Artificial intelligence (AI) is changing the way businesses work, but the emerging technology also has potential to change how people treat, trust and interact with each other.

Recent generative AI offerings allow organizations to make predictions based on data and generate new text, images or audio. But real-world biases reflected in the data used to train AI models can lead to the reproduction of those biases, New York University Professor Meredith Broussard pointed out during a VMware Explore 2023 panel discussion on responsible AI.

At the heart of AI's potential to contradict common ethics is the idea of technochauvinism – a pro-technology bias that supports technological solutions as superior to all others. But it usually makes far more sense to use the right tool for the task, regardless of how technologically advanced that tool might be, Broussard argued.

"People imagine that they're going to get a lot of status or power from being associated with the latest technological innovation. Often, that does happen, but it doesn't happen with every single [technology]. Often, the technological flavor of the month goes kaput, and then you have egg on your face," Broussard said.

Instead, the role of AI should be "to assist humans," not replace us, she said. "Being reasonable about the boundaries of the technology is a good place to hang out."

VMware VP of AI Labs Chris Wolf explained that identifying and removing biases from data used to train AI models is a priority. "AI ethics is really important," he said during the panel. His team works to understand how AI models are trained, what data sources are used and the types of biases present. "We we want to be able to help customers to understand correctness of AI, as well as explainability of AI results," he said.

The industry also recognizes the importance of prioritizing ethics when using AI. Businesses should approach AI through the lenses of transparency, accountability, cybersecurity, privacy, risk management, bias, fact-base erosion, environmental/social sustainability and human exploitation, according to Gartner analysts Bart Willemsen and Jo Karajanov, who co-authored a recent research report on the topic. "Generative AI bears great potential, but can also disrupt who, how and what people trust; affect how people interact and are treated; and impact the global environment," they warned.

The analysts recommend that enterprises implement "adequate controls" in regard to AI bias and risk management and ensure AI ethics initiatives avoid "anything that increases fact-base erosion, is harmful to the environment or society, or even remotely connects to human exploitation," they wrote.

Brousssard claimed the belief that technology is objective, neutral and unbiased falls victim to technochauvinist bias. "AI systems – automated systems – discriminate by default," she said. "It's not a question of if it's happening. It's a question of how it's happening."

Take mortgage approval algorithms, for example. A 2021 investigation found that algorithms used to approve or deny mortgage applications were far more likely to deny borrowers of color compared to white borrowers.

The data used to feed the model was based on who had been approved for home loans in the past – data that included the United States' legacy of financial discrimination, bias in residential lending and residential segregation. "It's not surprising that the model is making biased decisions," Broussard said.

The first step is admitting there's a problem. After that, "we can put our finger on the scale mathematically, ... and we can tweak the model's outputs to make it more fair," she said.

Business leaders using AI need to "confront the reality" and welcome discussions about race, gender and disability in the workplace, she said. "These are things we tend to not talk about in offices, but we need to have these hard conversations," and pretending that discrimination isn't happening won't do anyone any favors.