LLMs Do Not Have Human-Like Working Memories

Overview

Explore the limitations of Large Language Models in replicating human working memory through three experimental games, revealing key challenges for achieving artificial general intelligence.

Syllabus

• Introduction to Large Language Models
Overview of LLM architecture
Comparison to human cognitive architecture
Introduction to working memory
• Understanding Human Working Memory
Components of working memory
Working memory in cognitive processes
The role of working memory in human intelligence
• Experimental Game 1: Short-term Retention
Game design and objectives
Measuring retention capabilities
Analysis of LLM performance vs. humans
• Experimental Game 2: Sequential Processing
Game design and objectives
Understanding sequential task execution
Challenges for LLMs in sequential reasoning
• Experimental Game 3: Integrating New Information
Game design and objectives
Evaluating adaptability and learning in context
LLM limitations in context-sensitive tasks
• Key Challenges for Artificial General Intelligence
Limitations of current LLMs in emulating human cognition
Potential pathways to enhance LLM memory use
Discussion on future AI research directions
• Conclusion
Summary of findings from experimental games
Implications for the development of AGI
Final thoughts on AI and working memory
• Further Reading and Resources
Key papers and research on LLMs and working memory
Suggested books and articles on cognitive science and AI

Subjects

Computer Science