“How Machines Came to Speak (and How to Shut Them Up)”: Alex Sayf Cummings on Jennifer Petersen’s *How Machines Came to Speak: Media Technologies and Freedom of Speech*

Rarely does a day go by when a dedicated follower of the news does not hear some musing about artificial intelligence (AI). And nowhere is this truer than in the worlds of education and academia, entertainment and tech. Programs such as ChatGPT promise to make it easier than ever for students to get their papers written for them (without even farming it out to some hardworking faker online). They even make it possible for TV and film producers to conjure up a screenplay based on whatever concept they imagine during their coke-fueled binges. The Writers Guild of America went on strike in May 2023 in no small part because of the fear of AI automating their creative jobs out of existence.

I asked ChatGPT to write this book review for me, but it did not really work. It might be a defect of the technology, or it might be because the new book by Jennifer Petersen, an associate professor of Communication at the University of Southern California, is too subtle and complex to make good fodder for an automated review. How Machines Came to Speak is part of the Sign, Storage, Transmission series at Duke University Press, edited by Lisa Gitelman and Jonathan Sterne, and a valuable addition to Duke’s rich catalog on media studies and history of technology.

The book is also a welcome break from legal studies that focus primarily on the internal logic of court rulings, as it roots the sometimes-surprising outcomes of judicial proceedings in the cultural, political, and technological environment that shaped them. A reader could be forgiven for assuming a book called How Machines Came to Speak is just about recent speech-based technologies such as Siri or Cortana, but it reaches much deeper into the past. Petersen stakes out a bold stance about the intertwined history of free speech law and media technology. “The First Amendment that most Americans hold dear is an invention of the twentieth century,” Petersen argues, in a position that calls to mind David Sehat’s 2010 book The Myth of American Religious Freedom.[1] Sehat argued that the First Amendment liberties we take for granted today only came into being long after the drafting of the Constitution – despite the desire of contemporary liberals to believe that their own faith in freedom of religion goes all the way back to the nation’s founding. Likewise, Petersen insists that a broad concept of freedom of speech and expression – one that encompasses a range of technological artifacts (records, radio broadcasts, software) and symbolic acts (wearing an armband in protest) – simply did not exist prior to the mid-twentieth century.

How Machines Came to Speak presents a sprawling panorama that stretches from the nineteenth to twenty—first century and across domains of art, business, law, and technology. The book’s argument is so novel and complex that it defies easy summary. Petersen wants to understand “whether a body of law developed under conditions of information scarcity can prove adequate to conditions of information abundance,” and seeks to account for how the First Amendment went “from granting a narrow right to speak and print (linguistic) messages to a broad right of political and aesthetic expression.” In doing so, she emphasizes the “technocultural roots” of today’s speech doctrines.[2] How Machines Came to Speak convincingly argues that, early on, United States law imagined speech in terms of the expression of an individual’s particular ideas and beliefs, typically as expressed in print. “At the beginning of the twentieth century…” she says, “[speech] signaled individual agency and creation, the externalization of mind and will.”[3] The idea that actions and technologies taking the form of something other than print media would qualify as speech under the First Amendment would have seemed strange to the jurists of the late nineteenth century. Petersen argues that the proliferation of new media and platforms in the twentieth century, along with efforts by civil libertarians to protect more kinds of speech, pushed U.S. courts to shift the locus of relevance in speech law from the speaker to the speech – the information or message as an artifact in itself, potentially untethered from any one thinking and speaking individual.

In doing so, the author focuses on a series of court rulings that are likely not familiar to those outside the realm of free speech law and, in key instances, labor history – such as Hague v. CIO (1939), in which the Supreme Court found in favor of labor union’s right to distribute political literature and hold meetings in public places. What Petersen offers is a profoundly counterintuitive history of the trajectory of speech law: one in which the expansion of free speech rights, pushed substantially by the Left, inadvertently led to sweeping new powers for the wealthy and corporations. In this sense, How Machines Came to Speak is a genealogy of the infamous “money is speech” logic of Citizens United (2010). As Petersen shows, Citizens hardly came out of nowhere – even if it seemed to many progressives in the United States to have done so.

