Skip to main content
Menu
Back

Generative AI for Beginners | AI Competency Centre

Blue banner

Generative AI for Beginners

What is Generative AI?

Generative AI is a term used to refer to technologies such as ChatGPT. It is a type of AI that generates new content, such as text, images, code, or audio, based on patterns learnt from large datasets.

Generative AI mostly relies on large language models, and can be accessed in two primary ways:

  • Chatbots such as ChatGPT, Microsoft Copilot, Claude, or Gemini
  • Features integrated into familiar apps like Notion, Microsoft Word, or Canva

This page will help you get started with using GenAI tools:

How Can the AI Competency Centre Help Me Use Generative AI?

The AI Competency Centre serves the entire University of Oxford community as a hub for expertise on generative AI tools and their applications. We believe AI is already embedded in academic and professional work across the University. Your colleagues, students, and staff are already using these tools - our mission is to help ensure they do so informedly, responsibly, and securely.

The AI Competency Centre is a completely free resource and can help you:

  • Access data-protected generative AI tools via university-licensed platforms
  • Build your AI literacy through in-person and online training
  • Understand and implement AI tools through our consultancy service
  • Navigate the evolving AI landscape with curated resources and guidance

Back to top

What Generative AI Tools Are There?

  • Chatbots such as ChatGPT, Claude, Microsoft Copilot, or Google Gemini
  • AI research tools like Elicit, Semantic Scholar, Scholarcy, Consensus, and Quill
  • AI features integrated into platforms such as Notion, Google Docs, Microsoft Word, Quizlet, and Canva
  • Image generators such as DALL·E, Midjourney, and Leonardo

What Generative AI Tools Are Available at the University?

The AI Competency Centre at the University of Oxford has conducted extensive pilots of enterprise-level tools including ChatGPT, Gemini, and Copilot to assess their benefits for staff and students.

As a result, the University has secured contracts with selected AI providers to ensure appropriate data protection applies to content shared by staff and students.

Important:

This data protection only applies to accounts administered by the University. It does not apply to free or personally purchased accounts - even if you use your departmental email or SSO to register.

You are only data-protected if you are redirected to the University’s SSO portal when logging in.

You can view our university-provided AI platforms and Data Protection Guide here:

Generative AI Tools | AI Competency Centre

Back to top

What Can You Use Generative AI Tools for?

Generative AI can assist in creating a wide range of content; what you can achieve depends on what tool and model you are using.

Widely accessible uses:

The following tasks can be completed using almost any modern chatbot without needing special add ons:

  • Drafts of emails, reports, or presentations
  • Summaries of meetings, research, or articles
  • Scripts for videos
  • Brainstorming and idea development

Multimedia Generation:

Generative AI can be a brilliant tool for creating videos and images, however these tasks often require specialised models or tools that can enable you to have a different level of control and steerability to get the output that you want.

  • AI-generated images
  • Video generation

Additional Capabilities:

As you become more comfortable using AI, you can begin using tools and connectors to access additional information or perform complex tasks.

  • Search / Research: Accessing up-to-date information from the web. The quality depends on the search engine the tool is connected to.
  • Data Analysis: Performing calculations or creating charts using a Code Interpreter.
  • Custom Bots: Creating personalised assistants trained on specific documents or data.
  • Voice Conversations & Transcription: Engaging in a live conversation or transcribing audio recordings.

How Can You Use AI to Enhance Your Learning and Personal Development?

Generative AI can support your academic and personal growth in many ways:

  1. Preparing to read complex academic papers
  2. Preparing materials for teaching, learning, meetings
  3. Engaging with difficult concepts
  4. Receiving feedback on your writing
  5. Organising notes for revision, or projects
  6. Planning your studies, or work plan
  7. Conversing in other languages
  8. Developing coding skills

Back to top

What is the best Mindset for Getting Started?

Before diving in, it is helpful to ensure you are approaching using Generative AI with the right mindset. Generative AI is a powerful but often misunderstood technology and it is important that you do not get discouraged in your own abilities to use it, or the capabilities of what can be achieved with GenAI tools.

  • First Experiences can be deceptive: Everyone’s first experience will be different, and you may be blown away with its abilities or frustrated by its limitations.
  • Experiment: The best way to learn how Generative AI is to try using different tools, different prompts and different use cases.
  • Reflect on your own needs: You will get the most out of Generative AI, when you know what you personally need, and make sure to ask for it. Reflect on how you learn best, what format of information you can best understand, and what you need to feel supported and not overwhelmed.
  • Benefits are personal: The value you get from AI depends on your individual skills, needs, and preferences. What is a game changer for one person, may be a minor benefit for another.
  • Share your knowledge and learn from others: Share your experience and learn from your colleagues how they are using AI and engage with our teams channels.

