Generative AI, Large Language Models, and Higher Education

Resources for Artificial Intelligence

"From Alan Turing to GPT-3: The Evolution of Computer Speech" | Otherwords : This resource provides a quick overview of the technology behind ChatGPT, explained very clearly by OpenAI's Technical Director.

Ithaka S+R's Generative AI Product Tracker lists more than 100 tools, and provides links and vital details about the function, purpose, and cost for each tool.

AIPRM's AI Glossary, provided by Skye Olley and the STEM Club for girls from Fullers Library

How to Provide Citations When Using Artificial Intelligence Tools

The Chicago Style Guide FAQ provides guidance that covers how you can mention AI-generated text or images in the body of your own work and, when needed, as more formal endnotes or footnotes.

The Modern Language Association's guidance suggests that you note all uses of the tool, too, including if you use it as an editor or translator. Their examples provide guidance on text as well as images and other kinds of sources.

This short piece from the American Psychological Association (APA) offers clear guidance about how to distinguish between the various elements of the citation (such as author, title, and URL) and provides examples of parenthetical, in-text, and narrative citations.

Introduction to Generative AI and Its Implications in Higher Education

by Christl Caspar, former Coordinator, Point Park Center for Inclusive Excellence

Generative artificial intelligence has emerged recently as a powerful technology with profound implications for higher education. As we grapple with this evolution's difficult questions, the Center for Inclusive Excellence is dedicated to navigating this new landscape with Point Park by providing resources and support.

What is clear right now is that generative AI is here and many of these tools, like ChatGPT, are open access. While generative AI will undoubtedly change higher education, it is also important to note that there are many other industries that will also be impacted. We encourage everyone -- faculty, staff, and students -- to familiarize themselves with the merits, limitations, and vulnerabilities of this software. While this page is by no means an exhaustive guide, we hope it will help introduce you to generative AI, how it works, and the evolving conversations around it.

What Is Generative AI?

Generative AI is a subset of artificial intelligence that can create new content, data, or outputs. ChatGPT, DALL-E, and Bard are all prominent examples of generative AI. Generative AI learns patterns and structures from large datasets, enabling them to generate new content such as text, images, music, and videos. Content from generative AI can sometimes be indistinguishable from human-generated content because it relies on large amounts of human-generated data. However, this is not exclusively the case. Generative AI can have a generic voice and is still subject to hallucinations (see below), biases, misinformation, and cannot conduct high-level critical thinking.

What Are Large Language Models?

Large Language Models (LLMs) are a type of generative AI that use massively large datasets and complex algorithms to understand and generate human-like language. LLMs are trained on text from books, articles, websites, and other written sources. They learn the patterns, grammar, and meaning behind different words and sentences. In a sense, large language models (like GPT-3) are like highly sophisticated conversation partners. They can carry on discussions, provide information, assist in tasks like language translation or summarization, or generate stories, poems, and other types of written content. LLMs have a wide range of applications from helping with customer service chats, aiding in research, or providing writing suggestions.

What are AI Hallucinations?

AI hallucinations, also known as "adversarial examples", refer to situations where AI systems (particularly deep learning models, such as Bard or GPT-3), misinterpret or generate incorrect outputs in response to specific inputs that would seem ordinary to humans. AI hallucinations occur because LLMs have no real understanding of language or the reality it describes.

AI hallucinations might appear plausible but can also appear nonsensical. Hallucinated information might perpetuate harmful stereotypes or bias, impact decision-making, or simply mislead a user if the user does not verify the information with a reliable source. With this in mind, it is important to emphasize critical thinking, reliable research, and media literacy when using LLMs like GPT-3.

Hallucination Example: Google ChatGPT Rival Bard Flubs Fact about NASA's Webb Space Telescope

Other Types of AI

While our cultural conversation is currently focused on generative AI, generative AI is not the only type of artificial intelligence. In fact, we are already using AI in our everyday lives:

Machine learning - A branch of AI where computers learn patterns from data, enabling them to make decisions or predictions without being explicitly programmed. Examples of tools that can use machine learning are virtual assistants (Siri, Google Assistant, Alexa) and spam email filtering.
Deep Learning - A subset of machine learning that imitates the human brain. Deep learning employs artificial neural networks with multiple layers to automatically learn and extract complex patterns for data. Examples include image recognition and natural language processing.
Natural language processing - A field of AI that focuses on enabling computers to understand, interpret, and generate human language. Google language translation, email filters, and search engine results all use natural language processing.
Artificial general intelligence - Artificial general intelligence (AGI) is a hypothetical type of intelligence that possesses human-like intelligence and is capable of understanding, learning, and performing any intellectual task a human can do. Examples include Samantha from the movie Her, HAL 9000 from 2001: A Space Odyssey, WALL-E from WALL-E, and Do Androids Dream of Electric Sheep? by Philip K. Dick. As mentioned prior, AGI is purely hypothetical and does not exist.
Artificial neural network - A computational model inspired by the human brain's interconnected neurons. Artificial neural networks are used in machine learning to process and learn patterns from data.
Expert system - An expert system AI is a computer program designed to replicate the decision-making and problem-solving abilities of a human expert in a specific domain. Some expert systems are DENRAL, a molecular structure prediction tool for chemical analysis, and CaDet, a system that can detect cancer in its earliest stages.
Robotics - Robotics involves the creation of robots to perform independent tasks. Robots can perceive, process information, make decisions, and perform tasks autonomously or semi-autonomously, enhancing their ability to interact with and navigate their environment.
Cognitive computing - AI systems that aim to mimic human thought processes in complex situations where the answers might be uncertain. An example of cognitive computing includes IBM's Watson.
Random forest - According to IBM, "Random forest is a commonly-used machine learning algorithm [...] which combines the output of multiple decision trees to reach a single result." Random forest can be used for classification, like determining whether an email is spam or not.