Week 04: Ethics, Bias and Legalities in AI

Lecture

This class will be focused on ethics, bias and the legal side of generative AI. We will discuss the ethical side of how these platforms are trained on vast amounts of data that was taken without the consent of the original creators. We will also talk about the issues with the outputs generated from these platform that tend to have a racial and socioeconomic bias. We will also be talking about the legal issues behind how these modals are trained as well as how to properly use the content that is generated by these platforms.

Guest Lecture: James Creedon

Sam Harris Podcast

Homework

Choose one (or more) below:

If you are interested in bias and stereotypes in AI training data:

Using a generative AI image model of your choosing, output at least 100 generations using a descriptive adjective (beautiful, ugly, scary, joyful, etc.) or job title (doctor, lawyer, teacher, police officer, nurse, etc). Analyze the outputs and see what stereotypes or trends you see. Create a blog post on your findings. Look at Professor Woo’s gender and race bias studies for reference.

If you are interested in the use of IP and copyrighted materials used to train image models:

Use Have I Been Trained to research 5 of your favorite artists to see how much, if any, of their art has been used to train an AI image model. Create a blog post on your findings.

If you are interested in ethical solutions and best practices around generative AI:

Using the United Nations AI Advisory Board report on Governing AI for Humanity as a guide, design a poster (11″ x 17″) around the suggested rules and regulations for adopting AI for a more positive impact on humanity. Create a blog post showing your process and final poster design. Feel free to use ChatGPT to help you analyze the report.

Leave a Reply