Large Language Models Prompt Engineering Intermediate 1 min read

What is groundedness?

The degree to which an AI system's understanding is based on real-world experiences and sensory data

groundedness explained in plain English

Groundedness refers to how well an AI system's knowledge and understanding are connected to the physical world, rather than just being based on abstract concepts or theoretical models

Analogy

Think of groundedness like learning a new language - you can read books about it, but you'll only truly understand it if you practice speaking and listening to native speakers, which grounds your knowledge in real-world experience

Example

A self-driving car that has been trained on real-world driving data has higher groundedness than one that has only been trained on simulated data

How is groundedness used?

Groundedness is used to evaluate the effectiveness of AI systems, particularly those that interact with the physical world, such as robotics or computer vision

Common misconceptions about groundedness

Some people think that groundedness is only important for AI systems that interact with the physical world, but it's also important for systems that process and understand human language and behavior

History

The concept of groundedness originated in the field of cognitive science and robotics, where researchers recognized the importance of sensorimotor experience in learning and understanding

Also known as

embodiment situatedness real-world experience