What is generative AI?

Before we look into how generative AI works and what it can do, it’s important to understand what artificial intelligence is in general.

This short video introduces you to the basics of AI and explains how generative AI is different from the AI you might already be using every day.

• Narrow AI powers many tools we use daily, like voice assistants and smart suggestion tools. They don’t think or understand like humans—they just follow rules or patterns to complete a job.
• Generative AI goes a step further and uses the patterns from existing data to generate something new, like text, images, and code, based on your input.

2. How does it work?

Now that you know what generative AI is, let’s take a look at how it works.

In this video, we’ll look at the models behind generative AI, how they’re trained, and the ethical concerns around the way in which they use information and their impact on the environment.

Generative AI uses a data-driven approach, meaning it learns from patterns, relationships, and structures, from vast amounts of data, rather than following predefined rules. Unlike traditional rule-based systems that rely on explicit instructions, generative AI models identify statistical patterns in the data to make predictions, generate content, or solve problems. This allows them to adapt to new inputs and contexts more flexibly than if they were following more rigid rule-based systems.

• Generative AI models are trained on vast datasets using pattern recognition.
• As noted earlier, these systems do not understand meaning in a human sense – they predict based on training data.
• Training these models requires huge amounts of energy and data, often scraped without consent.
• Ethical concerns include copyright, privacy, and environmental sustainability.

3. Bias and misinformation

Although generative AI can be a really useful tool, it’s important to remember that it is not neutral.

In this video, we’ll explore how the data used to train AI models can shape the outputs, and why it’s always important to approach these tools with a critical mindset.

• AI outputs reflect the biases and gaps in the training data. They can reinforce stereotypes, exclude diverse voices, or present misinformation.
• Engineers can influence outputs, sometimes improving them, sometimes accidentally or intentionally introducing new biases.
• Always treat generative AI as a tool, not a source of truth. Always follow up and use your own judgement.

It’s important to always reflect on any generative AI responses and consider potential bias.

Use this checklist below to critically evaluate AI-generated content by identifying bias, misinformation, and missing perspectives. It will help you think beyond the surface and assess how inclusive, accurate, and academically reliable the information given really is.

4. Data protected submissions

Generative AI tools also raise important questions about privacy, data use, and ethics. It’s important to understand how generative AI tools use your inputs so you can make informed choices and ensure that you use generative AI tools in a responsible way.
In this video, we’ll look at how your prompts, input and uploaded documents can be used to train AI models and how aspects of what you upload could be made available to others. We will also look at how UWE Bristol is addressing some of these concerns through Microsoft Copilot.

It’s important to remember that many free third-party tools will use your inputs to train their models meaning that aspects of the content that you upload can be made available to others. At UWE, everyone has access to the premium data protected version of Microsoft Copilot meaning that your inputs are protected and won’t be shared outside UWE.
We will look more closely at Copilot and how you can access it later on in the course.