Coming into the AI world from the male dominated pharmaceutical sector, one of the most surprising things I found was that the relative lack of women in the technology industry was present throughout the whole employment pipeline, and not just at senior levels. For example, just 11 per cent of software developers are female.
I believe a critical driver behind this is the low level of female engagement and encouragement in secondary and higher education – although the percentage of girls taking computer science at A-level in the UK increased this year, they still make up only 10 per cent of computing A level exam entrants.
It is perhaps unsurprising then that in our own company computer science varies from other scientific fields in terms of its diversity. We’ve been able to achieve a 50:50 split in Benevolent Bio, the life sciences division of BenevolentAI. However, when looking to recruit in our AI division, BenevolentTech, the percentage of female applicants is very low, making a gender parity harder to achieve.
This is not an isolated trend. It’s estimated that by 2020 there will be 1.4 million jobs available in computing related fields in the US. Graduates in the country are on track to fill 29 per cent of those jobs, only 3 per cent of which will be filled by women.
Conscious and unconscious bias feeding the problem
Restrictions throughout the pipeline of female AI talent are reinforced by unconscious, and even conscious bias – both of which I’ve witnessed in my career. The AI industry is no different. For example, a study found that GitHub, the San Francisco based open source code repository, approved code written by women at a higher rate than code written by men: only if the gender of the author was not disclosed, though. Assumptions can impact on various aspects of a career. I’ve known a female AI speaker, due to give a keynote speech, to be given a detailed explanation of AI by a fellow male speaker, who automatically assumed, as a woman, she wouldn’t know what AI was.
This is not just a problem in terms of encouraging already talented female AI researchers into senior roles, but about fulfilling their potential as role models for other women. These researchers also go onto shape how AI is being incorporated into our everyday lives too and how this can in turn impact on the next generation of AI specialists.
Currently most chatbots and personal home assistants now have female names and voices – think of Alexa. This is reinforcing stereotypes and bias about the roles certain genders should play. Throughout our lives and schooling, parents and teachers make assumptions about what boys and girls will want to do that conform to their own unconscious biases too. In turn, these habits are feeding back into the problem.
Why it goes beyond gender
The Royal Academy of Engineering says more than a million computing engineers will be needed in the next ten years if the country is to maintain its leadership in this area, making it vital that more women are encouraged into the industry. This problem is not just about gender though, but about diversity generally. It’s well documented that companies with a more diverse workforce are more successful, and it’s something I see the benefit of every day in our company, with more diverse viewpoints contributing to better problem solving.
The AI industry, more than any other, has the potential to tackle this problem – or, if we don’t consider diversity carefully, the potential to contribute to it. Bias in data sets and analytics can be counteracted or reinforced, at an enormous scale, so it is really important that the industry deals with the issue before it becomes a bigger problem.
What’s to be done?
So how can we do this? It starts with a co-ordinated effort across the home, at school, at university and into the workplace. There has to be more encouragement for people from a diverse background, whether this be gender or otherwise to take up computer science. Part of this can be driven through funding. Initiatives like the Science Foundation Ireland funding Girls Hack Ireland, a programme of free technology and science centred events for teenage girls and even their parents, are examples of how we can encourage a minority group in the industry into the field.
It’s only by engaging in these initiatives that we can ensure there are more positive role models in the industry at all levels, which will in turn help to inspire future AI specialists. Ensuring that diversity is on the hiring, promotion and company cultural agenda across the board is also critical, to ensure that once hired, employees from minority groups can prosper.