Twitter presented last Friday (30) a contest to seek improvements in the algorithm for automatic cropping of images on the social network.
The idea is to reinforce with external help the Artificial Intelligence system, which makes adjustments to the photos posted on the platform without gender or ethnic bias, for example. Very similar situations were detected first from a complaint and, later, in internal Twitter studies: especially in photos with more than one person, the system tended to highlight the faces of white people and women.
For this, the company will share the codes used to generate the image clipping and evaluate the changes made by the community using a variety of methods.
in search of the best
The contest will take place along the lines of security competitions, in which hackers are challenged to break into systems and applications. The announcement of the winners will even take place at DEF CON AI Village, on August 8th, with the presentation of the best works.
The best algorithm change will receive a prize of US$3,500, but other categories will also be awarded, such as Most Innovative and Most Generalizing (that is, it can be applied to other categories). Interested parties must fill out a form and read the entire regulation on the HackerOne platform.
So far, as an interim measure after the official apology, the service has started to display entire images in mobile apps.