Sharala Axryd, CEO, ASEAN Data Analytics Exchange (ADAX)
The vision of an all-knowing, omni-present intelligent being that forms the backbone of our everyday lives has been portrayed in movies that captivate the imagination of many. Today, that vision is not too far from reality, and we are seeing this at work through artificial intelligence (AI) – from AI-powered voice assistants like Alexa, to helping solve traffic issues, enabling the sequencing of DNA, tackling business problems and transforming industries such as tech, healthcare to logistics and fintech.
Current AI technologies are estimated to have the potential to automate about 50 percent of work activities in ASEAN’s four biggest economies – Indonesia, Malaysia, the Philippines and Thailand, according to McKinsey.
Even as AI increasingly finds its way into our everyday lives, the transformative power of AI is underpinned by a deeper issue that threatens the very fabric of our society – the presence of bias within it.
Uncovering the roots of bias
Much of AI’s capabilities as an intelligent, cognitive system rely on it being programmed and trained. At its core, AI operates on algorithms and data sets, the driving force of the digital economy in the 21st century. However, AI also unfortunately inherits and reflects the existing bias of its creators through the data it is given.
For example, when used in recruitment, a biased AI could be trained to shortlist potential candidates based on selected profiles of high-performing employees, which may not be representative of the company’s workforce nor consider diversity and inclusion as a factor for hiring, and potentially skew the hiring demographic.
The capabilities of AI are only as objective as the quality of the data inputs, as well as the assumptions around this data.
Bias is an inherent human trait and impossible to eliminate, however the goal is not to eliminate bias, but reduce it to a negligible level
When this data is not carefully selected, AI may not only validate the biases we hold, but further perpetuate them.
Leaving this unaddressed could pose issues for society, given that AI has already found its way into sectors such as telecommunications, medical, legal and finance. For example, a biased AI system might deny a bank loan simply because the borrower is located in a poorer neighborhood.
The possible scenarios are endless, though the conclusion resoundingly clear – bias in AI needs to be swiftly addressed while AI is still at its teething stages, before it progresses too far to root out issues that lie at the conception.
Part of the reason behind the existence of bias in AI points to the lack of diversity, particularly of gender, of the tech industry. Even more so for a highly specialized field like AI, and it has been found that just 18 percent of C-level executives in AI or machine learning companies are women.
Can AI be truly objective?
A key agent of change is to ensure that the data used to train AI is representative across various socioeconomic factors including race, religion, sexuality, education, career background and financial status.
Additionally, it is essential to expand and diversify the talent pool of people working on the next generation of AI. These should include women, creatives, sociologists and various industries that can together identify the lacking aspects of AI and provide the needed perspectives to weed out bias. Bringing in the alternative perspectives of women can boost creativity within the industry and cultivate gender diversity, and also prevent AI from becoming a skewed, gendered technology.
AI presents an exciting new frontier for the human race, and could possibly be the defining technology that will change our world like never before. The healthcare industry has been a key driver for AI, with recent breakthroughs including the world’s first AI-powered stethoscope, a Malaysian invention that enables precise surveillance and detection of heart and lung diseases.
However, as with any new technology, a cautious approach is needed in its design and implementation. While AI is designed to make our lives easier, our responsibilities and ethical obligations cannot be outsourced to machines. A global framework and increased governance of AI is essential to ensure that the very technology we are designing to help us does not do the contrary.
It has been said that bias is an inherent human trait and impossible to eliminate, however the goal is not to eliminate bias, but reduce it to a negligible level. When we bring together the greatest minds in the industry and involve partners across industries, communities and people from all walks of life, efforts to use AI to make the world a better place, transcending race, religion, color and gender, will deliver unparalleled outcomes for mankind.