One of the biggest surprises I had when I entered the AI industry was the lack of women that were present within it. It's not just at the senior levels either, it's a problem which seems to be present throughout every layer of AI organisations.
It isn't hard to see where this problem starts either. Although there has been a slight increase in the percentage of girls taking computer science at A level in the UK this year, it still only makes up 10% of entrants. Couple this with only 11% of software developers being women and you can see the problem we're faced with.
As a company, BenevolentAI has succeeded in achieving an even split of men and women in its life science division, BenevolentBio, which I lead. Achieving that split was much easier than it was for my counterparts in the AI division of the company, BenevolentTech. Simply put, the percentage of female applications for roles was low, making it harder to hit the even split.
Conscious and unconscious bias are the issue
Through my career I've witnessed how bias - both conscious and unconscious - has restricted the pipeline of female talent in a variety of sectors - unfortunately the AI industry is no different. If that bias is taken away it can offer a more level playing field on which women can compete. A recent study revealed the potential of this. A San Francisco based open source repository, named GitHub, approved the code women wrote at a higher rate than men. However, there was just one snag; this was only found to be the case when the gender of individual authors was withheld.
This bias can hit at any part of a career, from entry level all the way to the top. It may come as no surprise to hear that I've known a female speaker, due to give a keynote address, have AI explained to her by a male colleague, as the assumption was made that her gender meant she wouldn't know what AI was.
It's vital that we encourage those already in the industry to become the much sought-after female role models the industry needs to inspire future generations. In addition, these women can help to form a part of how AI is developed for decades and, more importantly, how it is incorporated into our daily lives. If you look at the most well-known examples of 'AI' in our society today, these are in the form of chatbots and personal home assistants like Alexa and Siri. Each has a female voice and help to reinforce the stereotypes and bias about what roles each gender should play. It's common throughout our society, with adults making assumptions about what girls and boys should conform to - it's this that keeps the bias prevalent.
Gender is not the only issue
It's not just a gender problem we need to fix, but one of diversity in general. It's predicted that more than a million computing engineers will be needed in the next 10 years if the UK is to keep its place at the top table in the industry, according to the Royal Academy of Engineering.
It's well documented that companies with a more diverse workforce are more successful, something I can attest to, with the benefits I see coming out of our company. The AI industry is at a crossroads and needs to tackle the issue of diversity head-on. Looked at from a purely technical point of view, the industry relies on data, so any bias in how data sets are procured or used - as the result of a lack of diversity - has the potential to reinforce the issue I've been speaking about with huge negative effects for different groups of people across the globe. If the data is biased, then its output will be too.
The next step
So, what's the next step to overcoming this problem? There needs to be a concerted and co-ordinated effort in schools, universities and businesses, encouraging people from diverse backgrounds into computer science. There are some programmes that are making a good effort to do this, such as the Science Foundation Ireland funding for Girls Hack Ireland. This is a programme of free science and technology centred events for teenage girls and even their parents can get involved too. It's this kind of activity that will ensure those role models that the industry desperately needs are developed - in turn bringing greater diversity to the AI industry.