Basics of GenAI – Learning and Teaching Consulting

Basics of GenAI

What is GenAI?

The Large Language Model (LLM) tools currently called GenAI are a type of machine learning program that creates new content by statistical analysis of a large set of examples. These LLMs extrapolate from the patterns in the training data and use this analysis to generate new text, images, or other content that match the statistically most likely “next step” based on a user-provided prompt. GenAI can thereby mimic human conversation, composition, and unfortunately, often human bias.

How Does GenAI Work? 

LLMs start by being trained on a very large dataset. For example, if a particular model is intended to generate text, it might be trained on millions of books, articles, and webpages. During this training phase, the system learns the patterns, structures, and styles of the content. The model uses statistical analysis to predict and generate the most likely response to a user-provided prompt, based on the patterns of the training data. With such a large dataset, the results can be very impressive, though contradictions and bias in the data can also lead to unexpected or undesirable results, or to the generation of easily disproven factual inaccuracies that have come to be called “hallucinations”. LLMs are not capable of the judgement involved in fact-checking information, and are generally not programmed to even attempt this, so all output must be checked and validated by human judgement.

Some Common Applications of GenAI

Here are a few of the most common applications of GenAI tools, and suggestions for how they might be used in the classroom. These are, of course, only suggestions, and every LSA instructor can set their own policy and standards for inclusion or exclusion of GenAI tools.

Text Generation

GenAI can compose essays, articles, or even poetry, though the quality of creative writing remains significantly lower than, for example, business writing. 

  • Classes that focus on writing pedagogy could use GenAI texts as samples for students to analyze.
  • Classes that only incidentally include writing could allow more use of GenAI to compose, while focusing student learning time on research or application of theory.

Image Creation 

GenAI can create images, such as realistic photos of people who don’t exist or art in a particular style. However, since the model does not truly “understand” content, and visual art relies so heavily on unspoken cultural conventions and a sort of complex object-recognition difficult for programs to duplicate, endeavors such as creating a realistic avatar of a person, to perform specific actions, have been more successful than creation of more classically understood artwork.

  • Classes that involve recognition, analysis, or creation of art could use GenAI-created images as an exercise in articulating students’ own understanding of a given style, by identifying what seems correct or out of place.

Coding

GenAI can generate code and troubleshoot, suggesting solutions to programming problems.

  • Classes that include programming activities might teach students to use GenAI as an extremely advanced rubber duck.

Whether you include or exclude GenAI tools in your class, students will rely on your guidance to understand what these tools can and cannot do, and to develop their own judgement and critical thinking skills. Read on for some recommendations on how to create your own class policies for GenAI use, and how to be prepared for issues or concerns students might have.

Whether you’re cautiously exploring or actively experimenting, we invite you to use these materials to reflect, discuss, and shape your approach to GenAI.

Scroll to Top