Is the Human-In-The-Loop AI still relevant at the era of Agentic AI?

This document discusses the evolving role of Human-In-The-Loop (HITL) AI in the context of advancing Agentic AI within geoscience.

Evolution of AI in Geoscience

Early AI systems in geoscience relied on supervised learning, requiring significant input from subject matter experts (SMEs) for training and refinement.

The introduction of Large Language Models (LLMs) shifted the SME role from training to prompt engineering, enhancing interaction with AI systems.

The emergence of Agentic AI, capable of autonomous decision-making, raises questions about the necessity of human expertise in AI development and deployment.

Trust and HITL in Agentic AI

Trust in Agentic AI, which relies on LLMs, is limited due to their lack of logical reasoning capabilities.

HITL is crucial for enhancing trust, allowing SMEs to intervene during data gathering, reasoning, and feedback processes to ensure accuracy and reliability.

HITL facilitates enterprise adoption, aligns with regulatory frameworks, and supports scalable learning through continuous human feedback.

Conclusion

As Agentic AI matures, HITL will evolve, enabling SMEs to define goals, seek human input for uncertainties, and incorporate feedback for ethical improvements.

HITL remains essential, positioning human expertise as a guiding force in the future of autonomous AI systems.