Ethics of AI Assignment 3
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Ethics of AI
Assignment 3
October 13, 2025
• Total points: 10 out of 100.
• Include your full name and student number at the beginning of your submission.
• Deadline: 6 Nov 2025, 4:00 pm (as indicated on the Moodle assignment page)
• Submit a PDF file via Moodle (within 3 pages), using the Assignment 3 submission link.
• Please ensure compliance with the university’s Generative AI (GAI) policy. You may use GAI tools only for brainstorming or language polishing. But do not use them to directly answer the questions. Also, if you use any GAI tools, you must clearly state at the beginning of your submission which tool(s) you used and for what purpose.
1. (15%) You are free to choose either A or B.
Choice A (I hope you choose this one though): Click this link: https://playground.tensorflow.org/, and train the fourth dataset in the bottom right (the spiral one). Make sure that both the training loss and the test loss are below 0.05. Please attach a screenshot of the entire webpage after you complete the training. Additionally, answer the following questions:
• How many hidden layers are there?
• How many nodes are in your second hidden layer?
• What are the weights and bias of the second node in the first hidden layer (you should have this node)?
Choice B: Take a look at this image and answer the following questions:
• How many hidden layers are there?
• How many nodes are in the second hidden layer?
• Do you think this neural network is a black-box model? Why or why not?
2. (15%) Click this link: https://banana-projects-transformer-autocomplete.hf.space/doc/gpt2-large. Generate three lines of text using it, and copy the generated text in your response.
• Explain how it generated this text.
• Do you think the text looks reasonable? Is it something that could appear in everyday life?
3. (25%) Read this article: https://www.bbc.com/news/technology-54175359. Based on the content of the article (this accident is mentioned on page 112 of our required reading), answer the following:
• 5% What happened?
• 5% Was the driver held responsible? What reasons does the article provide?
• 10% Should Uber be held accountable for this? Why or why not? (cf. conditions for responsibility)
• 5% What is your main opinion on this matter? Mention any points you think are worth discussing.
4. (15%) Provide an example of a potential negative consequence involving opaque AI appli- cations (e.g., medical AI decisions).
5. (15%) Regarding the trade-off between accuracy and explainability: Do you think re- searchers have an obligation to develop explainable models, even if they are less accurate? Alternatively, if they continue to develop highly accurate but unexplainable models, should they be held responsible for possible serious consequences? Why or why not? (cf. condi- tions for responsibility)
6. (15%) Considering the different audiences involved, what do you think a good explanation should look like? Do you think it is possible for such a good explanation to exist?
2025-11-04