In a recent event, VB Transform: Get Ahead of the Generative AI Revolution, leaders from Mastercard, eBay, and Capital One gathered to discuss the future of generative AI.
Building a Foundation for Equitable Generative AI
Emily Roberts, SVP, head of enterprise consumer product at Capital One, emphasized the importance of building continuously learning organizations and incorporating diversity of thought and representation into AI products. She highlighted the need for including the right people in the conversation and asking the right questions to ensure equity is at the foundation of AI projects.
JoAnn Stonier, fellow of data and AI at Mastercard, pointed out the challenges with the data used by public large language models (LLMs). She noted that the data these models learn from is often flawed, reflecting societal biases and inequities. As companies build their own products on top of these models, it’s crucial to understand potential biases in the data sets and focus on outcome-based usage.
Investing in Guardrails from the Start
Xiaodi Zhang, VP, seller experience at eBay, stressed the importance of establishing prompts and constraints to ensure equitable and unbiased results. She highlighted the need for continuous learning, flexibility, and openness to experimentation in the rapidly evolving sphere of generative AI.
Encouraging Internal Innovation
Companies are investing time in internal innovation to explore the potential of generative AI. eBay recently held a hackathon focused entirely on generative AI, leveraging the creativity and imagination of their teams. Mastercard is also encouraging internal innovation but has recognized the need for guardrails for experimentation and submission of use cases.
Navigating Regulatory Challenges
With the rapid advancement of technology, regulations have been modified to include generative AI. However, companies are still trying to understand what documentation will be required going forward and how they will be required to explain their projects as they progress. Zhang noted that the technology has leapfrogged regular regulations, so flexibility and readiness to respond to regulatory decisions are crucial.
Building in a Well-Managed, Well-Governed Way
Roberts noted that Capital One rebuilt its fraud platform from the ground up to harness the power of the cloud, data, and machine learning. She emphasized the importance of building the right experiments and applications in a well-managed, well-governed way, with human-centered guardrails to ensure responsible use of AI.
The discussion at VB Transform highlighted the opportunities and challenges in the rapidly evolving field of generative AI. As companies continue to explore this space, the focus remains on building equitable, well-governed AI systems that can drive innovation while ensuring fairness and transparency.