Twitter has moved to introduce a radical change to its bug bounty programme by announcing the tech industry’s first-ever competition to determine bias in its artificial intelligence algorithm. The event will be part of the hacker convention DEF CON AI Village this year.
In a blog post, the social media giant said it was difficult to find bias in machine learning models and it is common for companies to discover ethical harms, although unintended, only after they have reached the public.
In May, Twitter had announced its approach to identify bias in the company’s saliency algorithm — image cropping algorithm — and made the code public. “We want to take this work a step further by inviting and incentivizing the community to help identify potential harms of this algorithm beyond what we identified ourselves,” Twitter said.
The company added in the blogpost that the manner in which the hacker and research communities had helped establish best practices to identify and remove vulnerabilities to protect the public was an inspiration. It added that the company wanted to cultivate a similar community that focused on ethics of machine learning and help identify issues on a far broader scale than what it would be able to on its own. “With this challenge we aim to set a precedent at Twitter, and in the industry, for proactive and collective identification of algorithmic harms.”
The San Francisco-based company has also announced cash prizes of $3,500 (approx. Rs 2,60,242), $1,000 (approx. Rs 74,369), and $500 (approx. Rs 37,184) for first, second, and third places, respectively. It has also announced cash prizes of $1,000 each for most innovative and most generalizable solutions. The winners of the competition will be announced during a workshop hosted by Twitter at DEF CON AI Village on August 8. The contest is open for entries till August 6, 2021 (11:59PM PT or 12.29 pm on August 7 in India).
