As artificial intelligence (AI) models become more widespread, it is crucial for tech leaders to consider how they’re training AI models and how public perception of AI will impact the technology’s adoption.
CEO and Chairwoman of Bast AI Beth Rudden and IBM Consultant Phaedra Boinodiris are co-authors of the book “AI for the Rest of Us.” They spoke during our Talent Forward 2023 event to share how to tackle issues of trust, transparency, and diversity when working to train AI models.
Building Public Trust in AI Is a Social and Technical Problem
As organizations implement AI tools more widely, one of the biggest obstacles is earning the public’s trust in how these tools are created and applied.
“Explaining what data is used in the models and how algorithms work is the most practical way to get AI implemented and adopted,” said Rudden.
Building the public’s trust in AI will require a multi-faceted approach.
“Anytime you talk about something that requires a holistic approach, you need to be talking about three key things: People, processes, and tools,” Boinodiris explained.
According to Boinodiris, developers need to focus on governance processes for responsible AI use, as well as the right tools for building engineering frameworks. However, she emphasized that the most difficult aspect is creating organizational cultures that foster responsible AI development and use.
Diverse Teams Develop Stronger Solutions
To create a culture that approaches AI curation responsibly, Boinodiris asserted that the teams building AI models should be diverse and multi-disciplinary.
“It's not just in terms of who's developing AI responsibly, but who's crafting the governance around AI,” she explained.
At IBM’s Center of Excellence, people from a wide variety of backgrounds are invited to work with data and machine learning scientists to build AI models that can approach problems from various angles.
“We don't care what your background is,” emphasized Boinodiris. "You have a background in psychology, sociology, anthropology, or design? You're interested in social justice? You're interested in law? You're welcome here. You have a seat at this table.”
As AI Gets Personal, Transparency Is Key
AI is quickly developing the ability to deliver highly personalized results and recommendations. However, individualized responses aren’t inherently more trustworthy.
“The promise of AI is hyper-personalization,” explained Rudder, “but we have to change how we look at artificial intelligence, and we have to show the world that we need to see the sources that are being used in order to train it.”