The fear of robots taking over the world might not be so far-fetched.
While artificial intelligence is praised for its responsiveness to large amounts of information, AI also seems to absorb bias. An experiment published in June demonstrated that robots trained in artificial intelligence displayed racism and sexism in their decision-making.
While sifting through billions of images, the robots in this experiment consistently categorized black men as “criminals.” Similarly, the label “housewife” was attributed more regularly to women than to men.
In the programmers’ efforts to create a machine that could eliminate human error, they failed.
AI is made by humans and deployed in existing systems and institutions – none of which are immune to sexist and racist practices throughout history. Without studies like this, such ethical concerns might have been overlooked altogether.
This technology is already so ingrained in our lives, in a way that seems to have relatively low stakes. AI can be used to restock grocery shelves, create online shopping algorithms, or even play chess.
Systemic issues arise when AI systems are encrypted in addition to just storage robots.
Artificial intelligence tools are being used to screen potential tenants or job applicants, relying on criminal history and evictions to make their choices. These screening process tools reflect long-standing racial disparities in the criminal and justice system, which greatly affect members of marginalized communities against whom these systems consistently discriminate.
And while this may only be a setback in AI development, it has the potential to become something more without this question being explored.
Zac Stewart Rogers, professor of supply chain management at Colorado State University, told the Washington Post that new coding software is often built on pre-existing structures. This, he said, explains the prevalence of persistent problems in new robotic systems.
Time and time again, systemic injustice has taken root in our institutions – even in the realm of technology. This problem could become rooted in the future development of artificial intelligence unless we act while AI is still under development.
Artificial intelligence has been called the future of technology in the world and it has many potential benefits if produced and introduced correctly. Many cite the benefits of expanding the use of artificial intelligence, including its potential for economic growth, cost-effective increases in productivity, job creation and increased GDP.
With a growing population, job creation is important. Biases need to be removed from data used in AI algorithms, starting with a more representative group of AI developers. Either the AI algorithms need to have corrective action, or the solutions to our growing world have become much more complicated.
This raises a question: is it really possible to have a robot without human error if it is made by humans?
To get the day’s news and headlines delivered to your inbox every morning, sign up for our email newsletters.