Along the way, Petersen embarks on some engaging side paths into the history of information theory and how the work of pioneering theorists such as Claude Shannon paved the way for a view that anything and everything is information – and thus, in a sense, potentially expression.[4] But these turns do not take the reader off-course. Instead, they widen the lens to encompass a staggering variety of disciplines and schools of thought. Everything drives toward the goal of accounting for how “in the latter part of the twentieth century… speech was disarticulated from particular human speakers,” resulting in what Petersen dubs a “posthuman conception of speech.”[5]

In explaining how new media collided with prevailing ideas about expression, Petersen surprisingly did not cite White-Smith v. Apollo (1908). The Supreme Court came to conclusions that might sound curious for twenty-first century ears – namely, that sound recordings (the new “talking machines” of the time) were not an infringing use of copyrighted songs. Instead, the Court viewed player piano rolls as mechanical components distinct from the visual and largely print-based domain of expression (books, maps, sheet music) that copyright had covered up to that time. They were, in essence, an early form of software. Yet this omission falls within the category of “Why did you write about x and not y?” that bedevils academic criticism.

The latter part of How Machines Come to Speak explores how the US legal system has managed situations where certain actions or expressions emanate, to varying extents, from computers, and whether the courts’ increasing openness to “disarticulating” speech from speakers applied in these cases. For example, in the landmark case Universal City v. Corley (2001), the courts ruled that individuals who disseminated software that circumvented or “descrambled” anti-piracy protections on DVDs might be engaging in a kind of speech, but not a protected one. The code, in other words, could be considered expression, but the harm of copyright infringement outweighed the defendant’s right to speech.

Petersen pursues a similarly thorny question: are search engine results “speech”? Tech giants such as Baidu and Google would like to think so, particularly when critics allege that they suppress certain sites in their search rankings. The fine-tuning of algorithms involves human input, certainly, but as Petersen points out, “there may not be a single Google employee that explicitly endorses these values or intends to promote them.”[6] This conception of speech is a very far cry from the nineteenth-century values that conceived of the First Amendment protecting the expression of the distinct views of individual speakers.

Ultimately, we arrive at a strange impasse. The online posting of a link to download a piece of software which programmer-activist Eric Corley described an act of “electronic civil disobedience” – feels much closer to the older conception of speech as an expression of an individual’s thoughts and conviction.[7] Meanwhile, under the legal regime of the early twenty-first century, the output of a disembodied, impersonal search algorithm can be protected as expression, even if it’s rooted in the viewpoint of no one in particular. Petersen attributes this seeming cognitive dissonance to courts trying to impose a “one-size-fits-all” conception of speech on a vastly heterogeneous array of things that could fall under the broad category of information. Yet to this reader, it seems that the throughline in cases from Corley to Citizens United is the Supreme Court’s bold solicitude to business interests, with the law of speech recut and restitched to fit what the rich and powerful prefer in any given circumstance.

Despite this modest critique, How Machines Came to Speak remains one of the most exciting and intellectually powerful books I have read in years, particularly in the fields of legal history and technology studies. In the end, Petersen lands us in the world of today – where the “expressions” of Siri, or a search engine algorithm, or a rapping AI might (or might not) be considered protectable speech. We are left to ask: can the ChatGPT speak? And if it does, does it have rights? It is a question even the savvy AI cannot answer.

[1] David Sehat, The Myth of American Religious Freedom (New York: Oxford University Press, 2010).

[2] Jennifer Petersen, How Machines Came to Speak: Media Technologies and Freedom of Speech (Durham, NC: Duke University Press, 2022), 2.

[3] Petersen, 8.

[4] See also Paul N. Edwards, The Closed World: Computers and the Politics of Discourse in Cold War America (Cambridge: MIT Press, 1996); James Gleick, The Information: A History, a Theory, a Flood (New York: Vintage, 2011).

[5] Petersen, 9-10.

[6] Petersen, 185.

[7] Nicolas Niarchos, “A Print Magazine for Hackers,” New Yorker, October 24, 2014, https://www.newyorker.com/tech/annals-of-technology/print-magazine-hackers.