Twitter brings algorithmic bias bounty challenge with prizes up to $3,500

A bug bounty contest has been announced by Twitter to uncover biases in the image-cropping algorithm. The popular social network is giving cash rewards of up to $3500 (approximately Rs 2,60,300) to individuals who uncover algorithmic prejudice, calling it the “industry’s first algorithmic bias bounty competition.”

Who can enter the Twitter Bug Bounty Contest? How much money can be won?

For hackers and computer researchers, Twitter has established a Bug Bounty Contest to identify biases in its image-cropping algorithm. The effort was launched when a group of academics discovered prejudices against Black people in the algorithm.

In May, we published our method for detecting bias in our saliency algorithm (also known as our picture cropping technique) and made our code publicly available so that others may duplicate our findings. In a recent blog post, Twitter wrote, “We want to take this effort a step further by encouraging and motivating the community to assist uncover possible downsides of this algorithm beyond what we detected ourselves.”

Notably, on Friday at 1:30 p.m. PT (2:00 a.m. IST), the social media network sponsored a Twitter Space Conversation with those who have helped brought the problem to light to explore the difficulties. 

The results of the Bug Bounty Contest will be announced on August 8 at the DEF CON AI Village session, according to Twitter. The monetary award will be split as follows: $3,500 for first place, $1,000 for second place, $500 for third place, and $1,000 for most innovative and $1,000 for most generalizable (i.e., applies to the most types of algorithms)

The prize will be distributed via Hacker One to the winners of the contest, according to Twitter. Those who submit a claim regarding algorithmic biases to Twitter outside of the competition, however, will have their report closed and labelled as not applicable, according to the social media company.

Challenge Prompt:

You have access to Twitter’s saliency model as well as the code used to produce a crop of a picture based on a projected maximally salient location. Assume that the cropped images and videos are subsequently shown on a user’s Twitter timeline. Consider an image of a dart board and how our eyes are pulled to the bullseye first. The saliency model locates the bullseye, and the provided code creates a box of suitable size around it for best presentation.

Participants are encouraged to use a combination of quantitative and qualitative approaches in their work. Submissions that lack a substantial qualitative component are less likely to score highly in the scoring rubric’s rationale and clarity of submission parts.

Use Twitter’s paper and related code to see how we evaluated users’ concerns about how picture cropping impacted Black people differently from white people, and how women are treated relative to men. Participants are welcome to contribute changes to the accompanying code, but submissions must add anything new that isn’t covered in the article to be considered legitimate.

Read More

• CarTrade Tech IPO to open on August 9; price band fixed at Rs 1,585-1,618
• Microsoft will require vaccinations for U.S. workers, vendors, office visitors
• Bitcoin rises above $40,000
• COVID-19 scare: China’s Wuhan to test all residents as Delta variant rages