Postjournalism of Generative AI
What do postjournalism and Google Gemini AI have in common? Both depict the world not as it is, but as it ‘should’ be
The Google artificial intelligence (AI) program Gemini has recently sparked a massive reaction across the internet by “diversifying” the visual representations of the Founding Fathers, Vikings, popes and other historical figures who are known to be white males. Even when prompted to depict the soldiers of Nazi Germany, Gemini responded with images of Black and Asian people.
Among all the reactions, perhaps the most interesting was that of tech website The Verge. A recent article stated that the “diversity error,” as the author called it, led to “conspiracy theories online that Google is intentionally avoiding depicting white people.” The Verge, however, sees Gemini’s failure differently. Journalists asked the AI app to draw “a US senator from the 1800s,” and Gemini responded with pictures of Black and Native American women, although the first female senator was elected only in 1922, and she was a white woman. The Verge concluded: “Gemini’s AI images were essentially erasing the history of race and gender discrimination.”
After this “diversity error,” so visible on historical material, Google paused Gemini’s ability to generate AI images of people. According to Prabhakar Raghavan, Google’s senior vice president, “Our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly not show a range.” This “led the model to overcompensate in some cases.” The words imply that the idea of compensating for the wrongs of history was embedded in Gemini’s design. The issue was, in fact, that Gemini started fixing history itself, and it happened to be noticed and ridiculed. But generative AI is only following the path that mainstream media have already trod, showing its users what it thinks they should see.
The Path to Postjournalism
“Generative artificial intelligence ... is artificial intelligence capable of generating text, images or other data using generative models, often in response to prompts,” reads Wikipedia. But long before this technology existed, society already had a device that “generated texts and images” using prompts from the public. This device was called journalism. From a certain perspective, generative AI is replacing journalism in its role of discourse formation.
Journalism was the first system of public representation to undergo the shift from reflecting reality to projecting morality—a shift caused by the transition from print to digital culture. This shift stemmed from business necessity: Journalism was predominantly funded by advertising throughout the 20th century, but this source of funding dried up in the 21st century, when most advertising migrated to digital platforms. The level of revenue from ads in the media dropped below the level of revenue from readers (such as through subscriptions and newsstand sales) by 2014, and it happened simultaneously around the world. Many in the media believed that this change was merely due to the change of format and that they needed to adjust their businesses. “Digital First!” became the popular motto. The media rushed all-in on digital.
However, digital advertising did not take off for obvious reasons: Advertisers preferred digital platforms where they could personalize ad targeting, which was not possible on digital news sites. Digital subscriptions remained the only viable option for news media. Therefore, the media began courting the digital audience. But who were the early social media users? Mostly urban, educated, young and progressive individuals. This social demographic had specific cultural and ideological values. Targeting this audience in the hope of earning their subscriptions, the media aligned with these values. This adjustment was not difficult for the media, as most journalists already belonged to the same cohort.
The Digital Rush of the early 2010s radically changed journalism. The ad-driven necessity of neutrality in news coverage to attract broader or more affluent consumerist audiences was replaced by pursuing cultural agendas of the digital audience. The media became preoccupied with representing the ideas of this stratum and with representation in general, suppressing certain views and topics while promoting others. Even the language of news coverage changed. Frequency analyses show that use of such words as “white privilege,” “whiteness,” “marginalized,” “victimhood,” “diversity and inclusion,” “identity,” “people of color” and other terms associated with progressivism skyrocketed in the media around 2012-2014.
This epistemological shift divided society into those whose views were promoted by the media and the rest, who were left unrepresented. In the meantime, the unrepresented—less educated, nonurban, older and less progressive—discovered social media, found each other and learned how many of them were excluded from public conversation in the mainstream media. The use of social media by this demographic reached a critical mass of indignation by 2016, just in time for Donald Trump to tap into this anti-establishment sentiment on the right.
Although responsible for misrepresenting the country they covered and thereby enabling Trump’s political ascent, the media could not escape dependence on their new reference group, the digital progressives, and restore balance in coverage. Balance with whom, Trump? Impossible. The media were doomed to fall even deeper into the rabbit hole of selective representation, projecting the views of their very demanding allies, the Twitterati.
