AI in the Academic Setting

Generative AI tool usage and integration into academic and daily life is rapidly evolving.  This guide is frequently updated!  If you have any comments or suggestions, please contact us at tlrs@oxy.edu

Keep in mind: "...generative A.I., relies on a complex algorithm that analyzes the way humans put words together on the internet. It does not decide what is true and what is not."  When AI Chatbots Hallucinate—The New York Times. (2023, May 8). Archive.Ph. https://archive.ph/07rpm

What is Generative AI?

• Generative AI is a type of artificial intelligence that has been trained on massive amounts of data from which it "learns" patterns and structures.  It then can generate new data (text, images, etc.) when prompted by a user input. (Generative AI overview from Wikipedia).  It is a subset of deep learning combined with natural language processing.
• ChatGPT (generative pre-trained transformer) and others like it are powered by Large Language Models(LLM's): A very gentle introduction to large language models without the hype 

Image source: Large language models An introduction to fine tuning and specialization in LLM's

Watch this short video to learn how large language models (LLM's) work.  LLM's make predictions based on their training data.  It's statistical- not magic!

Chatbot is a generic term for an artificial intelligence (AI) tool that uses natural language processing techniques to respond to user generated prompts.  Chatbots have been created by Google, Microsoft, Amazon, and Stanford University to name a few and have been around in some form since the mid 1960's.  Heard of ELIZA? 

History of Chatbots Infographic (up until year 2016)

ChatGPT is a chatbot created by OpenAI and was released to the general public as a "research preview" in late 2022.  It is a form of generative AI.  The usage and interest in ChatGPT has been an active subject in the scholarly and news media since then. 

GPT-4 was released in March 2023 and now powers ChatGPT Plus (paid version of ChatGPT) and is also integrated into Microsoft 365, the successor to Microsoft Office.  Microsoft calls its generative AI Copilot.  Amazon now has one for the workplace called Amazon Q.

OpenAI released in May 2024 GPT-4o which can accept audio and visual inputs as well as text and has access to the internet.  In September 2024, OpenAI released GPT-o1 that takes its time to "reason" through complex math and science problems.  It can be access via a $200 per month subscription to ChatGPT Pro.

There are currently multiple companies producing GPT-4 class models.  In addition to their training data, they have various levels of connection to the internet to include up to date information in their output.  They can still "hallucinate" incorrect answers. Gen3 (GPT-5 models) are on the horizon.   Also, note that both ChatGPT and Gemini can now (December 2024) see live video and interact with voice simultaneously, typically called "Live Mode."

• Claude 3.5 Sonnet: no access to the internet
• GPT-4 Microsoft CoPilot (Bing), ChatGPT Plus, or GPT-4o: has access to the internet
• Google Gemini Advanced (formerly known as Bard)/Gemini 2.0: has access to the internet
• Grok 2:  from X.AI, has access to the internet
• DeepSeek R1: open source, has access to the internet
• Llama 3.2: from Meta

The following is a current (January 27, 2025) listing and comparison of these models (from an opinion piece* by Ethan Mollick):

*Ethan Mollick, One Useful Thing Substack, January 26, 2025