There’s no arguing that artificial intelligence plays an important role protecting the and alike. But when it comes to naming video games, AI seems to have a special hidden talent.
AI programmer Andrew Eckel decided to train his recurrent neural network (RNN) — a machine-learning system — to create some fictional video game titles of its own.
“I downloaded an implementation of an RNN that was already set up, and I trained it on a list of 149,664 video game titles I found online,” Eckel wrote in his blog on July 1.
Eckel’s RNN came up with a whopping 5,000 imaginary titles ranging from plausible video games like Alien in the Fires, Lots of Doom, Pirate Devil, Lake Fighter, Blood Waster and Spy Blast Wars to the downright bizarre, like Worms of Garfield: Shadows of Championship, WWE Mink Blood (Conquest & A Might) and Pac-Man Scenery.
Some of the more unusual game titles might offer a little insight into what AI thinks about human sports. For example, it seems to realize that the sport of golf could use some livening up with titles like The Putt Raider, Hard Boxing Golf and Mega Golf Hockey.
“I felt like I was shopping from a mail order catalog,” Eckel wrote. “I think that’s the best way to enjoy these titles. Imagine it’s 1990. There are a bunch of new games on the market, but there’s nowhere to read reviews. Would you spend money on one of these games, sight unseen?”
Personally, I would pay good money to play Dark vs. Frog — a game that sounds like it pits an innocent frog against the forces of darkness. Perhaps Frogger meets Slenderman?
Also, if anyone decides to make the AI game idea called Galaxy in Love, let me know. We could all use more games that remind us of the kinder side of the universe.
Special Reports: CNET’s in-depth features in one place.
Culture: Your hub for everything from film and television to music, comics, toys and sports.