Remember the horror movie The Fly, in which a brilliant scientist comes up with an groundbreaking technology -- only for his good intentions to head south when he accidentally gets genetically spliced with a housefly, turning him into a hideous monster? That's kind of the vibe we get hearing about a project that aims to teach an AI to hold a polite conversation by feeding it message threads from Reddit. Nonetheless, that's the working theory held by researchers at OpenAI, the nonprofit Artificial Intelligence company backed by Silicon Valley heavy-hitters like Elon Musk, which lists its mission as "to build safe AI, and ensure AI's benefits are as widely and evenly distributed as possible." All kidding aside, the impressive-sounding project focuses on creating a deep learning neural network able to build up a "probabilistic understanding" of conversation, and maybe even construct enough of a language model to be able to communicate itself. To help with this, OpenAI is employing a DGX-1 supercomputer developed by Nvidia.