Tech tackles gender bias
Updated: Aug 28, 2020
As tech tries to emulate the real world, we realise that there are some real world issues that are better left behind.
Google removes gender labels from its image recognition to reduce inbuilt bias and reduce misattribution. This has been a large problem for people who don't have typically gendered appearance or don't identify as cisgender. AI gender identification success also varies depending on the race of the subject.
Just how objective is the news? 'Normally' created a filter that highlighted all news content written by men or women, the plugin quickly saw a huge imbalance, spotting that 60% of bylines are written by men. In breaking news it found this figure jumped to 70%. While this filter won't be released, its another clever way of visually highlighting the inequity that still exists
The music sounds better with Q A new genderless voice assistant has been developed to tackle the typically gendered Alexa's and Siri's of the world. 'Q' was created using the recordings of 5 non-binary people and fine tuned in a process involving 4,600 people who rated options from 1 (male) to 5 (female).