Dehumanizing Agents - Why Explainability is Crucial in the LLM Era

Explore the critical importance of explainable AI in the era of large language models through this comprehensive conference talk that addresses the inherent biases and limitations of artificial intelligence systems. Examine how AI models inherit human biases from training data and learn why explainability becomes essential when LLMs confidently present information, even when generating inaccurate content. Discover the evolution from traditional machine learning explainability methods to novel approaches specifically designed for generative AI systems. Delve into various implementation methods ranging from established techniques to cutting-edge research proposals, while understanding the main risks and challenges encountered when implementing explainability in AI systems. Learn practical solutions for overcoming these challenges and gain valuable insights into integrating explanations into LLM-based workflows, including strategies for working with third-party services that lack native explainability features. Master the techniques needed to break open the "black box" of AI models and understand their decision-making processes in an era where trust in AI output is paramount.