A Practical Guide to Generative AI Tools: Understanding, Using, and Acknowledging AI

The landscape of AI generation tools is vast and continuously expanding, encompassing a wide array of applications. These tools can be broadly categorized into those solely focused on AI generation, such as text generators like Google Gemini and ChatGPT, and broader applications that integrate AI generation features. Many existing software and learning platforms now incorporate generative AI capabilities, offering specialized functionalities within their existing ecosystems. The most prevalent forms of generative AI tools are those for text and image generation. Text generation tools, often powered by Large Language Models (LLMs), can produce conversational responses, documents, code, and more. Image generation tools allow users to create visual content from textual descriptions, often with specific stylistic requirements. It's important to note that while some tools are free, many image generation tools operate on a credit system or require a subscription. The University of York, for instance, recommends Google Gemini and provides access to Microsoft Copilot, emphasizing that using these tools with a University account ensures data privacy and prevents prompts from being used for model training. For image generation, while specific tools aren't endorsed, Google Gemini can be used, though it's cautioned that AI may struggle with generating highly specific real-world locations or objects accurately, making real photographs preferable in such cases.

The effectiveness of generative AI tools hinges significantly on the quality of the prompts provided. Crafting well-structured prompts involves more than just asking a question; it requires providing clear instructions and context to guide the AI towards the desired output. A common approach is to start with an initial prompt, evaluate the response, and then refine the prompt iteratively. Strategies for improvement include giving explicit instructions rather than simple questions, being specific about the desired output format, length, and style, and defining an audience or persona for the AI to adopt. For instance, instead of asking 'War and Peace,' one might prompt: 'Summarize the book War and Peace in three sentences for a ten-year-old.' Google proposes the 'PARTS' acronym as a framework for constructing effective prompts: Persona (setting the AI's role), Aim (defining the objective with a verb), Recipients (specifying the audience), Theme (detailing style, tone, and other relevant information), and Structure (outlining the desired format). Understanding that different AI models and tools have unique training data and architectures means that prompts effective for one tool may need adjustment for another. Experimentation and learning the nuances of each tool are key to maximizing their utility.

The rapid proliferation of generative AI tools brings forth a complex array of ethical questions and considerations. These range from the potential for bias and inaccuracies in AI-generated content, stemming from the datasets used for training, to concerns about intellectual property and copyright. Universities are actively developing guidance to help students, researchers, and staff navigate these challenges responsibly. For example, the University of York provides specific guidance for students and postgraduate researchers on the use of generative AI in assessed work, as well as for researchers in their scholarly activities. IT Services also offers general advice on recommended tools, data privacy, and security when using generative AI. It is imperative for users to critically evaluate all AI-generated outputs, cross-referencing information with trusted sources to verify accuracy. When using AI tools, especially those not officially sanctioned or integrated with institutional accounts, users should assume that their data may be used for model training and may not be secure. Adhering to university policies and ethical best practices ensures that generative AI is used as a tool to enhance learning and research, rather than a means to circumvent academic integrity or spread misinformation.