Where are LLMs heading?

Concetti sui Large Language Models (LLM)

Vidhi Chugh

AI strategist and ethicist

Journey so far

Basics of LLMs

Concetti sui Large Language Models (LLM)

Journey so far

LLMs applications in NLP

Concetti sui Large Language Models (LLM)

Journey so far

Training processes involving data consideration, ethics, and environmental impact

Concetti sui Large Language Models (LLM)

Journey so far

Future of LLMs

Concetti sui Large Language Models (LLM)

Model explainability

 

  • How do they arrive at their outputs?

 

  • Road-trip planning
    • Why this particular route?
    • Why these specific spots?

 

  • Builds trust and transparency
  • Identify and correct the biases or errors

 

Trust and explainability

1 Freepik
Concetti sui Large Language Models (LLM)

Efficiency

  • Computational efficiency

    • High-quality output with less compute

  • Faster and efficient

    • Model compression
    • Optimization

  • Benefits: better storage, lower energy use

  • Accessibility and sustainability

    • Promotes green AI
    • Reduces operating costs

Efficiency

1 Freepik
Concetti sui Large Language Models (LLM)

Unsupervised bias handling

  • Biased data -> discrimination

  • Unsupervised bias handling

    • Bias detection and mitigation techniques, automatically
    • No need of explicit human-labeled data
    • Identifies and reduces by analyzing patterns

  • Challenge

    • Subtle, difficult to detect
    • Might introduce new biases

Supervision

Concetti sui Large Language Models (LLM)

Enhanced creativity

  • Creativity in text-based and visual art forms
  • Artistic content: learned patterns, not emotional understanding
  • Lack human-like comprehension of art or emotions
  • Demonstrate human-like emotional behavior
  • Future: emotion inference

Four images generated by AI

1 https://arxiv.org/pdf/2302.09582.pdf
Concetti sui Large Language Models (LLM)

Let's practice!

Concetti sui Large Language Models (LLM)

Preparing Video For Download...