AI has been both a blessing and a curse ever since it became a commonly used technology. Although people are happy to use it to make their lives a little easier, companies and developers still can't iron out a few kinks which include bias and discrimination in AI models. An expert believes this could lead to grave consequences.
AI Bias and Its Risks
Even though AI technology still has a long way to go before it can become a reliable source of data, a lot of organizations and agencies already use it for certain functions like facial recognition. The problem is that AI can be inaccurate sometimes.
Since AI is trained using the available data online, it picks up other discriminatory and biased opinions as well, which are then embedded into AI models. Faculty of Law Assistant Professor at the University of Alberta, Dr. Gideon Christian warned people of the potential risks.
Dr. Christian was given a $50,000 Office of the Privacy Commissioner Contributions Program grant, which funds research called Mitigating Race, Gender and Privacy Impacts of AI Facial Recognition Technology, as reported by Interesting Engineering.
The study looks into bias in AI-based facial recognition technology in Canada, particularly the issues it has with identifying race. He stated that there is a "false notion that technology unlike humans is not biased," which is not accurate.
He also explained that facial recognition had a 99% accuracy rate in white male faces, while recognizing people of color, especially when it comes to women, has a high error rate of about 35%. This leads to others being arrested for a crime someone who looked like them committed.
The issue is not limited to the US. There have been cases in Canada where immigrants are stripped of their refugee status, all because their faces were matched to another person using facial recognition technology.
As a result, the government believed that the immigrants were using false identities. These cases involved black women in the country, which just confirms the statistics that people of color were more at risk when AI misidentified a person.
It Got a Pregnant Woman Falsy Arrested
A recent incident occurred back in early August, where a pregnant woman, Porsche Woodruff, was arrested by the Detroit police after being identified by facial recognition. The victim already had her photo in the police database due to a previous arrest.
She was held by the authorities for 11 hours and was questioned. Woodruff was charged in court with what she was alleged with, as well as paid a $100,000 personal bond in order to get released, all while she was innocent of the crime.
What's worse is that the victim of the crime said that Woodruff was the perpetrator when presented with a photo lineup, which was the confirmation the police needed to arrest her. The falsely accused already filed a lawsuit against the city after the incident.
As mentioned in Ars Technica, City Police Chief James E. White, expressed his concern over the mistake, stating that the matter was being taken seriously. Before Woodruff, there have already been three other false arrests.
