Skip to Main Content

GenAI

Generative AI information

What can ChatGPT help me with?

While Large Language Models (LLM) like ChatGPT can provide valuable assistance, they are no replacement for human guidance and support. 

Further, there are reliability and ethical issues surrounding these models and their outputs which are explained on subsequent pages of this guide. Always evaluate the information in front of you. ChatGPT has no truth filter!

ChatGPT can help with the following:

  • explain or define terms, concepts and topics
  • summarise information or find key points
  • proofread for grammatical and spelling errors
  • create practice quiz questions
  • create practice interview questions
  • brainstorm new projects
  • evaluate mathematical problems, computer code and formulas

In order to get useful results you will need to build the right questions and set parameters for the reply or dialogue you are seeking. This concept is also called prompting, prompt design or prompt engineering.

How to prompt ChatGPT

Knowing how a Large Language Model (LLM) like ChatGPT works can be useful for creating prompts and understanding its limitations. LLMs use their vast amounts of training data to try and predict words. These predictions are based on learnt patterns and content around words. In the examples below, the words 'colourful' and 'garden' in the prompt help ChatGPT predict the missing word. You could almost think of it as a very sophisticated autocomplete. If confronted with less patterns or content in the prompt the output is less specific.

 

The examples above demonstrate that the prompt you enter will influence the reply you will get. 

 

What makes a good prompt?

A good prompt generates basic text that needs tweaking; a great prompt is specific and useful to the user. Great prompts have some or all of the following elements:

  • Give the AI a persona. 
  • Be specific, clear, and concise. Define what you are looking for.
  • Provide content and avoid ambiguity. Open-ended prompts are not narrow enough to generate a meaningful response. 
  • Provide specific constraints, such as a target length or tone, to guide the model's response.  
  • Make it conversational and use complete sentences. 
  • Additionally, a good prompt should have a clear and achievable goal that the model can work towards.

Try the following two prompts in ChatGPT and see the difference: 

  • What are some ideas for a birthday party? 
  • You are a high-end birthday planner. Plan a birthday party for a five-year-old girl with a budget of $5,000. Her favourite colour is pink and she likes dinosaurs. There will be 20 guests of the same age and the party will be held outside in May in Brisbane, Australia. 

(by QUT, licensed under CC BY-NC-SA)

Prompting examples

  • Tell me in simple terms how the 'Theory of relativity' works. Use no more than 50 words.
  • Create a 10 question quiz about the 'Theory of relativity'. Provide feedback for my answers.
  • Act as a speech pathologist. Provide an assessment of a three year old child based on the speech sample: "Mummy get in ka."
  • Summarise in one sentence: [paste text]. / Summarise in one sentence preserving information about dates / people / locations.

Test your prompting knowledge