Current Students The Learning Centre During the study period Academic Integrity Generative Artificial Intelligence

Generative Artificial Intelligence

What is Generative Artificial Intelligence and can I use it at JCU?

Generative Artificial Intelligence creates unique data, images, texts, sounds, and video based on existing information using complex algorithms.

There are two main models, Generative Adversarial Networks (GANs) and Generative Pre-Trained (GPT) models.

GAN uses Machine Learning techniques to generate data and test it against existing data. Models like GPT use massive datasets (ChatGPT-3 is thought to contain over 45 terabytes of text) to generate natural sounding language that is based on the text it has in its data banks. However, this can never truly be original information as it is derivative of the material that is pre-existing in a databank.

At JCU the use of Artificial Intelligence (AI) in learning and assessment must be ethical, transparent, and purposeful. You may engage in the uses of AI only where they uphold the principles of Academic Integrity. You must avoid the use of AI in anyway which may contravene the principles of Academic Integrity. Generally this means you cannot use AI to complete your assessment for you. Read the information below to ensure you are clear on the ethical uses in your studies.

NLP is an algorithm that uses a Language Model (LM) that can analyse patterns in language, and predict what words should go into a sentence based on probability. For example, the LM will know that a verb should follow a subject in a sentence, and will write a sentence to follow this rule. The Language Model is the basis of common tools used every day such as Google Translate, Alexa, and Siri.

Generative Artificial Intelligence based on the Natural Language Models makes a series of predictions based on the information in its databank. For example, when predicting the weather an algorithm predicts tomorrow’s weather based on the current situation and the past weather behaviour data that is accessed. This is the same with an AI like ChatGPT, it will write a text based on past data relating to the question it is given by the user – which is known as “the prompt”.

Artificial Intelligence can be beneficial to assist with time consuming tasks like data processing or background information gathering. We need to be aware that this technology exists and that it can assist with improving our efficiency as scholars.

Generative Artificial Intelligence output data is synthetic. It is not new information or research. AI programs like ChatGPT are non-human and use existing datasets to write predictive text, and because of this their output is not always reliable, nor is it always correct. An AI could also produce incorrect information if its dataset is incomplete or when it guesses that information could fit the context. An early example of this was in February 2023 when Google’s parent company share price lost $100 billion in value immediately after it became public knowledge that their new AI chatbot had provided inaccurate information to user enquiries. Some AI output could also be harmful to human life if applied incorrectly, or may be biased. A human must apply ethical reasoning to AI output to ensure it is suitable for the purpose it will be used for.

We also do not always know who wrote the algorithms that control the output, or what their initial requirements were. We also do not know how the database was created, or if it is frequently updated. ChatGPT-3 for example only uses data to 2021.There are also serious security concerns with using AI, any information you put into an AI could be used by others. Some actors may use AI to create deceitful information such as ‘fake news’.

Another thing to be aware of is that because ChatGPT is a predictive language model, it can write a reference that looks correct  and follows the correct style (such as in APA), on any topic you like, but which does not actually exist in any research database.

You should check with your Subject Coordinator or lecturer about the appropriate use of AI in your subjects. In general, AI use must not contravene the Academic Integrity Policy. You should also ensure you have completed the Academic Integrity Modules in LearnJCU if you have not already.

The first thing you must do is check the assessment item information in your Subject Outline. This should make it clear if you are permitted to use an AI in your assignment. If you are still unsure, you must check with your Subject Coordinator or lecturer.

If you do use ChatGPT or another AI in your assignment you must acknowledge it. For example, you can add a disclaimer on the front page of your essay which includes an acknowledgement: “ChatGPT was used in this assessment as a tool for background research” or “ChatGPT was used to check the spelling and grammar in this assessment”.

Some lecturers will ask you to include the prompt that you used in ChatGPT, any text you copied into ChatGPT, and the original text output you received from ChatGPT in the Appendix of your assignment.

See the  information below on how to create your own Declaration.

A Declaration of AI generated material identifies the extent to which a GenAI Tool was used. Generally it includes a written acknowledgement that one or more GenAI Tools were used, and descriptions of how the information was generated (including the prompts used).

A suggested format:

I acknowledge the use of [insert AI system(s) and link] to [specify use of generative artificial intelligence]. The prompts used include [list of prompts]. The output from these prompts was used to [explain use].

Example 1

I acknowledge the use of [1] ChatGPT (https://chat.openai.com/) to [2] generate materials for background research and self-study in the drafting of this assessment. I entered the following prompts on 4 January 2023:

  • [3] Write a 50 word summary about the Mabo native title case. Write it in an academic style. Add references and quotations from Eddie Koiki Mabo.

[4] The output from the generative artificial intelligence was adapted and modified for the final response.

Legend: [1] AI system(s) and link; [2] specific use of generative AI; [3] prompt; [4] explanation of use.

Example 2

I acknowledge the use of [1] ChatGPT (https://chat.openai.com/) to [2] generate materials that were [4] included within my final assessment in modified form. I entered the following prompts on 4 January 2023:

  • [3] Write a 50 word summary about the Mabo native title case. Write it in an academic style. Add references and quotations from Eddie Koiki Mabo.

Example 3

I acknowledge the use of [1] ChatGPT (https://chat.openai.com/) [2] to refine the academic language and accuracy of my own work. On 4 January 2023 I submitted my entire essay (link to google document here) with the instruction to [3] “Improve the academic tone and accuracy of language, including grammatical structures, punctuation and vocabulary”. [4] The output (here) was then modified further to better represent my own tone and style of writing.

Example 4

If AI was permitted to be used in your assessment, but you have chosen not to use it, the following disclosure is recommended.

No content generated by AI technologies has been used in this assessment.

No, you cannot use any AI to write your whole essay for you. Submitting assessment that is not your original work is a form of academic misconduct and this is strictly prohibited (see procedure). If you use text directly from ChatGPT you must reference it as a block quote and include the reference in your discipline’s style.

At JCU students must complete an authorship declaration when submitting an assessment. This declaration states that you are the original author of your assessment and that you did not copy or reproduce from any other sources (including an AI) without attribution or referencing. The declaration says:

This assessment is my original work and no part was copied/reproduced from any other person’s work or any other source (including artificial intelligence), without appropriate attribution. This original work adheres to assessment rules including group assessment.  I have not submitted any part of this assessment previously (or concurrently) to, nor been awarded a grade in, any other subject or course. Academic Misconduct Procedure

N.B. You will need to check your preferred style guide for the most up to date information. The following relates to the APA format as an example:

It is not possible for other readers to access the outcomes of your "chat" with ChatGPT, this means the output is considered to be unrecoverable. In APA Style papers, unrecoverable data or quotations are often recognized as personal communications, but in the case of ChatGPT-generated text, there is no human author. We treat the output as if it were from an algorithm, and credit the author of the algorithm, OpenAI. For in text citations follow the Parenthetical citation: (OpenAI, 2023) or Narrative citation: OpenAI (2023) styles. In the reference list credit the author of the algorithm, the year, then the title, the access date, and a link.

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

If it is possible to link directly to the text of the chat or output, you would use that as the URL.

However, it is important to note that this style likely to change and students should check their preferred style guide for updates.

You should reference according to the style above.