Readings

For Students

• Chiang, Ted – ChatGPT is a Blurry JPEG of the Web
  • This article does a good job explaining the limitations of AI’s underlying technology and why we should use it with care. Here, Chiang compares machine learning to file compression formats like .mp3 or .jpeg. This form of file recreation is often imperfect, creating artifacts like odd sounds or pixelated imagery, and machine learning, similarly, creates approximations of human language that can have odd, distortionary artifacts. 
• Lingineni, Vishal – Beyond the Search Box: Large Language Models (LLM) vs. Search Engines and Unlocking LLM Potential
  • This is a really quick reading that explains the differences between LLM-based tools (i.e. ai software like ChatGPT) and search engines like Google. I think this source can help students understand that LLMs and search engines are not the same kinds of tools and thus require different user inputs to function. LLMs need inputs written as syntactical statements that they can process. In short, AI needs a fully written command, one that resembles everyday human language. Search engines, conversely, only need keywords, and they’re designed to find sources that match the words that the user inputs.
• Walker, Jill – ChatGPT is Multilingual, and It’s Learning your Values
  • This is a blog post by Jill Walker, a professor of digital culture from the University of Bergen. This is a longer source that might take students a while to read if you assign the entire thing, so I’d suggest they focus on the first two sections titled “What is ChatGPT trained on?” and “How deep learning models make sense of the world.” I found that these two sections did a fantastic job illustrating how ChatGPT produces knowledge by mimicking texts that it scrapes from large sets of samples. This source does a great job demonstrating that so-called “AI” doesn’t actually have intelligence–at least not the same kinds of intelligence that we humans have.

For Teachers

• Knowles, Alan M. – Machine-in-the-Loop Writing: Optimizing the Rhetorical Load
  • Here, Knowles discusses how students can better incorporate AI into their writing processes. Rather than arguing that writing is either entirely synthetic or human, Knowles reminds us that we must always collaborate with non-humans to write. Therefore, instead of outright rejecting the use of AI in composition, Knowles describes ways that students can be better collaborators with AI tools.
• Cummings, Robert E. – Generative AI in First-Year Writing: An Early Analysis of Affordances, Limitations, and a Framework for the Future
  • This article offers a method (here called the “DEER” praxis) that teachers can use to develop effective composition activities that incorporate AI tools. Here, Cummings argues that teachers won’t successfully integrate AI into the classroom if they ask students to use AI to compose entire projects. Instead, AI works best when teachers ask students to use AI to work through specific tasks in the writing process. Thus, teachers should: Define the compositional stages of the project, Evaluate what AI tool best addresses the stage that students need to work through, Encourage students to explore the tools, and prompt students to Reflect on their use of the AI tool.
• Lorenzo Aguilar, Gabriel – Rhetorically Training Students to Generate with AI: Social Justice Applications for AI as Audience
  • This article offers a method for using AI not as a replacement for the student writer, but rather as an assistant that students can use to automate parts of the research process. The goal, therefore, is to compose through AI by allowing AI to take on the parts of research that would otherwise be overly tedious or difficult for students to complete with the time constraints available to them.