Saturday, 6 January 2018

category: athletes with maps

The intrepid team at Lewis & Quark have run some notable experiments with neural networks in the past year (previously here, here, and here) and begins 2018 with an equally insightful drill that invited their AI familiar to come up with plausible Wikipedia articles—based of course on the contributions of human encyclopaedists, and the results were predictably weird though revealing about our own editorial conventions.
Parallel to the headings that the machine will generate, we learn that Wikipedia has curated a collection of submitted articles that humorously failed to adhere to the site’s standards and were consequently never published, which could have been the product of the machine itself but Wikipedia’s rejects could not hold a candle (we think) to some of the ones in the computer-generated sandbox: Monster Diseases, Tire bear (country), Near Dogs and Tree Wars. Be sure and visit Lewis & Quark on Tumblr at the link above for more of computers trying and subscribe so you’ll never miss a post.