The University of Chicago Press Podcast
Sep 2, 2026 · 41 min · 14 segments
Since the advent of computers, society has fantasized about conversing with machines. In…
Xiaochang LiGuestMiranda MelcherHost
So I am an assistant professor in the Department of Communication at Stanford University.

And my work really evolves around questions at the nexus of the history of science, history of computing, science and technology studies, and media studies.

And so I really... work on questions surrounding how is it that our information technologies shape the conditions of knowledge production, and how does that relationship then shape our understanding of the social world? And so this book really came out of a set of questions that began quite early on during my Ph.D.

studies, which, you know, this was maybe, oh, boy, a little over a decade ago when I began working in my Ph.D. And I was.

You know, I found myself kind of thinking about these questions around how it was that this form of predictive knowledge all of a sudden seemed to be everywhere.

When everyone was talking about these questions about big data at the time, right, before we now have the language of, you know, machine learning and large language models and foundation models, the main terminology at the time was big data.

And it seemed to me that all of a sudden things like financial models, climate models, surveillance, all of these things were being modeled in the same way.

And one of the things that were also being modeled this way was language at the time, right? This was also around when people were joking about how autocomplete was like kind of like collective psyche, right? A kind of crystal ball into society.

And so it struck me that it seems strange that it was reasonable that we would model markets and climate and all of these things, right, protein sequencing and language all the same.

And so it started out with a set of questions about why is this form of predictive knowledge suddenly everywhere? How did it become reasonable and desirable to do everything this way? And how did it kind of leak out into everyday life, into sort of our most mundane day-to-day communications, right? Why is it reasonable to do these large-scale simulations to think about like how you text your mother and things like that? And so that really kind of grew over the years into what the book is now, which is a history of speech and natural language processing about how these things became subject to data analysis.
How things become normal and taken for granted, I think, is always an interesting kind of investigation starting point.
So when exactly are we going to to find the origins of natural language processing? When and where are we talking?

Yeah, I mean, so I think, you know, as the kind of historian mindset is that, well, it kind of depends, right? It depends on how you define that object.

And there were a lot of ways that I'd seen it done, right? Which is that, you know, people have talked about the evolution of the field that we know of as natural language processing, which is a kind of applied area of computational linguistics, right? But that never seemed quite satisfying because it never quite explained how it was that this field became data-driven.

But I was curious, like, well, where did those ideas come from? And so I started to look and trace the particular kinds of techniques, the conceptual frameworks that shaped what we understand to be natural language processing today in this kind of very data-driven variant that we see as a centerpiece of machine learning now.

And so I tried to look for, like, how was it that we started to think about language as data, as something that's not only, you know, just electronically encoded to be sort of stored, transmitted, and even manipulated digitally.

digitally and computationally, but something that could be actually analyzed and interpreted and now, of course, generated algorithmically, right? That's something in which the substance of language could be understood through something like enumeration in the same way that we would, you know, calculate the flip of a coin? How was it that language became thought of that way? And so I ended up tracing it back to when language kind of gets brought into the laboratory, the experimental laboratory first.

in experimental and acoustic phonetics in thinking about, well, all of a sudden language as something in the body becomes measurable through these new physiological instruments.

And then that folds directly into this rise of industrial research at Bell Labs in the early 20th century in the US and research into acoustics processing for speech transmission and signal processing, where we really start to see the specification and encoding of language as something distinct from just the transmission of sound.

So I am an assistant professor in the Department of Communication at Stanford University.

And my work really evolves around questions at the nexus of the history of science, history of computing, science and technology studies, and media studies.

And so I really... work on questions surrounding how is it that our information technologies shape the conditions of knowledge production, and how does that relationship then shape our understanding of the social world? And so this book really came out of a set of questions that began quite early on during my Ph.D.

studies, which, you know, this was maybe, oh, boy, a little over a decade ago when I began working in my Ph.D. And I was.

You know, I found myself kind of thinking about these questions around how it was that this form of predictive knowledge all of a sudden seemed to be everywhere.

When everyone was talking about these questions about big data at the time, right, before we now have the language of, you know, machine learning and large language models and foundation models, the main terminology at the time was big data.

And it seemed to me that all of a sudden things like financial models, climate models, surveillance, all of these things were being modeled in the same way.

And one of the things that were also being modeled this way was language at the time, right? This was also around when people were joking about how autocomplete was like kind of like collective psyche, right? A kind of crystal ball into society.

And so it struck me that it seems strange that it was reasonable that we would model markets and climate and all of these things, right, protein sequencing and language all the same.

And so it started out with a set of questions about why is this form of predictive knowledge suddenly everywhere? How did it become reasonable and desirable to do everything this way? And how did it kind of leak out into everyday life, into sort of our most mundane day-to-day communications, right? Why is it reasonable to do these large-scale simulations to think about like how you text your mother and things like that? And so that really kind of grew over the years into what the book is now, which is a history of speech and natural language processing about how these things became subject to data analysis.
How things become normal and taken for granted, I think, is always an interesting kind of investigation starting point.
So when exactly are we going to to find the origins of natural language processing? When and where are we talking?

Yeah, I mean, so I think, you know, as the kind of historian mindset is that, well, it kind of depends, right? It depends on how you define that object.

And there were a lot of ways that I'd seen it done, right? Which is that, you know, people have talked about the evolution of the field that we know of as natural language processing, which is a kind of applied area of computational linguistics, right? But that never seemed quite satisfying because it never quite explained how it was that this field became data-driven.

But I was curious, like, well, where did those ideas come from? And so I started to look and trace the particular kinds of techniques, the conceptual frameworks that shaped what we understand to be natural language processing today in this kind of very data-driven variant that we see as a centerpiece of machine learning now.

And so I tried to look for, like, how was it that we started to think about language as data, as something that's not only, you know, just electronically encoded to be sort of stored, transmitted, and even manipulated digitally.

digitally and computationally, but something that could be actually analyzed and interpreted and now, of course, generated algorithmically, right? That's something in which the substance of language could be understood through something like enumeration in the same way that we would, you know, calculate the flip of a coin? How was it that language became thought of that way? And so I ended up tracing it back to when language kind of gets brought into the laboratory, the experimental laboratory first.

in experimental and acoustic phonetics in thinking about, well, all of a sudden language as something in the body becomes measurable through these new physiological instruments.

And then that folds directly into this rise of industrial research at Bell Labs in the early 20th century in the US and research into acoustics processing for speech transmission and signal processing, where we really start to see the specification and encoding of language as something distinct from just the transmission of sound.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.