Computing

  This file is restricted to University members. You must login to download it.

For a video running through these slides, see here.

.

ChatGPT (and other generative AI)

You may have read about ChatGPT and similar generative AI which use large language models to respond to user prompts with statistically probably words and sentences which have the appearance of human generated output.  They are often seen as shortcuts when conducting academic research.  You may even have experimented with such tools or used them to assist with your studies/research.

Current generative AI, such as ChatGPT, cannot and do not think, analyse, reason or argue.  All they can do is sort words into familiar patterns.  This allows them to create persuasive but potentially unsound rhetoric.  Anything produced by the current generation of AI should therefore be viewed with critical scepticism.  AI cannot find or replace genuine research using trusted information sources.

Understanding how to use these tools correctly is essential to avoid breaches such as plagiarism that could impact your successful course progression. 

• You must ensure you appropriately cite and reference any text or output generated by AI in an assignment.
• You must check the legitimacy of the information contained there.
• The final product must be your own work and not just copied from an AI generator.

Examples of usage that might be considered legitimate could be in assisting with planning an assignment, generating ideas for further investigation, or helping find additional search terms or keywords.  You may wish to check with tutors whether such usage is acceptable.

The University has issued a statement on the use of Generative AI such as ChatGPT in assignments and the Library has produced guidance on citing and referencing generative AI.  You might also want to note the general student guidance offered here.

You should be aware that ChatGPT and similar tools invent some academic references.  They are large language models which have been trained which words should appear and in what order - hence similar authors/titles/journals get linked with each other, producing very plausible sounding references that actually don't exist - the words simply have connections to each other within its training data.  Using such fabricated information in your work may be considered academic misconduct so you should always double check your reference lists.

When prompted "How does ChatGPT predict which words come next?" ChatGPT stated that it "predicts which words come next based on the patterns it learns from the training data. When given a prompt or context, the model uses its knowledge of language patterns to generate likely sequences of words that would follow the input" ... "it predicts one word at a time, based on the previously generated words" ... " and then calculates the probability distribution of the next word in the sequence based on the patterns it has learned from the training data" (OpenAI, 2023).

OpenAI. (2023). ChatGPT (Mar 23 Version) [Large language model]. https://chat.openai.com/chat