Generative AI for students

Although many generative AI tools offer 'freemium' (free of charge) versions, access to the full range of features can be costly. People who cannot afford to pay for these tools are disadvantaged, as this lack of access hinders their ability to compete in the AI-powered economy, stifling creativity and innovation. Existing inequalities are increased.

Without careful interventions, generative AI risks becoming a tool for further entrenchment of existing inequalities, rather than a force for social good.

To make the most of this remarkable new technology, we’ll need to both guard against the risks and spread the benefits to as many people as possible.

Bill Gates, The Age of AI has begun.

• Generative AI sometimes generates false responses and incorrect references (known as 'hallucinations'). This is because "their goal is to generate plausible content, not to verify its truth."
• Generative AI outputs can't be considered a reliable source for use in assessments. The credentials and expertise of the author can not be verified, the currency of the information generated in the output can not be verified and the information in the output may not be accurate.

Content generated by AI needs to be fact checked and verified using other more reliable sources (such as textbooks, peer-reviewed articles and reference works). Evaluating results in this guide offers handy tips to help you check out tools and the results they produce.

• Generative AI tools are trained on large datasets. ChatGPT's large language model, for example, is trained on internet-sourced data.
• The inherent biases that exist in the content the large language model is trained on may be reflected in Generative AI outputs.
• A recent report by UNESCO on Challenging systematic prejudices: an investigation into bias against women and girls in large language models discusses one type of bias.

Further reading:

Leon Furze's blog post on Teaching AI Ethics: Bias and Discrimination

Copyright is different from academic integrity (see previous tab in this box). Copyright is a type of legal protection that exists to protect the way others can use people's creative works such as books, artworks, sound recordings and film. It is important to remember that Australian copyright law is still catching up with advances in AI.

[NOTE Sometimes an activity permitted by copyright might NOT be permitted within your studies, because your unit convenor wants you to practice skills or demonstrate what you have learned. ALWAYS check your assessment instructions or ask your unit convenor about what is and is not permitted, for EACH assessment.]

Here are some of the copyright issues you need to be aware of when using generative AI tools in your studies:

1. Under a special copyright exception known as Fair Dealing for Research and Study, you may be able to upload resources (for example, pages from a book or an article from a journal) into AI platforms to generate summaries of those works.
2. All creative works (such as images, music or text you entered) you use to generate a new work on an AI platform should be fully attributed (referenced).

• It is important to review the data privacy policies for the Generative AI tools you intend to use.
• Do not include people's names or other confidential information in prompts. The generative AI tool may use the data in prompts to help train the large language model.

Further reading:

Teaching AI Ethics: Privacy (Leon Furze)