Skip to Main Content

Generative AI (ChatGPT, Gemini, & More): Home

Guide & Resources on generative AI in education, including examples, explanations, and more.

Introduction

What is Generative Artificial Intelligence?


Introduction

Welcome to our comprehensive LibGuide on Artificial Intelligence (AI), a transformative field that has captivated the world with its immense potential and thought-provoking implications. In this guide, we will embark on an enlightening journey through the history of AI, explore the ethical considerations surrounding its development and implementation, and delve into its fascinating applications in academia.

Join us as we trace the roots of AI, from its inception as a visionary concept to its current state as a driving force behind cutting-edge technologies. Uncover the milestones, breakthroughs, and key figures that have shaped the evolution of AI, forever altering the way we interact with machines and the world around us.

While AI offers boundless possibilities for progress and innovation, it simultaneously raises ethical dilemmas that require careful contemplation. We will explore the ethical considerations surrounding AI, such as privacy concerns, bias and fairness issues, and the impact of AI on society. Delving into these complexities will help us navigate the responsible and equitable use of AI technology.

In academia, AI has emerged as a potent tool, revolutionizing research, teaching, and learning. We will showcase how AI is transforming various academic disciplines, enabling more efficient data analysis, personalized learning experiences, and groundbreaking discoveries. From natural language processing to machine learning algorithms, discover how AI is enhancing academic pursuits and shaping the future of education.

Whether you're a student, researcher, educator, or curious individual, this LibGuide is your gateway to understanding the multifaceted realm of AI. Let's embark on this illuminating expedition into the history, ethics, and academic applications of Artificial Intelligence. Embrace the opportunities, grapple with the challenges, and expand your knowledge in this dynamic and ever-evolving domain. Happy exploring!

This introduction was created using an AI text generator named ChatGPT - specifically ChatGPT 3.0.

If you were not told that text was generated, more than likely you thought the author of this guide wrote it from scratch. The output generated by these predictive generators is complex, comprehensive, and surprising in how natural it can appear. The above was generated by a Large Language Model (LLM) text generating chat bot called ChatGPT. Below is the prompt I used with the chat bot. Read below to learn how these bots (and prompts) work.


 

Generative AI Chatbots (such as ChatGPT) are not:

  • connected to the Internet in real time - some can access Internet search, but not spontaneously (paid versions of some chatbots have access to the Internet)

  • generating discrete work - your prompts can generate different results with identical prompts

  • unbiased - the chatbots are as biased as the training materials (data set)

  • encyclopedias - chatbots are not authoritative sources and, due to the derivative generated nature of its content, are incomplete and/or inaccurate

What are Large Language Models?

Terminology & Glossary


The term "artificial intelligence" (AI) is an overly broad term. The "AI" we can interact with today are mostly "Large Language Models" (LLMs), which are a derivative of "machine learning" (ML) algorithms & processes. While these are under the umbrella of artificial intelligence, instances of these chat bots are no more intelligent than your phone assistant. These chat bots (and similar applications of AI) are generative and predictive, but have no true intelligence.

Glossary:

Artificial Intelligence (AI) - "machines that respond to stimulation consistent with traditional responses from humans, given the human capacity for contemplation, judgment, and intention" - Shubhendu, Vijay, International Journal of Scientific Engineering and Research. (Source)

Artificial General Intelligence (AGI) - "Algorithms that perform a wide variety of tasks and switch simultaneously from one activity to another in the manner that humans do." (Source)

example: we do not currently have any publicly known comprehensive AGI

Artificial Narrow Intelligence (ANI) - most of currently publicly accessible AI that perform a single task, but cannot generalize & synthesize the outcome outside of the AI's given function.

example: Looking for a particular face in iOS's Photos app or asking Google Assistant the weather.

Deep Learning - "A subset of machine learning that relies on neural networks with many layers of neurons... deep learning employs statistics to spot underlying trends or data patterns and applies that knowledge to other layers of analysis." Deep Learning requires large amounts of power and data. (Source)

Machine Learning (ML) - a way for artificial intelligence to classify data, pictures, text, objects, or other media without detailed instruction, and is able to continually learn on new data sets introduced. New data sets are comprehended in the context of older data sets.

Large Language Model (LLM) - generates text based on large data sets. LLMs use deep neural networks (Deep Learning & Machine Learning) to process and create text based on specific criteria.

Generative Pretrained Transformer (GPT) - a state-of-the-art language model based on the Transformer architecture. It is "pretrained" on a vast corpus of text data, learning patterns and language representations in an unsupervised manner. GPT can "generate" human-like text by predicting the next word in a sequence, making it capable of completing sentences, answering questions, and performing language-based tasks. (Prompt from ChatGPT. [In 100 words or less, explain what a generative pretrained transformer (GPT) is.])

