“Empowering Minds: The Essential Role of All Creatives in Shaping AI Design”

Diversity represents a significant advantage, particularly in the realm of artificial intelligence design. Consequently, individuals from various professional fields, encompassing law to the arts, are essential to the AI design and development process. Nonetheless, numerous teams striving to collaborate on AI frequently encounter a substantial hurdle: misunderstanding.

James Landay, a professor in computer science at Stanford University, championed participative AI during a recent podcast, emphasizing that comprehensive design of AI systems upfront is now the most vital element of AI implementations. Without human-centric values, Landay argued, AI endeavors are unlikely to succeed in his conversation with Lareina Yee, a senior partner at McKinsey.

Achieving a human-centered AI approach involves more than just the applications, according to Landay, cofounder and codirector of the Stanford Institute for Human-Centered Artificial Intelligence (HAI). He explained, “It is also about the manner in which we create and design those AI systems, who we involve in that development, and how we cultivate a process that is more human-centric while we create and assess AI systems.”

The unpredictability of AI poses a significant challenge. For instance, it differs substantially from PCs, and “in some respects, it is less reliable,” Landay stated. This is because AI systems operate on probabilistic principles, causing them to yield varying results based on the input data, unlike deterministic systems, “where the same input consistently produces the same output. We must rethink our approach to designing AI systems.”

Provide data to AI’s probabilistic models, “and receive varied results depending on how that data is interpreted within that vast neural network,” he noted. Furthermore, probabilistic models can generate hallucinations, or assertions that are false. “We don’t fully understand the reasons for their occurrence, which is a significant issue regarding the identities of those constructing these models.”

Consequently, managing AI systems becomes increasingly complicated when they deviate from expected outcomes. “This is why we must reconsider our design approach for AI systems, as they are destined to permeate all aspects of our daily existence—ranging from healthcare to education to governance,” Landay remarked.

“Currently, we primarily have groups of engineers, like responsible AI teams or safety departments, tasked with reviewing products prior to their launch. Regrettably, there’s substantial motivation to expedite the process. Moreover, these teams often lack the social capital to halt anything.”

Instead, it is essential to incorporate diverse expertise into the design and development phases. “We require teams comprised of individuals from various fields – social scientists, humanists, ethicists – as this will enable us to identify issues at an earlier stage. With these individuals as team members, they will possess the social capital necessary to effectuate change.”

A notable challenge within an open, interdisciplinary approach to AI is the potential for too many contributors in a single environment. “Individuals from different fields communicate in varied terminologies, meaning identical terms can carry different implications for distinct individuals,” Landay warned. “For instance, I’m collaborating on a project with an English professor and a medical school colleague. And what they refer to as a ‘pilot study’ is not the same as what I would categorize as a ‘pilot study.’”

Simultaneously, this type of confusion may not necessarily be a negative aspect — it could lead “to innovative concepts and alternative perspectives,” he elaborated. “For instance, we’ve had individuals working on extensive language models examining natural language processing. When they encounter an ethicist with a background in political science, that person raises inquiries regarding certain practices or the manner of releasing their software without suitable safeguards.”

AI is transforming our enterprises, workplaces, and society at large. It is crucial that this become a cooperative process.