Class 1 (What an LLM Actually Is)

What Is an LLM

My name is Wesley S. Bonafé. Think of me as Wes. I spent 50 years as a mechanical engineer. I designed systems for buildings. I led engineering teams. I passed professional licensing exams in two states.

None of that prepared me for artificial intelligence.

When I first started using AI tools, I did not know what I was working with. I did not understand how they worked. I did not understand why they sometimes gave me wrong answers. I did not understand how to get better results out of them.

That changed when I started learning what was actually happening under the hood.

This session covers one thing: what is a Large Language Model, which is what most AI tools are built on.

You do not need a technical background. If you can read and follow instructions, you can understand this.

 

 

Start With What You Already Know

Before we talk about what a Large Language Model is, think about how you learned to talk.

You did not read a grammar book at age two. Nobody handed you a list of rules. You listened. You repeated. You made mistakes. People corrected you. Over time, patterns became clear. You learned that certain words go together. You learned that some sentences sound right and some sound wrong, even if you could not explain why.

That is, in very simple terms, how a Large Language Model learns.

It does not read a rulebook. It reads text. Enormous amounts of text. Books. Articles. Websites. Conversations. Code. Scientific papers. Instruction manuals. And from all of that reading, it learns patterns.

It learns which words tend to follow other words. It learns which ideas tend to connect to other ideas. It learns how people ask questions and how answers are typically structured.

That learning is the foundation of every AI tool you have likely used or heard of.

What “Large” Actually Means

The word “large” in Large Language Model refers to scale. It refers to how much text the model was trained on and how many internal connections it built while learning.

To give you a sense of the scale: some of these models have been trained on hundreds of billions of words. That is more text than any human being could read in several lifetimes.

What “Language” Actually Means

The word “language” refers to the fact that the model works with text. It reads text. It produces text. Everything goes in as words and comes out as words.

What “Model” Actually Means

The word “model” refers to the fact that it is a mathematical system. It is not a database that looks up answers. It is a system that has learned patterns and uses those patterns to generate responses.

That distinction matters. A database retrieves. A language model generates. The response it gives you is not stored somewhere waiting to be found. It is produced, word by word, based on patterns the model learned during training.

Why That Matters to You

Here is where this becomes practical.

Because an LLM generates responses rather than retrieving them, it can be wrong. It may produce an answer that sounds completely confident and completely reasonable, and it may still be incorrect.

This is not a flaw in the same way that a broken engine is a flaw. It is a characteristic of how the system works. The model does not know what it does not know. It generates based on patterns, and sometimes those patterns lead it in the wrong direction.

It has been my experience that people who do not understand this get burned. They take the output at face value. They act on it. Then they find out the answer was wrong, outdated, or simply made up.

The other thing that matters is this: an LLM does not remember you between sessions unless you give it a way to do so. Each conversation, it starts fresh. It does not know who you are, what you have discussed before, or what your situation is, unless you tell it.

That limitation has a practical solution, and we will cover it in a later session.

What an LLM Is Good At

Despite those limitations, an LLM is genuinely useful. Here is a short list of what it does well:

      • Drafting written content: emails, letters, reports, outlines, scripts

      • Explaining concepts in plain language

      • Summarizing long documents

      • Answering general knowledge questions

      • Helping you think through a problem by asking it to consider different angles

      • Translating between languages

      • Writing and reviewing basic computer code

      • Helping you organize your thoughts into a structured format

    What these tasks have in common is that they involve working with language. That is the lane an LLM operates in. It processes text and produces text.

    It is not a search engine. It is not a calculator, though it can do basic math. It is not a database of verified facts. It is a language-based reasoning tool, and when you use it for language-based tasks, it tends to perform well.

    A Large Language Model is a system trained on enormous amounts of text. It learns patterns from that text and uses those patterns to generate responses.

    It does not retrieve stored answers. It generates them.

    That means it can be wrong. It does not remember you between sessions. And the quality of what it produces depends largely on the quality of what you give it.

    In the next session, we will cover a simple way to think about an LLM’s strengths and limits, and how to give it the background information it needs to actually help you.

    Thank you for your time.