This Algorithm Generates Crazy and Terrifying Disease Names
A new neural network makes lists of diseases we thankfully don’t have
/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer/ff/d7/ffd78588-8a62-4c50-9976-de5bfd61ed88/doctor_consults_with_patient_7.jpg)
Know someone with a case of metal ringworm or eye stots? Chances are you don't, because these "disease names" were made up by a computer.
"Science fiction writers and producers of TV medical dramas: have you ever needed to invent a serious-sounding disease whose symptoms, progression, and cure you can utterly control? Artificial intelligence can help!" research scientist Janelle Shane writes in a blog post about her latest project, which uses the power of AI to create disease names that range from ridiculous to frightening.
Shane often tinkers with artificial neural networks to create recipes, craft beer names and even pickup lines with humorous results. As Kaleigh Rogers writes for Motherboard, Shane's latest project is disease names. To complete this task, she trained a network with a set of 3,765 names for diseases using an online database compiled by one of her blog readers. Unlike standard software that is programmed to complete certain tasks, artificial neural networks are loosely modeled on the human brain.
Neural networks are composed of a series of nodes that function together to generate an output. Each node reacts from a combination of inputs from other nodes. Because of this, the network can be "taught" to create unique responses based on information fed into the system. Working with these networks has resulted in everything from a "new" Game of Thrones book to music and art creations.
Early in the training of Shane's new system, the disease algorithm's made-up ailments were pretty bad, Shane writes on her blog. It's a bit like the initial products of anyone learning a new skill. These early names included allergic antibody, sexurasoma ear and the ominously vague complex disease.
As the system chugged along, the results slightly improved, Shane says, but they're still pretty laughable. Seal breath sounds unpleasant, but survivable, while machosaver seems like a bad fitness product, not an illness.
When it competed its training, Shane says the algorithm essentially settled on creating four different kinds of disease names. The first group were all based on body parts (like ear or lower right abdomen) or the names of other diseases (cysts of the biles and swollen inflammation). The second category, according to Shane, is "reasonably convincing." This includes things like hemopheritis or osteomaroxism. The third category was reserved for names that are implausible but serious like cancer of the cancer or super syndrome. Finally the fourth category was comprised of sexually explicit disorders and those with "unprintable four-letter words," Shane writes.
She concludes: "They usually sound ludicrous, and entirely uncomfortable, all at the same time." But if you really want to know them, Shane provides a link on her blog to gain access.