Explainability of LLMs

Explainable Artificial Intelligence (XAI) Concepts

Folkert Stijnman

ML Engineer

Large Language Models (LLMs)

  • Understand, generate, and interpret human language
  • ChatGPT, Bard
  • Trained on vast amounts of text data
  • Write articles, poetry, code
  • How to understand its decisions?

 

a graphic of an LLM

Explainable Artificial Intelligence (XAI) Concepts

LLMs and explainability

graphic depicting how data goes into a large language model and prompting the model

  • Complex due to:
    • Depth and scale of neural network layers inside model
    • Large amounts of text, i.e. terabytes of data
Explainable Artificial Intelligence (XAI) Concepts

Methods for LLM explainability

Creator

  • Transparency in training data
  • Use simpler models
  • Incorporate human feedback

End user

  • Self-explanation
  • Tweaking the input
Explainable Artificial Intelligence (XAI) Concepts

Future use of LLMs

  • Enhancing explainability will remain a requirement
  • Building trust and confidence

a graphic depicting a brain with different aspects representing language processing

Explainable Artificial Intelligence (XAI) Concepts

Let's practice!

Explainable Artificial Intelligence (XAI) Concepts

Preparing Video For Download...