Introduction to Generative AI
We have not written any code yet. This chapter only explains what Generative AI is, because the rest of the LangChain course assumes you have seen a chat tool at least once.
What is Generative AI?
Older software classified things: spam or not spam, cat or dog in a photo. Generative AI creates output — text, images, code, audio — from a request you type in.
Open ChatGPT, write a question, get an answer. That question is the prompt. Claude and Gemini follow the same pattern with different models behind them.
Rough mental model: autocomplete on your phone, but scaled up to full paragraphs and trained on far more text.
An Example
A Class 8 student stuck on Life Science might type:
- Plants use sunlight, water and CO₂ to make glucose.
- Chlorophyll in the leaves absorbs the light.
- Oxygen is released — the gas we breathe.
The answer is assembled on the spot. It is not pulled from one saved webpage. The model has seen many explanations during training and drafts a new version for this prompt.
Strengths and Limits
Good for first drafts — emails, short notes, rough code, breaking a chapter into simpler points. Handy when you are stuck on wording or structure.
Poor substitute for fact-checking. Dates, numbers, citations, and exam answers still need your own review. Models can sound certain while being wrong; that is often called a hallucination.
Training data has a cutoff, so do not expect live scores, today's headlines, or post-cutoff events unless the tool has search built in.
Where You See It Today
- ChatGPT, Claude, Gemini
- Gmail / Outlook sentence suggestions
- GitHub Copilot in the editor
- Canva, Photoshop, and similar image tools
Most chat products sit on a large language model (LLM). The next lesson covers OpenAI specifically — ChatGPT, the API, and keys. LangChain comes later, when we wire models into Python.
What's Next
What OpenAI is, how ChatGPT differs from the API, and where your API key lives.