Twitter plans to change the way photos are cropped based on racial bias
Twitter plans to change the way photos are cropped based on racial bias

Twitter announced on Friday that amid concerns that the company's machine learning algorithms are racially biased when it comes to cropping images, and prioritizing white faces over black faces, the company will cease its image cropping function. Change its location. The Road.

The company's chief technology officer and design director (Dantley Davis) explained in an article how the company tests the model for racial or gender bias before implementing it. However, Twitter has not announced how these tests are being conducted for external analysis of what the company calls "oversight".

To address this issue, Twitter is currently conducting "further analysis to make our testing more rigorous, advocating for sharing our results, and looking for open analytical methods so that others can help us do our jobs," Twitter company officials said. .

The company is also making more concrete changes to its website to ensure the images are displayed the way users want them to. In particular, Twitter promises to shift away from focusing on using machine learning-based technology to crop images, and instead develop tools that will show users what an image in a post looks like as they type. Tweet. The company also said: It has already started trying new options for cropping and previewing images to give users more control.

Going forward, Twitter promises to develop a "what you see is what you get" policy so that the image that a user attaches to a Tweet is the one that other users see when they view it. Twitter. Twitter needs to solve some advanced issues like: Images Too Long or Too Wide. However, the goal is to ensure that users know how their photos will appear on the website from scratch.

The company switched to a machine learning-based cultivation system in 2018, which uses a neural network to preview images based on the "celebrity" crop. Unlike the company's previous system that focused on faces, the current model aims to predict where the average person will appear for the first time when looking at the full picture and focus on recognizing the algorithm as "the most interesting part of the picture."

There is no information on when these changes will be published to Twitter's photo cropping system.

Previous Post Next Post