Janelle Shane works with artificial intelligence, experimenting with training neural networks to generate data that may or may not resemble anything in the real world. She posts the funniest results on her blog AI Weirdness. Shane put out a call for mid-century American recipes (the kind that rely on Jell-O and other processed foods) last week to see what a neural network would learn from them. She received over 800 recipes, and used them to train an artificial intelligence algorithm to generate its own recipes. The results are just as horrifying as you might imagine.

WHAT HAS OCCURRED CANNOT BE UNDONE

I have trained a neural net on a crowdsourced set of vintage jello-centric recipes

I believe this to possibly be the worst recipe-generating algorithm in existence pic.twitter.com/cwQwOpUNDv

— Janelle Shane (@JanelleCShane) February 7, 2020

Shane admits that this algorithm has previously trained on internet text to learn how to read, and she was surprised at the random bits of information it had retained from, say, fan fiction. Read about the experiment at AI Weirdness, which contains ten of the recipes. There are quite a few more of them at Twitter.

[via Mashable]

Tags: algorithm, artificial intelligence, Janelle Shane, neural network, recipe