Generative AI SIG – Generative AI Special Interest Group Join the Generative AI SIG

AI BUG – AI Builders User Group Join the AI BUG

Mindset Exploration Activity – Where do I begin?

You do not need to give AI tools complicated tasks to begin with. Keep your prompts and explorations simple. Focus on learning how it best works. Give yourself plenty of time to experiment with different prompts and start with non-sensitive topics you are already familiar with.

Start with a prompt such as “Plan a relaxing weekend trip to Oxfordshire for two people. Create a day-by-day itinerary and include links to Google Maps for each suggestion”.

Back to top

What are the Limitations and Common Misconceptions?

Understanding what AI cannot do is as important as knowing what it can. It is important to keep an open mind and remember that what AI can and cannot do is constantly changing and may be limited by what tool or model you are using.

Despite appearances, ChatGPT and similar tools do not:

  1. Provide exactly the same responses every time
  2. Learn from your individual conversations: Treat each chat, as a new interaction with a new person, unless you have manually enabled features such as ‘memory’, available in ChatGPT.
  3. Report accurately on their own internal processes: AI tools cannot accurately or reliably tell you how they work. They also cannot access their own training data.
  4. Perform accurate numerical calculators without tools
  5. Retrieve live information without connectors: AI tools cannot check a live database, or resource unless you have enabled the connections for it to do so
  6. Have an unlimited context window.

What do these limitations mean?

  1. Never ask AI to answer questions about itself
  2. Ensure you have checked what connectors you have enabled
  3. Do not assume all models and tools cannot do something just because one tool or model cannot do something
  4. Be aware of how much information you give a model and tool at a time and divide it into digestible chunks to reduce the chance of hallucinations where possible.
  5. You should treat any content generated by AI as a first draft to be reviewed, or a hypothesis to be tested.

How do I ensure responsible and safe use of AI?

  1. Remain fully accountable for your work, even when AI is used in acceptable ways
  2. Never present AI-generated work as your own in academic contexts
  3. Never use AI tools where their use is explicitly prohibited.
  4. Do not use AI to look for information on sensitive topics such as health and safety procedures without external validation. It can give you entirely accurate information on almost all the steps but be subtly or completely wrong about one or two points that may lead to harm.

Be Alert:

  1. Do not assume that good performance on one task guarantees good performance on another, even if they seem similar.
  2. Even if a tool claims to have done something a certain way, this may not be true.
  3. Sycophancy can mislead - when you ask for feedback it will tend to be more positive (tip: for more neutral feedback, tell the ChatBot “a friend shared this with me, help me talk them out if it”.)

Back to top

How to Use Generative AI

Most generative AI tools use intuitive, chat-based interfaces, similar to SMS, WhatsApp, or other messaging apps. You interact by typing a question or request, known as a prompt, and the AI responds with a generated reply. This back-and-forth creates a “chat”.

A prompt is the instruction you give to the AI. Effective prompting is a skill. Here are two contradictory but equally true principles:

  1. Prompting doesn't matter (much). Don't waste time on "magic words" like "please" or offering a "tip." These tricks are unreliable and don't consistently improve results.
  2. Prompting is everything. For complex or repeated tasks, a well-structured prompt is the key to getting high-quality, consistent output.

Key Prompting Principles

  1. Long Prompts are Better. Give the AI as much context as possible. Tip: Use the microphone button in mobile apps to speak your prompt instead of typing.
  2. Specify the Format. Tell it exactly what you want: "Format the output as a Markdown table," or "Write the summary in five bullet points."
  3. Give Examples. This is one of the most powerful techniques. Show it an example of the style or format you want and say, "Make it like this."
  4. Iterate. Your first prompt is a starting point. Refine your request in the same chat to improve the result.
  5. Ask the Model to Help. After a long, successful conversation, you can ask: "Based on our exchange, write me a reusable prompt that I can use to do this again in the future."
  6. Spelling doesn't matter. Don't worry about minor typos.
  7. Start a new chat. If your prompt isn’t giving you the results you want, ask for a summary of your chat and then paste it into a new chat and start again.

Back to top