A decade ago, BuzzFeed was mostly known for popularising the listicle format and for its bevy of weird quizzes. Hell, there’s even listicles of weird quizzes there to keep users scrolling forever. But with the AI craze flooding the internet, BuzzFeed is trying to carve a new niche for itself in the often-surreal and ever-shifting sands of AI-generated content.
On Tuesday, BuzzFeed launched its new AI-powered “Infinity Quizzes” that were first hinted at in an internal memo last month, seen by Gizmodo. The quizzes are part of the company’s head-first dive into using AI to recreate its business model, shifting away from those inane and mundane human-created BuzzFeed quizzes to something that can be produced even faster with various different pre-established formats. The idea is to create content that “could not exist without artificial intelligence,” according to the company’s press release.
The company is advertising that quiz writers are still involved in creating the idea, and the framework. The “Buzzy” AI, which was created using the ChatGPT-creator OpenAI’s API, is supposed to combine both inputs from the quiz creator and user to generate a personalised response. A BuzzFeed spokesperson further told us that not all of the quizzes are going to be AI-based and that writers can also choose how they want to craft the quizzes being created.
So there’s a couple early versions of these AI-powered quizzes on display. One of them gets the AI to create a breakup text for you, which I personally found very demeaning when it tried to insinuate my ex just didn’t respect Nelson Mandela enough. Other quizzes are much more innocuous, like one to create a personalised “cinematic universe.” I chose one where AI will design the perfect “houseplant soulmate” for me. OK, here goes. The AI thinks my “plant soulmate” is a “Fritz Cactus.” That’s good to know, I guess, even if I can’t actually find any example of a plant called “Fritz Cactus.” I checked on the U.S. Department of Agriculture’s Natural Resources Conservation Service website and found nothing. I also checked on the World Flora Online database and found no examples of that specific cactus. So perhaps it’s just a cactus named “Fritz.”
I’m sure I wouldn’t mind putting a cactus on my shelf, but more than that, the plant will be waking me up with a “playful nudge.” Of course, I don’t want to be nudged by a spiky cactus, but I’m not sure the AI cares. We’ll do yoga together, apparently, even though I don’t do yoga and don’t plan to. Then on a hike through the mountains, the cactus will break in and ask me “Can you imagine if we could fly like Aang?” I told the AI Avatar: The Last Airbender was a huge part of my childhood. This, I now see, was a huge mistake.
“Sure can,” Kyle says with a grin. “We could soar higher than ever before. I loved his carefree spirit and fiery determination.”
“I always had a thing for Fire Nation Zuko,” Fritz winks playfully and Kyle laughs.
This is the first time I have ever shivered when using an AI. I don’t like the idea of anybody creating a fanfic where I’m the main character, let alone what an AI thinks a quasi-romantic conversation about Avatar would look like.
So the AI creates some pretty strange answers, but more than that, is it capable of just making up plant species out of whole cloth? Knowing that Microsoft’s Bing AI, which was developed in part by OpenAI, can gaslight users about the current year. BuzzFeed is using this as an opportunity to court advertisers as well. The plant-friend survey is a branding effort with Scott’s, the makers of Miracle-Gro. How good the system is at keeping things clean will be a big factor determining how well those sponsorships go.
Since the start of 2023, tech companies big and small have been rapidly trying to integrate AI into their systems. Dating app OKCupid has been piloting AI-generated dating prompts. And of course, Microsoft’s Bing and Google are both desperate to be the first to shove ChatGPT-like interface into their search engines, even if those chatbot systems are known for making up facts.
Indeed, that’s just one of the concerns with this kind of AI generated response to a quiz. My response was just plain weird, but what if it shares some kind of falsehood? BuzzFeed said the system has been through several phases of “testing, learning, guardrailing, and retesting” it still
“marks the early stages of experimentation and conversation with our audience.”