The future is approaching fast, providing a rare tabula rasa-esque opportunity to correct previous mistakes in order to shepherd in a more diverse future with technology that actually works.
The issue of the lack of diversity within STEM has come into sharp relief of late. Much-needed public debate has been had about homogeneity in tech conference line-ups and the reasons behind the disparity in the number of women and minorities entering STEM fields versus how many actually end up in STEM roles.
It’s hard to pinpoint exactly what has led to the lack of diversity in STEM as there are a myriad of forces at work that have contributed. Recently, evidence has been produced to suggest that not only is increasing diversity important as a matter of principle but that companies can expect a healthy boost to their bottom line if they seek out a more diverse workforce. The excitement about this potential has, in turn, inspired many HR and recruitment professionals to make diversity their top priority.
So, good moves are being made to correct these missteps of the past. Now, we also are faced with a unique opportunity to author the future and ensure that history is not repeated.
Leading AI researcher Andrew Ng has said that the introduction of AI will bring about a revolution comparable to the proliferation of electricity in the latter half of the 19th century.
It has the potential to transform the landscape of our entire society and could affect everything from how we work to how we travel.
Yet a potential threat looms that could kneecap this technology and cause our world to edge away from this new and improved way of existing: a lack of diversity in who programs AI, explained RefineAI CEO Bharat Krish at Inspirefest 2018.
Krish emphasised the ubiquity AI will have in the future, telling the audience that the technology will be present in everything from finance to healthcare to detecting fraud. In the next 10 years, it is estimated that AI will have a global impact of between $1.5trn and $2.95trn.
He posed a question worthy of grave consideration: “Who is thinking of minorities when building the systems?” He went on to explain what motivated him to start talking about this.
Krish and his co-founder set out to create a facial recognition technology. They engaged with the top universities in the US and dipped into open source data to construct what they thought was a good algorithm fit for purpose. However, the technology hit a rather glaring snag: it failed to properly recognise the faces of people with darker skin tones.
“I’m sure you can see the irony here,” Krish said, gesturing broadly to himself and eliciting titters from the audience. “Here I am, building a company based on a technology that recognises faces, but it doesn’t even recognise my face.”
After laughing it off, the situation gave Krish pause. He was quick to recognise the technology’s bias due to being dark-skinned himself, but it led him to wonder whether the same would be the case in other AI start-ups, many of which wouldn’t have any dark-skinned people on their founding team.
“If all of these biases are not checked, they will end up in AI and explode.”
The leading AI technologies today, Krish went on to say, do an “exceedingly poor job” of recognising women with a darker skin colour. When you consider the fact that these algorithms are being leveraged by police forces to do virtual line-ups, the true danger of an AI possessing these kinds of blind spots becomes readily apparent. “Who do you think will be identified, or misidentified?”
To hear more about how to correct the issues with AI before they arise, check out Bharat Krish’s talk in full above.