What is generative AI? Creating content from learned patterns
Generative AI produces content such as text, images, audio or code using patterns learned from data. It differs from a system that only assigns a label or predicts a numerical value. A generated answer can sound convincing while still being unsupported by the information available to it.
SchoolWhool's current learning path covers Python, Data Tools, Maths, Statistics and classical Machine Learning. Open lessons' parts and key terms are free to read.
Generation, prediction and retrieval answer different questions
How do language and image generators work?
Many text generators use language models that predict tokens conditioned on preceding context. Generating a longer response repeatedly extends that context with selected tokens. The training objective does not by itself make each statement a verified fact.
An image generator can use a different process. In diffusion-based systems, generation involves a sequence of denoising steps, often guided by a text prompt. A large language model is therefore one kind of generative system, not the definition of the whole field.
Prompting, retrieval and fine-tuning
A beginner project: draft answers from a small FAQ
Define the evidence
Write a small fictional FAQ with facts you can check manually. Use an exercise dataset rather than private customer documents.
Build a search baseline
Retrieve the relevant entry and show it directly. This gives you a useful reference before adding generated wording.
Specify unsupported questions
Include questions the FAQ cannot answer, as well as paraphrases and conflicting requests. Define when the application should say that it lacks enough information.
Evaluate the whole answer
Check factual support, omissions and source references. Keep the original query and retrieved text so you can distinguish retrieval mistakes from generation mistakes.
How should you judge a generative AI demo?
Read the evaluation examples, not just the best output. Ask whether the application handles a missing source, an irrelevant source and a request that contradicts the source. Compare the result with a simpler search or template-based workflow.
In the FAQ exercise, record unsupported claims separately from formatting mistakes. A response can satisfy the requested format and still state a policy the FAQ never contained. If you revise the prompt, rerun the same cases and inspect any new failures.
How to start learning generative AI
You can explore a small application while learning the foundations. Python helps you handle inputs and outputs; Data Tools helps you inspect datasets; Statistics helps you design comparisons; and Machine Learning explains training objectives and evaluation. Maths becomes increasingly useful when studying the models themselves.
SchoolWhool publishes those five foundation tracks. It does not yet publish a Generative AI or LLM track. Use the learning path to build the skills behind the demo, then continue with specialist model and retrieval material.
Where this guide takes you
- Use the examples to understand the task and its evaluation. They are suggested learning exercises, not measured results or production-ready systems.
- The linked SchoolWhool courses teach the foundations. Specialist deep learning, computer vision, NLP, generative AI and reinforcement learning tracks are not published.
- Open lessons' parts and key terms are free to read. A free account opens quizzes and saved progress, plus practice and other material in the free sample modules.
Questions
Is generative AI the same as an LLM?
No. A large language model can generate text, while generative AI also includes systems for images, audio and other content.
What is a hallucination in generative AI?
It is a commonly used term for generated content that is unsupported or incorrect. A fluent tone or an included citation does not establish factual support.
Does prompting train the model?
Ordinary prompting supplies input at use time without changing learned weights. A provider may have separate data-use policies, which should be checked independently.
Does retrieval remove incorrect answers?
No. Retrieval can provide useful evidence, but the system can retrieve the wrong passage or generate a claim the passage does not support.
Do I need to train a model to build a generative AI application?
No. A learning project can use an existing model. You still need to define the task, handle data appropriately and evaluate the output.
Sources
Technical references and further reading from the authors, educators and maintainers below. Read dates record when these references were checked; the examples and suggested exercises in this guide are SchoolWhool's own.
- Google: machine learning and generative AI overviewRead 2026-10-14
- Hugging Face: LLM courseRead 2026-10-14
- Hugging Face Diffusers: text-to-image generationRead 2026-10-14
Related reading
Natural language processing explainedAI vs machine learning vs deep learningTransfer learning vs fine-tuningHow to learn AIBuild the foundations for this subject
Follow SchoolWhool's learning path through Python, Data Tools, Maths, Statistics and Machine Learning. Start with the skills you need, read open lessons without an account, then sign in free for quizzes and saved progress. Free sample modules also open their practice and study material.
