Skip to Main Content

Generative Artificial Intelligence: Language Models

An overview of generative AI and some suggestions on appropriate use

What is a Large Language Model (LLM)?

A large language model, or LLM, is an artificial intelligence that has been trained to understand and generate text in a human-like fashion. Think of it like a super-smart version of autocomplete. Based on having "read" an enormous body of text, called a corpus of text, a LLM can produce similar text in response to a question or other prompt. The biggest LLMs have been trained on a corpus consisting of most of the text freely available on the internet, which takes a huge amount of processing power (and money). However, they are still prone to hallucinate, or make up probable-sounding (but false) information, so you should always fact-check any information you get from a LLM.

LLM Options

  • ChatGPT has two versions, a free version and a paid version. The paid version uses a more advanced language model and enables you to use plugins and connect to the internet. 
  • ChatGPT is good at brainstorming and providing explanations. Because it doesn't link to material elsewhere on the web, its responses are usually very comprehensive.
  • Microsoft Copilot uses the GPT-4 language model, the same model used by the paid version of ChatGPT.
  • Microsoft Copilot is connected to the internet. It answers questions more slowly than the other LLMs because it is actively searching the internet. It usually provides source links or suggested links for more information on a topic.
  • Google Bard is connected to the internet.
  • As of July 2023, Bard is more concise and less creative than other LLMs. It tends to provide responses in bullet points.

Use Cases for Education

LLMs are good at many tasks that make teaching and learning easier, such as explaining concepts, providing examples, and designing assignments. However, if you are an instructor, make sure you check any LLM-generated content for errors before giving it to your students. If you are a student, make sure you check with your instructor before using a LLM to help with an assignment. Here are some possible use cases for education:

  • Brainstorming topics for a paper
  • Outlining a paper
  • Checking your grammar
  • Coding a web app

For Instructors

The following articles provide some suggestions on how to use generative AI effectively in the classroom. Whether and how you choose to use AI in the classroom is up to you.

For Students

The following articles provide some suggestions on how to use generative AI appropriately in the classroom if your instructor says that you may.

Prompt Engineering

As with a human, when you give a LLM a task, providing specific instructions yields better results. Here are some ways to be specific:

  • Provide a role: "You are a scientist who studies dark matter."
  • Provide an audience: " . . . for an audience of second graders."
  • Specify a tone for the response: "In a scholarly tone, . . ."
  • Specify a format for the response: "In five bullet points, . . ."

AI Tools for Specific Purposes