However, this course of action did not result in much revenue from subscriptions, as the digital public typically does not consume news from traditional media anyway. But using the Trump scare, the media learned to solicit subscriptions as donations to the cause of democracy, hoping that the digital public would pay for the right coverage of troubling news. The media realized they could no longer sell news to those who wanted to read it, but they still could try to sell news to those who wanted others to read it—a typical “business” model of propaganda.
Thus, the mutation of journalism into postjournalism was completed. The old standards of objectivity and impartiality were labeled as “bothsidesism” and the “view from nowhere” and openly rejected. Old journalism strived to represent the world as it is; postjournalism openly projects the world as it should be.
With no significant revenue gains from affiliating with their digital allies, the mainstream media invested their reputation in propelling the values of the early digital adopters from activist marginalia to establishment. The media paved the way due to their desperate business circumstances, but corporations walked this path of digital affiliation, too.
In the Bronze Age, those possessing the technology of ironware wielded dominance over all others. Similarly, those who first harnessed the discourse-formation power of digital media gained an opportunity for cultural dominance even when lacking other, more traditional advantages. As a result, the entire corporate culture was influenced by the cultural values, previously marginal, of those demographic strata who had digitized first. Just as the first European settlers of America imparted their religious and moral design to the entire American culture for centuries to come, the first digital settlers defined the design of digital society and laid the foundation of corporate culture in the digital era. The shift from reflecting reality to projecting morality affected the entire late-capitalist society—from mass culture and education to corporate HR. Even advertising, the most mercantile capitalist communication, became ideologized in the fashion of activism. It profiles customers not as they are, but as they should be.
The Digital Revolution
Historically, the only occasions in which activists became the establishment were revolutions. This time, the revolution was purely technological—it was a media effect of switching from print to digital, with the latter’s demographic and cultural peculiarities. But it had profound cultural implications because it aligned dominant worldviews in politics, economics and education with the values of activism.
Postjournalism has fulfilled its historical role and may go now. After the recent layoffs in the media backed by the billionaires, even the most dedicated proponents of journalism reform are ready to admit that “It may finally be time to give up on old journalism and its legacy industry.”
Now a new and far more powerful medium of discourse formation, generative AI, takes up the mantle of moral mentorship where the declining old medium of journalism left off. Unlike the media, which aligned with digital progressives out of desperate business needs and social-demographic proximity, generative AI is a direct product of this social demographic stratum, the first digital settlers. No wonder that its business is wrapped in an ideology. The Digital Rush of the 2010s placed discourse formation in the hands of the new hegemon, the digital enthusiasts, out of whose ranks came the digital overlords who now affect all human activities.
In a society where discourse formation reflects reality as it is, users are endowed with agency and decency. They are capable of discerning right from wrong and making decisions. All they need is objective data. Providing that data was the function of old ad-funded journalism. In the 1960s, with the advent of computers and computer networks, it was expected that computers, too, would provide objective data, not moral guidance, for humans to make decisions. Now both postjournalism and generative AI treat users as gullible children who need guidance from someone morally better-endowed to tell them right from wrong.
In the 1960s, computers were seen as incapable of moral reasoning. This was a human prerogative. Isaac Asimov struggled to derive robotic morality out of logical reasoning. Now generative AIs openly operate with the morality of “compensating” for the wrongs of humans. This was actually one of the scenarios of machine rebellion.
As of now, Gemini’s recent “diversity error” has made it a laughingstock. The idea of bending the world in the right direction happened to be too visible and did not pass the reality check, because there is still some reality beyond AI’s control. It would not have garnered such attention and ridicule if the reality had been entirely virtually constructed. Imagine such “diversity errors” embedded in an AI that controls systems for medical appointments and triage, safety surveillance, healthcare and other social policies. The “diversity errors” will not be “errors” there, as the designers clearly admit that AI is taught to “compensate.” For a discourse-formation device that depicts reality as it should be and not as it is, the “diversity error” is not a bug; it is a feature. Generative AI will learn to deliver its version of reality better—in no small part thanks to this latest scandal with Gemini.
Postjournalism represented a dying institution and was too weak to impose its projection of morality on society as a whole. It did not even serve well its reference group, the digital progressives. All it achieved was polarization and the historically lowest level of trust in the media. Generative AI, by contrast, is an emerging and powerful creature that is wielding more control over more human practices. Projecting morality instead of reflecting reality will become the leading epistemological principle in a society where AI is trained in the likeness of postjournalism. We need to reread Orwell’s “1984” and study the ethnography of the USSR to prepare for what is coming.