AI and society today.
About Aaron Mendon-Plasek
Aaron Mendon-Plasek is a historian of science and U.S. history, and a Ph.D. candidate at Columbia University. His work examines how schemes of quantification, including their material, cultural, technical, and institutional instantiations, have been used to imagine, enact, and justify social order. His dissertation has been supported by a variety of institutions, including fellowships from the National Science Foundation, Columbia University, and the Charles Babbage Institute. He holds an M.Phil and MA in history from Columbia University, an MA in humanities and social thought from NYU, an MFA in writing from the School of the Art Institute of Chicago, and a BS in physics and astronomy and a BA in writing from Drake University.