Can’t Come up With a Good Costume? @JanelleCShane has Some AI Generated Ideas! #costume #halloween #MachineLearning #ArtificialIntelligence #GPT2
Haloween is right around the corner, do you have a costume yet? Are you looking for something fun and VERY unique? Then Janelle Shane (@JanelleCShane) has the costume for you. For the last three years, Shane has used various machine learning algorithms to come up with some pretty memorable costume ideas. In 2017, it was the char-rnn neural network that learns words ‘from scratch, letter by letter’. The char-rnn costumes were a bit weird including “Grankenstein”, “Scorbs Monster” and “Sexy Avical Ster of a balana Aly” but fun nonetheless.
In 2017 I tried to generate Halloween costumes with a neural net that didn’t have any prior knowledge of English words. Its costumes often made no damn sense.
In 2018, Shane teamed up with the New York Times (NYT) to create some more ‘artificial’ costumes. This time using the textgenrnn which can imitate text. The idea was to predict new costumes people would come up with.
…we turned to a machine-learning algorithm called textgenrnn that can learn to imitate text. Its author, Max Woolf, designed it as a blank slate ready to learn any kind of text; the text we gave it to imitate was a list of 7,182 costumes that people sent to aiweirdness.com over a year.
The NYT post was beautifully illustrated by Jessia Ma and brings the ideas to life, the costumes include classics like, “vampire chick shark” , “dragon ninja” or “cat witch”. It turns out that there is actually a sexy dragon ninja costume out there so I think we’re getting close!
Earlier this week, Shane announced her newest line of AI-generated costume ideas using the GPT-2 algorithm.
Now, as of 2019, there are much more powerful text-generating neural nets around. One of these is GPT-2, trained by OpenAI on a huge dataset of text from the internet.