Making GenAI Safe, Trustworthy and Fit for Purpose with Auto Alignment

Explore a groundbreaking alignment templating technology for enhancing the fairness, robustness, and safety of generative AI in this 56-minute Toronto Machine Learning Series (TMLS) conference talk. Delve into methods for understanding expected model behavior, measuring underperformance using synthetic test data, and iteratively improving models with minimal human intervention. Discover how this alignment platform can be applied to specific use cases, including promoting fairness towards underrepresented groups, reducing toxicity and misogyny, tailoring political viewpoints, customizing tone for customer service applications, preserving PII, and preventing harmful responses. Learn about the technology's ability to interact with users for contextual decision-making in intention-understanding, data generation, testing, and tuning processes. Presented by Rahm Hafiz, CTO and Co-Founder, and Dan Adamson, CEO and Co-founder of Armilla AI.