Chatbots - interactive artificial narrow intelligence tools based on large language models and a generative pretrained transformer. Chatbots are interacted with via a text prompt or dialog box. Chatbots create responses using a vast corpus of "digested" data from large amounts of data sets.


This is not an exhaustive list of terminology related to AI. A very thorough source of terminology can be accessed through AIPRM's Ultimate Generative AI Glossary. While AIPRM is a for-profit business-oriented company, the glossary is informational, easily understandable, and thorough.

How Do They Work?


Large language models, such as GPT-3, work by leveraging deep learning techniques and vast amounts of data. They consist of neural networks with multiple layers that process and understand text input. During training, these models learn patterns, relationships, and representations of language from the data. Once trained, they can generate human-like text, answer questions, perform language-related tasks, and even exhibit a degree of understanding and context. The vast size of these models enables them to capture intricate linguistic nuances, making them powerful tools for natural language processing tasks.

- Written by ChatGPT

Chat bots such as ChatGPT work on a prompt system. The user, such as yourself, writes a prompt into a chat box. The chat bot - we'll use ChatGPT as the main example - will then process your prompt. What it attempts to do is a "reasonable" procedural continuation of words to mimic what is in its data set(s). ChatGPT uses probabilistic analysis to rank each following or upcoming word in a response.

The chat bot is asking itself "what's the word with the highest probability rank" after each word choice. This probability stems from a corpus of data - e.g. data scraped from the Internet and websites such as Reddit or Wikipedia - processed by machine learning.

However, if AI only chose the top ranked word in association, the responses would be one-note (you could say dry or boring) and deterministic. There is then some wiggle-room in how the AI chooses the next word - sometimes the fifth "best" word or the eighth "best" word - to create a form of difference.

For an extremely detailed outline of how these Large Language Models work when paired with a GPT, look to Stephen Wolfram's explanation of how GPT-2 - OpenAI's transformer from several years (and software generations) ago - and LLMs work.

The Uses, Limitations, & Pitfalls of AI

Uses

 
  • It can be useful as a starting point for research. The conversational aspect of chat bot interaction can help you "organically" look for subjects or terms.
  • Can help you find alternate search words, similar terminology, and suggest alternate lines of research. A super-powered thesaurus.
  • Simple text editing. The chat bots are good at formatting and checking English. Great for grammar correction.
  • LLMs are very good at basic programming and coding, due to basic coding and its highly structured nature.
  • Can create outlines and brainstorming ideas for your research inquiries. Good for starting out.
  • Responds extremely quickly. Response generation usually occurs within 30 seconds.

Limitations

  • Can only assist in a prompt-response fashion. LLMs can respond with a previous prompt for context and some can incorporate longevity, but differs by product.
  • Has trouble synthesizing more complex ideas, subjects, or projects. Meant for more simple or brief interactions.
  • Cannot do math. Large language models rely on structural consistency and are geared towards language, such as English.
  • LLMs only have knowledge up to a certain date. Some LLMs can access the Internet, but is largely limited to paid tiers of access.
  • Some are free, some are paid access.
  • Different models will produce different results.

Pitfalls

  • The corpus of work processed by LLMs and GPT bots causes plagiarism issues. Nothing "generated" by these chat bots is "original".
  • Chat bots can "hallucinate", wherein the bot represents false information as truth. The chat bot does not know it is wrong (it does not have that capacity), but will proffer fake information in responses.
  • Data sets are biased. Search algorithms are also biased.
  • Anything you prompt a chat bot with (and anything generated) is added to its corpus of data.
  • Data sets have run out of original content to collect. Generative AI is now turning to synthetic (generated) data.

 

AI in your life

You already experience the effects of generative AI in your life.

Predictive Text

iPhone text message application with suggested responses

This form of AI suggests words you may insert in your text field. It uses a data set to predict the most common words to follow the previous words, keeping other words in your text field in context.

ChatGPT Free Tier Data Set Date

Through my prompt (used on another page of this Libguide), ChatGPT 3.5 has told me its data set ends in January 2022 (as of May 2024).

Due to the extremely quick development and fast iteration in this space, this guide is a perpetual work-in-progress. We will attempt to keep this guide up to date, but if you see out-of-date information, please let us know.

This guide is representative of the library's view on Generative AI in Education, with De Anza being the context. This is not representative of the College's perspective on the matter. Please ask your instructors and refer to College communications on up-to-date policies on the use of generative AI. The College has an Academic Integrity Policy. The use of Generative AI can be considered plagiarism, cheating, or a study tool - it depends on how you use it and cite it.