Building better AI through diversity
Artificial Intelligence (AI) and Machine Learning (ML) technologies are becoming more prevalent in our lives. Ashley Howden, CEO of Telstra Wholesale partner KJR, is backing a leadership camp for young women in AI, to help bring greater diversity to STEM (science, technology, engineering and maths).
Today AI and ML algorithms are being applied to diverse use cases from voice-based digital assistants, maps and real-time navigation, image recognition, song and movie recommendations through to medical diagnostics and industrial manufacturing.
“On paper, the benefits of AI are compelling. It can help us get the most out of Big Data, enable us to do our jobs better and more efficiently, and improve outcomes in areas such as healthcare. However, if we are going to gain the full benefits from AI, it’s imperative that we can first trust it to make the right inferences and decisions,” CEO of strategic technology consultancy KJR Ashley Howden said.
The problem of unintended biases
Artificial intelligences can process information and learn so fast that they often surpass the engineer’s ability to explain how it made the decisions or produced the outcomes that it did. Ashley says this means that it is vital to ensure algorithms are programmed without any inherent and unintended bias.
“Modern AI is all about teaching computers how to learn. Researchers feed huge data sets to computers and reward them when they find valuable patterns. The machine then learns and improves its processes,” Ashley said.
“Yet every data set has some degree of skew built in, and because the AI learns from that data, engineers have to be very careful that they don’t design bias into the AI algorithms. If they do, the AI will compound the effects of that bias well in to the future and it becomes hard to identify and correct.”
This is where the value of a diverse engineering team becomes apparent. “One of the ways to introduce rigour in decision-making process is to have diverse opinions. Individuals within homogenous teams tend to make similar decisions. Often this comes from not challenging each other’s worldview or perspectives,” Ashley said. “Gender is the obvious starting point for diversity in the workplace. In contrast, if there is gender inequity in the workplace, the uniformity of view may get carried forward into the AI, which will then self-reinforce the group’s unintended biases.”
Building in diversity
KJR is sensitive to the need for diversity in its own teams because of the work they do testing clients’ software solutions.
“It is very early days for many corporates when it comes to implementing an AI solution or ML-based service. Our work on client AI projects has focused on interrogating the processes to provide assurance for the outcomes from the AI decision-making processes and we need different perspectives in our team to do our job properly,” Ashley says. “However, there is a broader issue that we want to help solve, which is the gender inequity in STEM industries.”
Young Women Leaders in AI
In 2016, research by the Office of the Chief Scientist revealed that women make up less than one-fifth of Australians qualified in science, technology, engineering and maths. “If we don’t see an increase in the number of women coming into the industry, inherent bias in AI will continue to be a potential risk,” Ashley said.
KJR is acting by helping organise the Young Women Leaders in AI program. The program was created by IntelliHQ, a not-for-profit organisation committed to the establishment and support of an innovation hub for AI and machine learning technologies. It is supported by KJR and Amazon Web Services (AWS).
The program centres on an annual, six-day leadership camp, commencing in April 2019 on the Gold Coast, which will bring together up to 75 young women with a passion for AI. Graduates from the leadership camp will undergo a year-long ambassador program to secure local leadership roles in AI, while mentoring will continue on a national scale.
“AI is enabling rapid transformation of technology and automation and is poised deliver a seismic shift to industry in all its functions. At its greatest success, AI will fundamentally enhance the sophistication of the human condition. Programs such as the Young Women Leaders in AI can help the industry reach that potential by ensuring that we create strong, diverse base of skilled talent for developing AI in Australia,” Ashley says.
If you or someone that you know would be interested in participating in or supporting the program, you can find more information at womeninai.com.au
