Lesson 3: The AI Dictionary (Only the Words That Actually Matter)
Estimated time: 10 minutes
Difficulty: Beginner
Introduction
AI has a jargon problem.
Open any article about AI and you'll hit a wall of buzzwords: "LLMs," "fine-tuning," "tokens," "context windows," "neural networks." It's like everyone's trying to sound smart instead of being clear.
Here's the truth: you only need about 10 terms to understand 90% of what you'll encounter in the AI world. That's it. Ten words.
This lesson is your decoder ring. By the end, you'll recognize these terms everywhere, know exactly what they mean in plain English, and stop feeling lost when people throw around AI vocabulary.
Why Bother Learning Any Terms at All?
Fair question. Why not just use the tools and ignore the terminology?
Because understanding these 10 words helps you:
- Make sense of articles, tutorials, and tool descriptions
- Know what features matter when comparing AI tools
- Ask better questions when you're stuck
- Not feel like everyone else is in on a secret you're missing
Think of it like learning to drive. You don't need to know how a transmission works. But knowing "brake," "accelerator," and "steering wheel" is pretty essential.
Same here. These 10 terms are your controls. The rest is optional.
The 10 Terms You Actually Need
1. Prompt
What it means: The instruction or question you give to an AI tool.
In plain English: It's just what you type in. "Write me a thank-you email" is a prompt. "Summarize this article" is a prompt. Your prompt is your input.
Why it matters: Better prompts get better results. Most of what people call "AI magic" is really just learning how to write clear, specific prompts. (That's the entire focus of Course 2.)
Example: Instead of "recipes" (vague), try "Give me 3 quick dinner recipes using chicken, rice, and broccoli for 2 people" (specific prompt).
2. Model
What it means: The AI "brain" that powers a tool. GPT-4 is a model. Claude is a model. Gemini is a model.
In plain English: Think of it as the engine inside a car. ChatGPT is the car. GPT-4 is the engine. Different engines have different power and efficiency.
Why it matters: When someone says "I got better results with GPT-4 than 3.5," they're comparing different versions of the engine. Newer models are usually smarter and more accurate.
Example: ChatGPT runs on GPT-4 (or GPT-3.5 for free users). Claude runs on Claude 3. Same concept, different companies.
3. Hallucination
What it means: When AI confidently makes something up.
In plain English: AI will invent fake statistics, nonexistent books, made-up people, false facts - and sound completely sure about it. It's not lying or broken. It's generating what sounds plausible based on patterns, even when it's wrong.
Why it matters: This is the #1 thing that trips people up. You ask for sources, AI gives you five citations that look perfect, and none of them exist. You need to verify anything important.
Example: Ask AI for "recent studies about [obscure topic]" and it might generate realistic-sounding citations to papers that were never published.
4. Token
What it means: How AI measures text. Roughly 1 token = ¾ of a word.
In plain English: AI doesn't count words. It counts tokens. "Hello there" is about 2 tokens. A 1,000-word essay is roughly 1,300 tokens.
Why it matters: Most AI tools have limits on how much text you can send or receive, measured in tokens. If a tool says "8,000 token limit," that's roughly 6,000 words. You don't need to calculate it yourself - just know it's how the limit is measured.
Example: ChatGPT's free version has a limit of about 4,000 tokens per conversation. That's why really long conversations eventually "forget" earlier parts.
5. Context Window
What it means: How much text the AI can "remember" at once in a conversation.
In plain English: Think of it like a whiteboard. You can keep writing on it, but eventually you run out of space and the stuff at the top gets erased. That's the context window - once it's full, older messages fall off.
Why it matters: If you have a really long conversation with AI, it eventually forgets what you said at the beginning. This is why sometimes it seems to "lose track" of what you were talking about.
Example: If a model has a 16,000 token context window, it can hold about 12,000 words of conversation history before it starts forgetting the beginning.
6. Fine-tuning
What it means: Training an AI model on specific data to make it better at one particular thing.
In plain English: It's like taking a general-purpose AI and giving it a specialized education. You could fine-tune a model on medical textbooks to make it better at answering health questions, or on legal documents to make it better at contracts.
Why it matters: You probably won't fine-tune a model yourself (it's technical and expensive). But you'll see tools advertise that they're "fine-tuned for marketing" or "fine-tuned for customer support." That means they've been specialized.
Example: A company might fine-tune GPT-4 on their customer support history so it answers questions in their specific style and knows their products.
7. LLM (Large Language Model)
What it means: The technical name for text-based AI like ChatGPT.
In plain English: "Large" = trained on tons of text. "Language Model" = it works with words. It's the official category name for AI that reads and writes.
Why it matters: When you see "powered by an LLM," it just means it's using a text AI like GPT-4 or Claude under the hood. You can mostly ignore this term, but it's everywhere in articles, so now you know.
