Skip to Main Content
Albert S. Cook Library

Thinking Through Gen AI

Explore how Generative AI tools can support academic reading, writing, and research. Includes tool tutorials, citation guidance, accessibility tips, and more.

What is Gen AI?

Generative AI is a subset of artificial intelligence that creates (generates) content based on the data and examples it was trained on. Currently, many common generative AI tools utilize natural language as an input to create text, images, audio, video, code, and other media.

Core Concepts

Key mechanisms of Gen AI
Mechanism What it Does Real-World Analogy
Tokenization Breaks down sentences into small units like words or sub-words (tokens).  Like separating a sentence into individual Lego bricks to rebuild it.
Attention Mechanism Helps the AI focus on the most relevant parts of your input. Like using a highlighter to mark only the key phrases in an article.
Memory Module Remembers previous parts of a conversation to stay coherent. Like picking up a conversation where you left off with a friend.

How Does It Work?

Although responses from tools like ChatGPT might give the impression that the AI can critically analyze your prompt, it’s important to understand that:

  • Generated outputs are based on predictive models that identify patterns in training data.
  • Gen AI does not actually understand the content it produces.
  • It makes predictions about relationships between words, images, and sounds.
  • It mimics language structures learned from data to create coherent responses.
  • At their core, LLMs function like sophisticated versions of autocomplete.

LLMs have been trained on human-generated data, so they behave more like people than traditional computers. They:

  • Make human-like mistakes.
  • Excel at human tasks.
  • Aren't sentient, but they can feel like they are during interaction.

In other words, a Gen AI platform defines success as answering your question in a way that makes you happy and increases likelihood you’ll keep asking more questions over time. They:

  • Are prone to making “hallucinations,” or confident but false statements.
  • May fabricate citations, referencing or even creating sources that don’t exist.
  • Can be biased as a result of incomplete or discriminatory training data.
  • Cannot reason, critique, or verify facts independently.
  • May only have access to information that is outdated based on recency of training data.

Protect Your Privacy

Be mindful of privacy when using Gen AI tools, as most platforms use your inputs to train future versions of their systems. Avoid sharing personal information, confidential data, or anything that could identify you or others. And just like you wouldn’t copy someone else’s work, don’t feed their work to a Gen AI tool either. This includes your professor's materials, classmates' assignments, or copyrighted content. Remember: Once you input information into these systems, you may lose control over how it's used or stored.

Environmental Impact

Generative AI tools like ChatGPT, Claude, and Gemini require significant energy and water to operate their data centers. Training a single large AI model can consume as much electricity as powering San Francisco for three days, and data centers use millions of gallons of water daily for cooling. And because most big platforms don’t share how energy-hungry each query is, it’s hard to measure the cost. 

Bottom line: While individual queries may seem small, collective usage is substantial. Consider whether Gen AI is truly the best tool for your task, especially for simple questions you could answer through library databases or quick web searches.