AI Wednesdays: The One About AI Safety
From neural network interpretability to ethical AI deployment, our speakers from SASH, impress.ai, and Neo4j will explore different aspects of building safer AI systems. Whether you're interested in technical implementations or broader implications, join us for this comprehensive look at AI safety.
More about the Sharings
Clement Neo (Research Engineer, Singapore AI Safety Hub) will present an introduction to mechanistic interpretability in neural networks. The talk will cover the history and fundamental concepts of mechanistic interpretability, its importance for understanding AI systems for safety, and current research directions and open questions in the field. (Technical Level: 300)
Vaisagh Viswanathan (CEO & Co-Founder, impress.ai) critical examination of AI ethics in hiring - past, present, and future. He'll explore how recruitment AI has evolved from basic bias reduction to today's complex LLM-driven systems and discuss key considerations for both users and developers of AI recruitment tools. Discover what responsible AI adoption looks like in hiring and what ethical questions we should be asking as this technology advances. (Technical Level: 100-200)
Siao Si Looi, a communications specialist at FAR.AI—a non-profit dedicated to AI safety—will present an overview of the organisation's research initiatives and field building events. The talk will cover the robustness research agenda FAR.AI has pursued, including their work on beating superhuman Go AIs. (Technical Level 200-300)
Explore Neo4j's GraphRAG with Siddhant. See how graph databases can level up your RAG systems through improved vector search and context handling! (Technical Level: 200)
Do note that sharings will begin at 3pm!
More about the Speakers
Clement has been working on the field of mechanistic interpretability for the past two years, conducting research and advising others entering the field. He is currently a Research Engineer at the Singapore AI Safety Institute. He can be reached on Twitter @_clementneo or via email!
Dr Vaisagh Viswanathan is the CEO and Co-Founder of impress.ai, an AI-powered recruitment automation platform. With a PhD in AI agents from NTU Singapore, he combines his expertise in AI research and enterprise software development to lead impress.ai's expansion from Singapore to global markets. He is passionate about using AI to create fairer, more equitable workplaces and transform how we work.
Siao Si Looi is a communications specialist at FAR.AI, an AI safety research and education non-profit. She assists the research team with paper submissions, produces blog content, manages social media channels, and supports the events team by promoting field-building content from those events.
Siddhant Agarwal is a seasoned DevRel professional with over a decade of experience cultivating innovation and scaling developer ecosystems globally. Currently leading Developer Relations across APAC at Neo4j and recognized as a Google Developer Expert in Gen-AI, Sid transforms local developer initiatives into global success stories with his signature "Local to Global" approach. Previously at Google managing flagship developer programs, he has shared his technical expertise at diverse forums worldwide, fueling inspiration and innovation. Know more about him at meetsid.dev.