Getting Started with ChatGPT: Understanding Large Language Models in Simple Terms
I’m often asked, “What exactly is a language model, and how does ChatGPT work?” The best way to explain it is through real-world examples—like the government document collections at UC Berkeley’s Institute of Governmental Studies Library (IGSL).
These collections hold thousands of historical reports, policies, and planning documents—valuable, but often complex to navigate. Imagine trying to sift through hundreds of pages of legal or policy language just to find a key detail. That’s where ChatGPT and large language models (LLMs) come in. They help researchers, policymakers, and everyday people make sense of massive amounts of text quickly.
At its core, ChatGPT is powered by two key components that make this possible:
1. Large Language Models (LLMs): The Brain of AI
Think of a large language model (LLM) as an enormous knowledge repository—like a digital library filled with information from books, articles, websites, and more. In many ways, it’s like a search engine but much more advanced. Instead of just pointing you to links, an LLM processes language, understands patterns, and generates human-like responses based on the data it has been trained on.
However, there’s one major factor to keep in mind:
The LLM is only as good as the data it has. If it doesn’t get updated with fresh information, it can become outdated.
It doesn’t “think” like a human. It recognizes language patterns but doesn’t have personal opinions, experiences, or emotions.
It can infer meaning from text, but sometimes those inferences may not be completely accurate.
2. ChatGPT: The Interface That Brings LLMs to Life
If an LLM is the massive digital library, ChatGPT is the librarian that helps you navigate it. The LLM stores all the knowledge, but ChatGPT allows you to interact with it in a conversational way.
With ChatGPT, you can:
Ask questions in natural language and get detailed responses.
Refine your search by rewording or specifying details.
Use it for research, brainstorming, summarization, and more.
For example, if you simply ask, “Tell me about language,” ChatGPT might return broad results about human speech, programming languages, or even how animals communicate. But if you refine your prompt to something like “Explain ancient Egyptian languages spoken in 2000 BCE,” ChatGPT will adjust its response accordingly.
This ability to shape the AI’s response based on the way you ask your question is where prompt engineering comes in.
Bridging the Gap: The Power of Prompt Engineering
To get the best results from ChatGPT, you need to ask the right questions in the right way. This is called prompt engineering—the skill of creating effective prompts to get useful and accurate responses.
A clear, well-structured prompt leads to better, more precise answers.
A vague or unclear prompt results in broader, sometimes less relevant responses.
By understanding how to structure your prompts, you can maximize the accuracy and usefulness of ChatGPT’s output—whether you're conducting research, analyzing government documents, or just exploring new ideas.
Putting It All Together
When you use ChatGPT, you’re interacting with a system that combines:
The LLM (the brain) – where all the knowledge is stored.
ChatGPT (the voice) – the tool that helps you access and interact with that knowledge.
Prompt engineering (the key) – the method that helps you communicate effectively with AI.
Whether you're a researcher, policymaker, or just someone curious about AI, understanding these three components will help you unlock the full potential of ChatGPT in your work.
My name is Nick, and I enjoy teaching and speaking about the intersection of research, ChatGPT, and prompt engineering. My work focuses on developing easy-to-use frameworks and strategies that ensure AI doesn’t just generate answers, but also verifies and checks itself—helping researchers use ChatGPT more effectively and responsibly. If you have questions, need help setting up, or want to improve your prompts, feel free to reach out—I’d love to help!