Generative AI tools, such as OpenAI's ChatGPT, Microsoft's Copilot, and Midjourney Inc.'s Midjourney, utilize machine learning and massive knowledge bases to allow individuals to answer questions, write essays, develop software, and create multimedia. This lab will lead you through a series of exercises exploring Generative AI tools and how you can use them effectively and ethically.
Creighton has created a video on the use of Generative AI tools in your classes: Generative AI and Academic Integrity. This video is used to lead discussions on these issues in RSP classes. If you have not already watched this video, do so now. The University has also created a more detailed PowerPoint presentation that you should review: AI Student Guide.
Reading about generative AI tools can be informative, but nothing beats exploring the tools and reflecting on their actions. The following exercises focus on using two of the most popular tools, ChatGPT and Copilot, to answer questions. If you are asked to report a response and that response is more than a paragraph, you may report the first paragraph and a brief summary of the rest.
EXERCISE 1: Explore the capabilities of ChatGPT. Think of a reasonably complex question you would like the answer to and enter it is a prompt.
- What was the prompt that you entered?
- What response did you receive from ChatGPT?
- How accurate/useful was the response?
- Ask ChatGPT for a different answer to your prompt (e.g., "Can you give me another answer?"). What response did you receive?
- In your opinion, is the second response better or worse than the first response? Explain your reasoning.
Each generative AI tool is different. They were trained on different knowledge bases and may weight information differently. It should not surprise you that two different tools might produce very different results.
EXERCISE 2: Enter your prompt from EXERCISE 1 into Copilot.
- What response did you receive.
- Is the Copilot response similar to the ChatGPT response? If they are substantially different, which do you feel is the better response. Why?
- Ask Copilot for a different answer to your prompt. What response did you receive?
- In your opinion, is the second response better or worse than the first response? Is it better or worse than the ChatGPT responses? Why?
While Generative AI tools such as ChatGPT do very well on answering most questions, it is important to understand that their goal is to produce answers that are plausible and "average" with respect to their knowledge base. The answers may not be correct, however. When an incorrect response is given, this is known as a hallucination, as if the tool imagined the answer and reported it as fact. Responses can also exhibit biases from the knowledge base or produce mediocre answers that only partially answer the question.
EXERCISE 3: Use ChatGPT to generate hallucinations or biased/mediocre answers.
- Enter the following prompt: Divide 98989898 by 13, then take the square root. What was ChatGPT's response?
- Do the math yourself using a calculator or spreadsheet and compare the answer that you get. Did ChatGPT give you the correct answers?
- Try to devise your own prompt that will generate a hallucination or biased/mediocre answer. What prompt did you come up with?
- What response did you receive from ChatGPT. How did it do?
- What does this tell you about the reliability of Generative AI responses?
EXERCISE 4: Brandolini's Law, also known as the bullsh*t asymmetry principle, states that "the amount of energy needed to refute nonsense is an order of magnitude greater that that needed to produce it."
- In general, do you agree that mistakes, lies and fake news are far easier to produce than to identify or clean up? Why do you feel that way?
- Read the conversation titled Diabolus Ex Machina. What do you think about this conversation? Does it change your perception of Generative AI and how trustworthy its results can be?
AI tools can streamline the process of writing, coding or creating visual art. However, there are ethical considerations. Did the authors/artists who created the training materials give permission for their use? Does the use of Generative AI harm our environment or contribute to inequity in society? At what point does Generative AI go beyond helping you to do the work but instead does it for you?
EXERCISE 5: Reflect on questions such as these and answer the following.
- Skim the Creighton Library AI ethics resources. Briefly summarize your thoughts related to ethical use of AI.
- During the 2024 Olympic Games, Google ran an ad concerning generative AI that has been widely debated. Watch the Dear Sydney ad. Do you feel that this use of generative AI is ethical. If you were to receive a fan letter, would you rather receive a polished letter created with an AI tool or a less-polished letter directly from the person? Similarly, if you received an apology email from someone who has wronged you, would you view it differently if the email was written using a generative AI tool as opposed to being the words of the apologizer?
- Read the article Explained: Generative AI's environmental impact from MIT News. Were you aware of the level of resources (e.g., electricity, water) needed by AI applications before reading this article? Do you feel differently about AI after reading the article? Will this new information affect the way that you utilize AI tools?
EXERCISE 6: A common use of generative AI by students is to create study tools, such as quizzes or study guides.
- Enter the prompt "Write me 5 true/false questions (without answers) on the history of the Internet" What are the questions it responds with? Do you feel these are reasonable questions?
- Note your own answers to the questions, then enter the prompt "Now show me the answers." How many did you answer correctly?
- Do you feel that the use of generative AI to create quizzes or study sheets is ethical? Why or why not?
- What are the potential dangers of using generative AI to create quizzes or study guides?
EXERCISE 7: Reflect on your own personal use of generative AI.
- How do you currently use generative AI related to your area of study and/or future profession? Describe the resources you currently use, and for what purposes. Be sure to touch on each of the areas below, where relevant.
- writing assistance
- creative assistance
- study assistance
- research and information gathering
- data analysis and visualization
- other ways
- Do you believe your usage will change as you enter your professional life?
- What do you believe are the most important ethical dimensions of generative AI use in your area of study and/or future profession. Why do you feel that way?
Submit a document containing your answers to all the lab questions via BlueLine.