As the use of AI and voice assistants becomes increasingly prevalent both in the home and at work, 2020 could be the year that our approach to AI voice assistants becomes more diverse and inclusive.
While fully inclusive AI is yet to become a reality, there are tech companies exploring what it means to create voice assistants and other artificial intelligence with greater diversity. This means creating AI that includes – and hopefully ultimately embraces – a wider range of genders, voices and ethnicities.
So, which organisations are doing this work? Why is it important? And what does diversity in AI look – or sound – like?
Gender and AI: what you need to know
Although voice assistants have come under scrutiny for the way in which they depict female voices, tones and responses, gender and AI isn’t confined to male/female classification.
As society has come to more widely recognise all genders (and none), so too must AI. The development of non-binary voice assistants is another element that must be explored.
But it doesn’t stop there: when thinking about what would make AI or voice assistants truly diverse, it’s important to consider other aspects of diversity too. This means thinking about – and creating – AI that can respond equally to human voices from diverse ethnicities, as well as regional and non-native accents.
Moves are already being made towards more diverse AI. Firstly, there's Q, the first genderless voice assistant, whose voice has been recorded by people that don't identify as male or female, and pitched at frequency which further removes any sense of gender. Then there's F’xa – a chatbot created by a team with diverse gender and race identities, whose mission is to tackle the issue of reasserting the balance of the AI industry, in which it states “only 22% of the people building AI right now are female.”
Artificial intelligence and recruitment
But what about the real-world application of AI? One industry that’s using AI more and more is recruitment.
While AI can benefit human resources teams by freeing up time that would otherwise be spent on admin tasks, it can still create further problems. When Amazon used AI during the hiring process, it was found to negatively process CVs that referenced women. Why? Because the AI was being fed historical, man-made data sets, that already had these biases ingrained – it simply replicated the same patterns.
With AI increasingly used during the recruitment process (as seen in South Korea), there’s even more need to make AI that can respond equally and fairly to a diverse pool of candidates.
There’s already awareness around many of the biases that can affect hiring decisions. And now, this needs to include using AI that can understand a diverse range of accents and voices too.
Voice assistants and AI: The Future
How else might the technology be developed? And where do we go from here? As we know, the application of AI extends beyond the interview room or domestic use.
At CES 2020, Neon showcased its ‘artificial humans’, essentially human avatars that are intended to act as the ‘front end' of an AI. Think AI can be used constantly to automate tasks and outdo human productivity? Think again. Neon’s creations are designed to have flaws and are meant to ‘get tired’. And, with such human-like characteristics, when does AI go from being an ‘it’ to a ‘you’?
Similarly, it’s been said that for AI to advance even further, the technology needs to incorporate other elements of human behaviour, such as learning how to forget. The argument here is that by having to constantly reevaluate information, the AI brain can better mirror the human learning process.
But with “diversity, non-discrimination and fairness” one of the EU’s seven key requirements for ethical AI, progress can’t be at the expense of inclusivity.
AI Needs to Reflect Society
Clearly, greater diversity in AI and voice assistants is required. This can be through addressing AI and gender bias, such as non-binary representation. Also, by ensuring that the technology reflects the widest possible range of demographics, including accents and voices.
As AI and voice assistants continue to be used in the real world, especially in recruitment, the need to develop AI that’s as diverse as possible becomes even more crucial.
Plus, with the future of the technology looking set to focus on becoming more human, it’ll have to be more diverse to reflect society itself.