What does a totally genderless voice sound like?
Trying to imagine it is like trying to picture ultraviolet light. We understand it exists - but can’t see it. So, how do we know what it looks like?
The question of whether you can really imagine something if it doesn’t exist yet might be an ontological dilemma, but it’s also the challenge that drives us as humans to break the mold and create something entirely new – something more hopeful.
Take Project Q, in Denmark, which claims to be the world’s first genderless voice assistant for AI. When you consider what a genderless voice sounds like, you might think of one that is androgynous, or so automated it eludes our human ideas about gender altogether. Log on to Q’s website, you’ll discover an intriguing mixture of both.
Launched in 2019, Q’s voice sits between 145 and 175 hertz, a frequency range defined by their team of audio experts as gender-neutral. “Think of me like Siri or Alexa, but neither male nor female,” Q tells us. “I’m created for a future where we’re no longer defined by gender, but how we define ourselves.” Amazingly, when the team behind Q asked 4,500 people from across Europe whether it sounded male or female, answers were split 50/50 – right down the middle – proving that the machine really does defy definition.
Q isn’t the only idea of its type to be proposed in order to shake up ideas about gender and AI. Kriti Sharma, who is Vice president of AI and ethics at software company Sage, urged us to consider a gender-neutral assistant named Pegg back in 2018, who “doesn’t pretend to be human” and so – vitally – does not reinforce outdated and problematic human stereotypes.
In this sense, Q and Sage represent the potential beauty of AI: that we can use it to program a future without human bias. If AI responds to the data sets that it’s fed, programmers have the power to make that data more representative in order to radically transform outcomes. In other words, we can, right now, create AI that is not limited by our human ideas about race, gender, age or other factors, but rather AI that sets us on the fast path to a fairer and more equitable society.
The Common Voice project, launched by Mozilla in 2019, is one example. It recruits global voices as a data set to diversify AI speech through machine learning and to help AI better understand accents from across the globe.
However, in a world plagued with sexism in almost every country, gender is arguably the most urgent frontier. Until now, many of our best-known assistants and AI technologies have been given feminine voices or faces. Think Eliza, the first chatbot. Or Sophia, the famous robot created by Hanson Robotics, first switched on in 2016 and appearing on The Tonight Show Starring Jimmy Fallon a year later. Siri, Alexa, Google Home and Cortana all default as women, and research has found that a concerning 67% of digital assistants are coded female. It’s almost as though we’re most comfortable barking orders at feminine voices. Experts agree, explaining that the tendency to code AI as female is promoting and perpetuating the existent idea that women are more subservient. If in doubt, we could look to pop culture here too, films like Her and Ex-Machina.
The answer? The main way to truly transform AI is to also look at who is creating it. Bias is built into the AI industry with men massively outweighing women in the sector – 12% of leading machine learning researchers identify as female, and 22% of jobs in artificial intelligence are held by women. This means that AI is often built to optimize male users, and has been found to recognize and understand male voices and faces more accurately than women’s. The same goes for race. With many AI systems proven better at reading white faces, AI has a big discrimination problem. If white men are at the wheel of AI development, then when we enter singularly, we – humanity – will have created racist, gender-biased machines that will then forever take on our prejudice. The only way forward is to take human learnings about gender and ask AI to do better.
In some fields, this is already happening. According to UNESCO, “AI-powered gender-decoders help employers use gender-sensitive language to write job postings that are more inclusive in order to increase the diversity of their workforce.” This shows that change is possible – that AI has powerful potential to make hiring processes fairer, and could even be used to train workers themselves to improve gender equality in society.
Q and Pegg also represent solutions. Yet gender neutrality in AI doesn’t only pave the way for a fairer future when it comes to existing gender biases – it also reflects the diversity of gender expression we’re living with today. If one in five US adults in America say they know someone who goes by a gender-neutral pronoun, and Pew research for the Vice 2030 identity report found that 41% of Gen Z already identify as gender-neutral, we need AI systems and avatars that reflect the boundless ways that we are identifying already. A plethora of gender identities are out there, so why are AIs so often female?
“As society continues to break down the gender binary, recognizing those who neither identify as male nor female, the technology we create should follow,” Ryan Sherman, a developer at Project Q, told the Guardian. The team hope Q is an example of what the future holds: “a future of ideas, inclusion, positions and diverse representation in technology.”
For these reasons and more, over the coming years, research into more equitable AI is set to double down – whether it’s using AI to improve workplace equality between men and women, programming AI itself to be less biased, or doing away with gender in AI altogether. And when 84% of millennials now use voice-activated assistants in day to day life, there truly is no time like the present for this critical change of direction.
Soon, we’ll spend more time talking to machines than our own families, pointed out Sharma in the same speech where she presented the idea of Pegg. “The good news is the design of AI — whether it’s gender-neutral, unbiased, and nondiscriminatory — is entirely within our control.”