Eidnes developed algorithms that logically pick words, then trained the system using some "2 million headlines scraped from Buzzfeed, Gawker, Jezebel, Huffington Post and Upworthy." He dryly noted the titles don't need to make sense, because "it's not clear that clickbait needs to have any relation to the real world in order to be successful." As proof, he cited this gem from Buzzfeed: "22 faces that everyone who has pooped will immediately recognize."
After he crunched the results multiple times on a high-end NVIDIA GTX-980 graphics card, the resulting headlines are surprisingly good. "Most of them are grammatically correct, and a lot of them even make sense," said Eidnes, citing examples like "Romney camp: 'I think you are a bad president,'" or "Biden responds to Hillary Clinton's speech." (Others are hilarious nonsense, like "This guy thinks his cat was drunk for five years, he gets a sex assault at a home.")
As final proof of the concept, Eidnes built an entire machine-written clickbait website. His algorithms chooses photos from Wikimedia Commons based on keywords from his clickbait headines. The article body is also seeded with headline words, and the author's names are picked by his neural network. The glorious result is Click-o-Tron, a site that generates a new article every 20 minutes. "This gives us an infinite source of useless journalism, available at no cost... forcing other producers of useless journalism to produce something else." We'll be shocked if that happens next.