Example: ChatGPT, Claude, and Gemini are all LLMs. Midjourney (image AI) is not.
8. Generative AI
What it means: AI that creates new content - text, images, music, video.
In plain English: "Generative" = it generates stuff. As opposed to AI that just analyzes or sorts things you already have.
Why it matters: This is the category of AI that exploded in 2022-2023. Tools like ChatGPT (generates text) and Midjourney (generates images) are generative AI. That's different from older AI that just categorizes emails or recommends products.
Example: ChatGPT writing an email is generative AI. Gmail sorting your inbox isn't.
9. API (Application Programming Interface)
What it means: A way for apps and websites to talk to AI behind the scenes.
In plain English: When an app says "powered by GPT-4," it's using an API to connect to OpenAI's servers. The app sends your input, gets a response, and shows it to you. You're not directly using ChatGPT, but the app is using it for you.
Why it matters: You don't need to use APIs directly unless you're building something. But understanding that tools can plug into AI models helps you recognize when different apps are using the same underlying technology.
Example: Notion AI, Canva AI, and dozens of other tools use OpenAI's API to power their features. They're all accessing GPT-4, just through different interfaces.
10. Training Data
What it means: The text, images, or information the AI learned from.
In plain English: Before an AI can answer questions, it needs to study. Training data is what it studied - books, websites, conversations, articles, whatever the company fed it.
Why it matters: This explains why AI knows a lot about common topics (tons of training data) and less about niche subjects (limited training data). It also explains why AI's knowledge has a cutoff date - it only knows what was in its training data.
Example: GPT-4's training data includes information up to April 2023 (as of this lesson). Ask it about events after that date and it won't know unless it has web access enabled.
Three Terms You Can Safely Ignore (For Now)
People love to throw around technical terms to sound impressive. Here are three you'll hear that you don't need to understand to use AI effectively:
Neural network - The technical architecture behind AI. Sounds important. Irrelevant to daily use.
Transformer - A specific type of neural network. Cool. Still doesn't matter for you.
Reinforcement learning - A training technique. Neat. Also not your problem.
If someone uses these in conversation, just smile and nod. They're showing off. You're here to get stuff done.
Your Quick Reference Card
Here's the cheat sheet. Bookmark this or screenshot it:
- Prompt = what you type in
- Model = the AI engine (GPT-4, Claude, etc.)
- Hallucination = AI making stuff up
- Token = how AI measures text (~¾ of a word)
- Context window = how much AI remembers in one conversation
- Fine-tuning = specializing AI for one task
- LLM = text-based AI (like ChatGPT)
- Generative AI = AI that creates new content
- API = how apps connect to AI
- Training data = what AI learned from
That's it. Ten terms. You now speak AI.
Key Takeaways
- ✅ You only need ~10 terms to understand 90% of AI conversations
- ✅ A "prompt" is your instruction, a "model" is the brain, a "hallucination" is AI making stuff up
- ✅ Technical terms like "neural network" sound impressive but don't help you use AI better
- ✅ Most AI jargon is people showing off - focus on what's actually useful
- ✅ Understanding these terms helps you compare tools, troubleshoot issues, and ask better questions
Try It Challenge
🎯 Spot the terms in the wild (takes 5 minutes):
- Find any article about AI (search "ChatGPT news" or "AI tools")
- Skim through and count how many of the 10 terms you spot
- Notice how much more you understand now
Bonus: If you spot a term that's NOT on this list, copy it and ask ChatGPT: "Explain [term] like I'm a smart beginner learning about AI." You'll get a plain-English definition in seconds.
This is how you keep learning. When you hit a new word, just ask AI to explain it. The best teacher for learning AI is AI itself.
Quick Check
Question 1: What does "hallucination" mean in AI?
A) The AI is dreaming
B) The AI confidently generates false information
C) The AI shows you images
D) The AI crashes
Answer: B) The AI confidently generates false information
AI hallucinations happen when the tool generates information that sounds correct but is completely made up - fake statistics, nonexistent sources, fabricated details. It's not lying or broken. It's doing what it was designed to do: generate text that matches patterns. Sometimes those patterns lead to plausible-sounding fiction. This is one of the most important things to understand about AI because it looks so convincing.
Question 2: What is a "prompt"?
A) The AI's response to you
B) A type of AI model
C) The instruction or question you give to an AI tool
D) The AI's training data
Answer: C) The instruction or question you give to an AI tool
A prompt is simply what you type into an AI tool. It's your input - your question, instruction, or request. The quality of your prompt directly affects the quality of the response you get. That's why "prompting" (writing good instructions for AI) is such an important skill. Better prompts = better results. Course 2 digs deep into this.
Next up: Lesson 4 - What AI Can and Can't Do (Setting Real Expectations)