This guide provides an introduction to artificial intelligence (AI), with a focus on generative AI. It offers explanations of the benefits and limitations as well as support for using, citing, researching, and teaching with AI. The purpose of this guide is to support SFCC Faculty and students in exploring and using AI ethically and in productive ways that foster the development of critical thinking. In addition, the guide addresses topics such as AI and academic dishonesty, environmental impacts, copyright, and privacy.
Daiyuu Nobori, "Unraveling AI Complexity - A Comparative View of AI, Machine Learning, Deep Learning, and Generative AI", Wikimedia Commons, licensed under Creative Commons Attribution-Share Alike 4.0 International (CC BY-SA 4.0).
Artificial intelligence refers to the development of computer systems that can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and natural language understanding.
Types of Artificial Intelligence
A chatbot is a software application that uses natural language processing (NLP) and machine learning to simulate conversation with humans, either via text or voice interfaces.
Generative artificial intelligence refers to algorithms and models that can generate new content or data, such as images, videos, music, or text, based on patterns learned from existing information.
Machine learning is a subset of artificial intelligence that involves training computer systems to learn from data and improve their performance over time through experience.
NLP is a subfield of artificial intelligence that deals with the interaction between computers and human language, including text and speech processing, sentiment analysis, machine translation, and dialogue systems.
A large language model is a type of machine learning model that is trained on vast amounts of text data to generate language outputs that are coherent and contextually appropriate.
Disclaimer
These definitions were generated using the Llama 2 large language model and reviewed for accuracy by a Libraries staff member. Generating content like this can be done efficiently using a large language model, but it is important to remember to review the output carefully and acknowledge the source.
In the context of AI, hallucination refers to the phenomenon where a model generates inaccurate or imaginary output that cannot be explained by its training data, often due to overfitting or underfitting.
A prompt is a specific task or question that is given to an AI system to elicit a response or output.
Prompt engineering is the process of designing and refining prompts to elicit desired responses or behaviors from AI systems, in order to improve their performance and versatility.
Parameters are settings or values that are adjusted during the training process to optimize the performance of an AI model, such as the learning rate, regularization strength, or number of hidden layers.
In the context of generative AI, temperature refers to a parameter that controls the "randomness" or "diversity" of generated samples, with higher temperatures resulting in more diverse and less predictable outputs.
In Natural Language Processing and machine learning, tokens refer to individual words or phrases in a text dataset, which are used as input features for models to analyze and understand the meaning of the text.
Training data is the set of examples or inputs used to train an AI system, which helps the model learn patterns and relationships in the data and make predictions or decisions.
https://copilot.microsoft.com/
Generates text, code, and images using ChatGPT and DALL-E. Includes links to some web resources.
Licensed by SFCC
Website - log in with your SFCC Microsoft 365 account to access the SFCC licensed version of CoPilot.
Microsoft
Free through SFCC
When using CoPilot licensed through SFCC, your prompts and responses are not retained by Microsoft or used to train AI models, and your information is encrypted. It is not currently approved for use with confidential university data (e.g., FERPA, HIPAA, PCI, IRB).
ChatGPT
Generates text, code and images.
Proprietary
Web, iOS, Android
OpenAI
Basic tier is free and no account is needed. Upraded version is available for $20/month and requires a user account with OpenAI.
OpenAI Privacy Policy (for non-European users) and OpenAI Privacy Policy (for EU)
Will collect personal information and provide information to partners, in addition to analysis of user behavior.
Trained using the Common Crawl open dataset in addition to resources like Wikipedia, books, and news articles.
Gemini
Generates text, code and images. Includes links to some web sources.
Proprietary; not licensed by SFCC
Google Gemini website and through Google app integrations. Formerly known as Bard.
Basic tier is free but requires Google account. Does not work with @utexas Google accounts but does work with @gmail accounts. Upgraded version (Gemini Advanced) costs $19.99/month as of August 2024 and offers further integration within Google apps, 2 TB storage, priority access for new features, and other premium features.
Proprietary; not licensed by SFCC.
Gemini Apps Privacy Hub; Uses location, past conversation data to provide responses. Will save and share data with other Google products if linked.
https://llama.meta.com/ or https://meta.ai
Generates text, code and images.
Use it immediately through Meta AI or download Llama and deploy locally
Meta
Free
Llama is open with some restriction; not licensed by SFCC
Meta Privacy Policy- Requires submission of name and email in order to download the model. Once downloaded, the model can then be run locally without sharing data.
Trained with data sources similar to those of other LLMs, but only those with publicly available data that are “compatible with open sourcing”. See https://arxiv.org/pdf/2302.13971.pdf for full details on training data and list of data sources.
A more exhaustive list of generative AI tools that may be of interest to university students and faculty is maintained by Ithaka S+R at https://sr.ithaka.org/our-work/generative-ai-product-tracker/
literature searching, data extraction from PDFs, summarizes individual papers using LLMs, synthesize results from multiple papers to create an overall summary
Elicit
Free version provides basic services. Two paid tiers provide increased functionality.
Will collect personal information and provide information to partners, in addition to analysis of user behavior. See more information on their privacy statement.
Like all LLM summaries, the output draws from a limited scope of information and has a tendency to contain biases. All summaries should be read critically and should not be viewed as a total replacement of engaging with the literature. The focus of Elicit is predominantly on synthesizing empirical research. Is not the best fit for humanities research. Articles gathered from Semantic Scholar.
Consensus
literature searching, provides a "consensus meter" of how the results align with your question (yes, maybe, no), summarizes individual papers using LLMs, synthesize results from multiple papers to create an overall summary
Union Square Ventures
Free version provides limited services. Paid tiers provide increased functionality and increased access.
Will collect personal information and provide information to partners, in addition to analysis of user behavior. See more information on their privacy statement.
Like all LLM summaries, the output draws from a limited scope of information and has a tendency to contain biases. All summaries should be read critically and should not be viewed as a total replacement of engaging with the literature. The focus of Consensus is predominantly on synthesizing empirical research. Is not the best fit for humanities research. Articles gathered from Semantic Scholar.
Research Rabbit
Website
https://www.researchrabbit.ai/
Literature mapping - i.e. tracking citations to create a "map" of citations affiliated with a paper. Also provides citations for similar work.
Research Rabbit
Free to use.
Will collect personal information and provide information to partners, in addition to analysis of user behavior. See more information on their privacy statement.
Articles gathered from Semantic Scholar & PubMed.
Inciteful
Website
Literature mapping - i.e. tracking citations to create a "map" of citations affiliated with a paper. Also provides citations for similar work. Ranks affiliated papers by "importance" using PageRank scores.
Inciteful
Free to use.
At the time of writing, Inciteful does not have a published privacy statement. Based on similar products, one would assume that they will collect personal information and provide information to partners, in addition to analysis of user behavior.
Articles gathered from Semantic Scholar, Open Alex, Crossref, & Open Citations.
Generates images, video and audio. Free and paid versions. Available to download and run locally, or use online through Stable Assistant.
Generates images and video in Discord and on the Web. Requires a paid subscription.
Text to image generator from OpenAI. Incorporated into Microsoft CoPilot.
Text to video generator from OpenAI that was released for broader use in December 2024. Available through the ChatGPT Plus subscription which provides basic functionality and with expanded capabilities available for ChatGPT Pro users.
Generate and optimize code in many programming languages. Free and paid subscriptions available.
Generate and optimize code in many programming languages. Available for free.
This guide was created using "Generative AI Tools for Research and Learning" from the University of Texas Libraries, licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0).