Generative AI is a term applied to tools that create new things using mathematical models. Many generative AI tools have a very simple conversational interface that make them very accessible to a non-specialist audience, and developers have built interfaces that allow AI to be incorporated into existing tools like web browsers, word processors and image editing software. This is a rapidly changing area.

Generative AI tools, such as ChatGPT, Microsoft’s CoPilot (formerly Bing), and Google Bard, trained on vast databases to write coherent text in response to users’ prompts, are widely available, increasingly embedded in standard software packages, and already deployed across many professions. Such tools have the potential to perform an assistive role in routine language (and image) based activities common to higher education, but they also reproduce cultural stereotypes and bias, produce inaccurate statements, make things up (‘hallucinate’), have an
environmental impact and constitute a potential threat to the academic integrity of assessments.

What is generative AI?

A robot sitting on a bench reading a book It is helpful to think about what AI is to help set the context for its safe and effective use. It generally refers to technologies that are designed to accomplish tasks that would typically require some level of human intelligence. It has been around for some time, but what has really accelerated it to the forefront of people's minds is generative AI.

Generative AI represents a subset of AI. It uses algorithms designed to produce outputs that normally require some level of human creativity, for example from text, images, videos and even music.
Algorithms are step-by-step instructions or rules to perform a task or solve a problem, algorithms allow automatic processes and decision-making possible with computers.
Machine Learning (ML) is a type of AI that allows machines to learn from data and improve performance over time. It is used in many applications, such as image recognition (many of you may have this on your phone, also used in digital exam proctoring, and increasingly and controversially used by police with the public), natural language processing(a type of AI that gives machines a certain ability to interpret language – so virtual assistants, chatbots i.e. ChatGPT), and predictive analytics (used in Learning Management Systems).
Large Language Model Chat GPT4, the generative AI behind Microsoft’s Bing, is a class of machine learning natural language processing models known as Large Language Models, a type of AI trained on a large amount of text data that can generate new text in similar style and tone to the original.

Murray, M. (2023) 'A student guide to generative AI micro-course'. Available at: https://dcad.awh.durham.ac.uk/generativeai/#/ (Accessed: 3 May 2024).

Image: Andrea De Santis on Unsplash