Twitter Is Giving Away Thousands Of Dollars To Fix ‘Racist’ & ‘Sexist’ Algorithm
Image via Photo 133465603 © Michele Ursi | Dreamstime.com
Humans are to blame for the prejudice rooted in artificial intelligence, and humans will fix it. After much struggle, Twitter is now offering bounty rewards to researchers and hackers who can identify bias in its algorithm, as well as offer solutions to fix it.
A cash reward of US$3,500 will be presented to the winner, with prizes of US$1,000 and US$500 set aside for runners-up and other participants with meaningful entries.
A new blog post announcing the competition, running through August 6, invites “bounty hunters” who can discover “potential harms of this algorithm beyond what we identified ourselves” to not only abolish gender and racial bias in Twitter’s image-cropping feature but also help resolve a broader challenge of “algorithmic harms” faced by the industry. As such, there are also openings for ‘Most Innovative’ and ‘Most Generalizable’ (applicable to most algorithms) ideas.
Twitter acknowledges that the rewards were inspired by criticism around its image-cropping AI, which seemingly favored white people over Black figures. For instance, the tweet below includes photos of both former US president Barack Obama and former Senate Majority Leader Mitch McConnell, but only shows the latter. The social network overcame the issue by killing the photo-cropping tool in May.
Trying a horrible experiment...
— Tony “Abolish ICE” Arcieri 🦀 (@bascule) September 19, 2020
Which will the Twitter algorithm pick: Mitch McConnell or Barack Obama? pic.twitter.com/bR1GRyCkia
The challenge’s entries will be reviewed by a panel of judges—including Harvard computer science PhD student and research scientist Ariel Herbert-Voss, Ford Foundation Tech Fellow Matt Mitchell, network security expert Peiter “Mudge” Zatko, and data scientist Patrick Hall—on “both quantitative and qualitative” aspects of the research. Winners will also be invited to present their findings at Twitter’s DEF CON AI Village workshop in August.
“We are always looking for ways to improve our algorithms to reduce the unintentional harms they can cause on the people who use Twitter, and society at large,” Twitter noted in a tweet. “We can’t solve these challenges alone, and we want your help.”
Here’s the TL;DR.
— Twitter Engineering (@TwitterEng) July 30, 2021
What:
Our first-ever algorithmic bias bounty challenge. We’re openly sharing a Twitter algorithm so adversarial minds can do what they do best. The winners will receive cash prizes.
[via Insider, cover image via Photo 133465603 © Michele Ursi | Dreamstime.com]
Also check out these recent news
