Narayanan: AI chatbots becoming a 'truth oracle'

Narayanan framed his talk as "as much about the continuing effects of social media on newsrooms as it is about AI." He argued that the newsroom bundling reporting, analysis, fact-checking and distribution was a "historically contingent" development, not a permanent one — the ad-supported penny papers of the 1830s broadened what counted as news, objectivity arose in response to World War I propaganda, and paid reporters and interviews "had to be invented." He quoted Clay Shirky's line that "society doesn't need newspapers. What we need is journalism," and said Paul Starr's warning about unfunded public goods like accountability has been proven right "in many ways," though philanthropy has softened the blow.

His broader claim is that piece by piece, specialists do each job more cheaply than a newsroom can, and AI is hastening that. He described the internet as having "democratized distribution" while AI is now "democratizing production," making it possible to produce first-rate journalistic content with a one- or two-person team — "not so much a decline of news" as "the migration toward a bunch of other institutions." He also argued that agenda-setting by the media has backfired, saying the Washington Post's Watergate role was "an unalloyed good" but that similar efforts in the Trump era have fed polarization and hastened the collapse of trust.

The stakes hinge on whether publishers can close the power gap he describes — he says publishers need chatbot distribution more than AI firms need any one outlet, so licensing deals underprice journalism. His proposed remedy, democratic input into algorithmic tuning with journalists at the table, would have to be "broad-based" and not "perceived as partisan" to work, which is a demanding bar. If it fails, the outcome he leaves mostly unmentioned may prove the default: journalism as an organized, aggregated industry may be a phenomenon of the last two centuries that technology and economics are evolving away from.