Learn how general AI models are customized for specific uses like coding, customer service, or medical diagnosis
Part of Understanding ChatGPT & Modern AI
Educational content slides
Pre-training gives AI knowledge. It knows facts, can write code, understands language.
But raw pre-trained models are chaotic—they might be rude, unhelpful, or unsafe because they learned from all of the internet, including the bad parts.
Fine-tuning teaches AI how to behave: be helpful, follow instructions, refuse harmful requests, and match desired personality.
Test your understanding with this quiz.
GitHub Copilot suggests code completions extremely well, while ChatGPT is better at conversational help. They're both based on similar GPT models. What explains the difference?
Complete this exercise and get AI-powered feedback.
Now that you understand fine-tuning, what would you create?
Key insight: General AI + specialized fine-tuning = expert assistants. This is how most AI products are built.