“I don't think our readers care, to be totally honest."
Tech finance publication VentureBeat is using OpenAI's Bing Chat to assist in editing and writing articles. And the folks in charge of the effort really don't seem too shy about it.
"I think of it as like having another person on the team," VentureBeat's editorial director Michael Nuñez told Bloomberg, arguing that large language model (LLM)-powered tools of the kind can do tasks like information aggregation faster than humans can, seemingly to argue that AI tools simply help to speed up the journalistic process.
"I don't think our readers care, to be totally honest," he added. "Tweet that out, if you want."
Heart-Shaped Black Box
Considering that VentureBeat, which boasts the slogan "transformative tech coverage that matters," covers AI innovation pretty heavily — and is also set to release a special issue all about what they're calling "the quest for AI Nirvana" — it's not terribly surprising to see them experiment with the buzziest tech in Silicon Valley.
Still, Nuñez's enthusiasm certainly comes with some caveats.
For one thing, the editorial director told Bloomberg that as long as a writer is only using "sentences and fragments" from chatbot-generated text that are independently verified to be "truthful," he doesn't think that AI-assisted articles should have to be overtly marked as such.
But publications like CNET, Bankrate, and Men's Journal have all made similar fact-checking promises, and they've all churned out some pretty terrible AI-assisted content. And to that end, it's generally just good practice to let people know where info is coming from, be it from a human or a chatbot or both — especially considering that chatbots generally either fail to provide sources or just make them up on the spot. AI plagiarism is also of concern, and a tool that reliably senses whether content is AI-generated has yet to hit the market.
Regardless of the potential downsides, however, Nuñez says that the hesitation to allow AI into the news cycle at scale will be short-lived.
"I'm not trying to take a press release, feed it through the system, and get an article, although I can probably do that," Nuñez told Bloomberg, reportedly adding that he thinks that, pretty soon, no one will question the use of AI as an assistive tool.
"If you're taking sentences or clauses or small fragments of ideas from the large language model — in my view, we're going to be discussing it for six months," he continued. "I guarantee, in two years, it's not going to matter."
More on AI journalism: Tabloid Fires Editor for AI-Generated "Interview" with Injured Celebrity