This neural network has got some fascinating Halloween costume ideas
The machine-learning algorithm was fed with a dataset of an estimated 4,500 costume ideas to create new ones.
After producing more than 100 fascinating craft beer names with the help of artificial intelligence, research scientist Janelle Shane is again offering some help – this time by creating first-of-its-kind machine learning algorithm that gives out some really interesting Halloween costume ideas.
According to a report in Business Insider, Shane tried simplifying the costume-selection task by feeding a neural network with over 4,500 names of Halloween costumes.
She noted on her blog that the massive dataset was crowdsourced from the internet, which included classics as well as technically challenging names such as, Invisible Pink Unicorn, Whale-frog, Glow Cloud, Lake Michigan, Toaster Oven, and Garnet.
The information was then processed by the algorithm, which pulled out a bunch of new and interesting costume ideas. It is important to note that instead of analyzing the meaning of the words, the neural network learned word patterns from the scratch, letter-by-letter, to create new ones.
It initially produced some common names, but got better and pulled out names like Punk Tree, Disco Monster, Spartan Gandalf, Starfleet Shark, A masked box, Martian Devil, Panda Clam, Party Scarecrow, Potato man, Dragon of Liberty, Pumpkin King, and more. The complete list of costume ideas can be accessed through her blog post.
Detailing the algorithm to Business Insider, Shane said, "I would argue that the Halloween costume neural network is actually right up there at coming up with creative things that humans love. It can form its own rules about what it's seeing rather than just memorising."
A few months back, Shane's neural network did a tremendous job at giving away a plethora of novel craft beer names. One of those – The Fine Stranger – has actually been given to a beer.