AI applications and responsible usage

AI around you

Generative AI and large language models

Generative AI creates new content: text, images, audio or code. Large language models (LLMs) are deep learning models trained on very large amounts of text. Given a prompt, they generate a response one small piece (a token) at a time, each time predicting a likely next token based on patterns learned in training.

That explains both their strengths and their weaknesses:

• They are excellent at drafting, summarising, explaining and rephrasing.
• They do not look facts up unless connected to a search or data source, so they can produce fluent, confident statements that are wrong. This is often called hallucination.
• Their knowledge stops at their training data, unless the system gives them newer information.

Using AI from code

Applications usually use AI through an API: the program sends data (for example, a prompt) over HTTPS and receives the model's output as JSON, just like the REST APIs in the Full-Stack path. Production systems add checks around that call, such as validating the output, limiting what data is sent and logging usage.

Responsible AI

AI systems affect real people, so they must be built and used with care.

Bias and fairness. Models learn from historical data, including its biases. A hiring model trained on past decisions that favoured one group can repeat that unfairness. Teams test performance across different groups and keep humans in charge of important decisions.

Privacy. Don't paste passwords, personal details of other people, customer data or confidential company information into public AI tools. Prompts may be stored or reviewed by the provider. Follow your organisation's policy and data-protection law.

Accuracy. Treat AI output as a draft. Check facts against reliable sources, run and test generated code, and never deploy code you don't understand.

Transparency and accountability. Tell people when content or decisions involve AI where it matters, and keep a person responsible for the outcome.

Security. Applications that pass user text to a model must defend against prompt injection, where crafted input tries to make the model ignore its instructions or reveal data.

A practical checklist for students

1. Use AI to learn, for example by asking for an explanation of an error, then write the solution yourself.
2. Ask for sources or reasoning, then verify them.
3. Never share private or confidential data.
4. Follow your institution's rules on AI use for assignments and exams.
5. Credit AI assistance when your course or employer requires it.