The Related Press revealed requirements at this time for generative AI use in its newsroom. The group, which has a licensing agreement with ChatGPT maker OpenAI, listed a reasonably restrictive and common sense list of measures across the burgeoning tech whereas cautioning its workers to not use AI to make publishable content material. Though nothing within the new tips is especially controversial, much less scrupulous retailers may view the AP’s blessing as a license to make use of generative AI extra excessively or underhandedly.
The group’s AI manifesto underscores a perception that synthetic intelligence content material must be handled because the flawed software that it’s — not a alternative for skilled writers, editors and reporters exercising their greatest judgment. “We don’t see AI as a alternative of journalists in any method,” the AP’s Vice President for Requirements and Inclusion, Amanda Barrett, wrote in an article about its method to AI at this time. “It’s the accountability of AP journalists to be accountable for the accuracy and equity of the data we share.”
The article directs its journalists to view AI-generated content material as “unvetted supply materials,” to which editorial workers “should apply their editorial judgment and AP’s sourcing requirements when contemplating any info for publication.” It says workers might “experiment with ChatGPT with warning” however not create publishable content material with it. That features photos, too. “In accordance with our requirements, we don’t alter any components of our images, video or audio,” it states. “Due to this fact, we don’t permit the usage of generative AI so as to add or subtract any components.” Nonetheless, it carved an exception for tales the place AI illustrations or art are a narrative’s topic — and even then, it must be clearly labeled as such.
Barrett warns about AI’s potential for spreading misinformation. To forestall the unintentional publishing of something AI-created that seems genuine, she says AP journalists “ought to train the identical warning and skepticism they might usually, together with making an attempt to determine the supply of the unique content material, doing a reverse picture search to assist confirm a picture’s origin, and checking for stories with comparable content material from trusted media.” To guard privateness, the rules additionally prohibit writers from getting into “confidential or delicate info into AI instruments.”
Though that’s a comparatively common sense and uncontroversial algorithm, different media retailers have been much less discerning. CNET was caught early this yr publishing error-ridden AI-generated financial explainer articles (solely labeled as computer-made if you happen to clicked on the article’s byline). Gizmodo discovered itself in an identical highlight this summer time when it ran a Star Wars article full of inaccuracies. It’s not arduous to think about different retailers — determined for an edge within the extremely aggressive media panorama — viewing the AP’s (tightly restricted) AI use as a inexperienced mild to make robotic journalism a central determine of their newsrooms, publishing poorly edited / inaccurate content material or failing to label AI-generated work as such.