CSTAT Spring 2026 - Evolution of Language Models

Speaker: Dr. Wenjuan Ma, CSTAT
Date: March 27, 2026
Title: The Evolution of Natural Language Processing: From Rules to Large Language Models

This seminar offers a clear, accessible journey through how computers learned to understand and generate human language, tracing the evolution of Natural Language Processing (NLP). We begin with the early, awkward phases where programmers had to write strict, rigid rules for language, and then look at the first major shift to recognizing simple patterns and connections in text. The real revolution started when models learned to assign a unique, meaningful code to every word, allowing them to truly grasp context. We will focus on the most important recent invention—the "Attention" mechanism—which allowed AI to process entire paragraphs instantly, rather than reading word-by-word. This final innovation unlocked the creation of today’s powerful, enormous language tools, known as Large Language Models (LLMs) like the Gemini and GPT series, which can now perform complex tasks and communicate across multiple forms of media.