Effective Generative AI: The Human-Algorithm Centaur

Abstract

Advanced analytics science methods have enabled combining the power of artificial and human intelligence, creating centaurs that allow superior decision-making. Centaurs are hybrid human-algorithm models that combine both formal analytics and human intuition in a symbiotic manner within their learning and reasoning process. We argue that the future of AI development and use in many domains needs to focus more on centaurs as opposed to other AI approaches. This paradigm shift towards centaur-based AI methods raises some fundamental questions: How are centaurs different from other human-in-the-loop methods? What are the most effective methods for creating centaurs?

When should centaurs be used, and when should the lead be given to pure AI models? Doesn’t the incorporation of human intuition—which at times can be misleading—in centaurs’ decision-making process degrade its performance compared to pure AI methods? This work aims to address these fundamental questions, focusing on recent advancements in generative AI, and especially in Large Language Models (LLMs), as a main case study to illustrate centaurs’ critical essentiality to future AI endeavors.

Keywords: Centaurs, Generative AI, Large Language Models, Human-AI Algorithms