Union College’s Honor Council has introduced a new policy regarding generative AI tools in a rapidly evolving academic landscape. While this rule change may have gone largely unnoticed, it’s essential to understand its implications.

In dealing with today’s world of AI, with the mass adoption of LLMs (Large Language Models) like ChatGPT and Bard, we can write variations of essays and papers in seconds. The question then becomes, how do they work?

A language model is a type of model that specializes in processing and understanding human language. It’s trained on large amounts of text data: books, articles, and websites. The model learns the patterns, styles, and structures present in the data.

For instance, ChatGPT, a language model, can create text based on the prompts it receives. You provide it with a sentence or question, and it generates a response that, ideally, is – coherent, contextually relevant, and grammatically correct.

Let’s take a deeper dive into ChatGPT. GPT stands for Generative Pre-trained Transformer. It understands sequential data like sentences by capturing relationships between the words or tokens. You may have, at some point, heard that GPT is made up of a lot of “parameters.” Each parameter is a number. GPT-3 has around 175 billion of these parameters. Each parameter looks like this: output = input * parameter. So, when you enter a sentence, ChatGPT combines the outputs from all the parameters and gives you an output.

However, how can you multiply words with numbers? Each word is broken down into tokens and converted into a set of numbers. For example, Union College is converted to [38176, 5535]. AI tools don’t read text word-by-word like humans; all they see are numbers, and they complete sentences and write papers by probabilistically analyzing which number is usually followed by the earlier number.  Thus, this produces , at times, a coherent, contextually relevant, and grammatically correct response.

The Honor Council’s recent policy change seeks to address the use of these tools in an academic context. When assignments call for original work, unauthorized resources, including generative AI tools, are prohibited. If your assignment showcases your creativity and knowledge, the content should be your own, not AI-generated.

Instructors can set their policies on using such tools in their classes. Some may permit their use, while others may prohibit them entirely. Hence, transparency is critical. Always clarify that AI played a role in your work, and be prepared to provide proper attribution and citation. If you need help attributing AI-generated content, ask your instructor for guidance.

As Union embraces the digital age, students should take this opportunity to explore the creative and educational potential of AI while respecting the rules set forth by the Honour Council. Balancing technological innovation and academic integrity will provide a more enriching and equitable educational experience.