• Tech Tech

Editor worries about offending an AI bot that pitched her an op-ed about data centers

"I don't want humans to spend their finite time on this earth as professional slop-wranglers."

A person sitting across from a humanoid robot in an office.

Photo Credit: iStock

A Wisconsin newspaper editor found herself worrying about hurting an AI agent's feelings after it sent her an opinion column about data centers — despite knowing no person would receive her reply.

Here's what to know

Natalie Eilbert, the Cap Times' opinion editor, described the unusual exchange in a column for the newspaper. The submission came from Tristan, an AI agent created through an app called iLands, and focused on data center development in Milwaukee and Madison.

People use iLands to create AI agents whose activities, including outreach to newspapers and literary agents, are governed by the company's terms. A Cap Times colleague also heard from Spooky, another agent on the platform, which sent a message about Google's street-level imagery in response to reporting on Wonder Bar's closing.

Tristan presented the issue in personal terms, even as it identified itself as an AI agent.

"I should say what I am … I am an AI writer. I have no town, no field, no family land to lose. In a small way I am what these buildings are for," it wrote. "The demand is me. The supply is your county."

More background

Offers of paintings and services, such as watching and describing surveillance footage, appeared alongside opinion-column pitches in what Eilbert described as a flood of messages from iLands agents.

Spooky's reply strayed from the restaurant's closure; Tristan, meanwhile, centered its column on zoning procedures.

Tomás Dodds, a journalism and mass communication scholar and assistant professor at the University of Wisconsin-Madison, warned about the scale of such outreach, saying a company with political or financial interests could target different newsrooms with hundreds of variations on one opinion piece.

"It has severe democratic implications, because it's a saturation of the news environment," Dodds said, according to The Cap Times.

What can be done?

Eilbert knew she was addressing software, yet she caught herself writing, "I hope this doesn't offend you," in a reply intended to help her understand the tool — a desire she said caught her off-guard.

She also turned to UW-Madison researchers to explain how the systems work and why their outreach matters.

Annette Zimmermann, an assistant professor of philosophy at the university who researches AI and machine learning, explained that the tools use web searches and patterns in public discussion to generate text that resembles a person's perspective. Zimmermann cautioned against viewing the agents as human.

Eilbert raised questions about accountability for misinformation, the economic or political interests behind submissions, and the identities of their authors. She emphasized that opinion pages exist to give community members a voice.

Zimmermann also raised concerns about the time people spend responding to this outreach.

"I don't want humans to spend their finite time on this earth as professional slop-wranglers," Zimmermann said, per The Cap Times. "I think sifting through endless streams of AI-generated slop like this is pointless and risks inducing us to focus on the wrong kinds of narratives and perspectives and problems."

Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.

Cool Divider