1. Seattle Colleges home page
  2. Employees
  3. Faculty eLearning & Instructional Design
  4. AI Task Force
  5. AI Resources for Faculty

AI Resources for Faculty

Book laying open on dark background, pages are dark with glowing blue lines suggesting technical drawings

Introduction

AI (Artificial Intelligence) is not a new concept. The 1950s saw the academic conceptualization of AI in science and technology.  

ChatGPT (Chat Generative Pre-Trained Transformer) is a form of AI many became familiar with in late 2022. As librarian Kristen Palmer writes and quotes, "Chat GPT is a specific chatbot built as an online assistant that can talk back-and-forth with a user. The creators, OpenAI, describe ChatGPT by saying, 'We’ve trained a model called ChatGPT which interacts in a conversational way’” (Butler University lib guide). 

You can use AI tools to get classroom activity ideas and non-essential life advice. You can use generative AI (GenAI) platforms like Scribe to create an accessible step-by-step guide for any process. Do keep in mind the environmental impacts of AI technology. The video below by Katie Tarasov describes the energy required to run data centers. Feel free to watch the first five minutes, the whole 15 minutes, or as much as you'd like:

Equity and AI 

There are many ways in which inequities are produced and reproduced by AI tools. Racial, gender, and other biases are present in the data sets used to train GenAI tools. Critical analysis of AI tools through an equity lens helps to clarify their risks and best use. This equity in AI page serves as a starting point for learning about racial bias in the AI industry as well as organizations that tackle equity issues, such as the DAIR Institute and Black in AI.  

The following video explains the concept of algorithmic bias: VIDEO

Despite its many limitations, including, but not limited to the perpetuation of racism, sexism, and other biases and unreliable information, a lot of people are currently using GenAI, including students. Consequently, an AI course policy concerning its usage is very useful to help students make informed decisions. You can browse these 200 examples of AI course policies for inspiration.

AI Course Policies 

If you haven’t crafted an AI policy for your courses, there are many existing examples to pull from. Alternatively, co-creating policy with students generates the most buy-in and can lead to more transparent communication about AI use. It can also be helpful to poll your students on AI use in their lives and discuss the topic with colleagues. For further exploration, here is an extensive collection of syllabus statements with many discipline-specific examples. You can submit your own as well! 

Additional resources: 

FAQ

I am new to AI. Where should I begin? 

Here is a recommended sequence for becoming familiar with AI tools: 

  1. Quick guide to Artificial Intelligence – overview of six types of AI
  2. If you learn well by listening, the podcast "In Machines We Trust" offers discussions about automation and the impact of AI on everyday life.
  3. These 10-minute videos by PBS about AI are helpful for visual learners. 

Intro to LLMs

Large Language Models (LLMs) are a type of GenAI that create text in response to written commands, called “prompts”.

  • LLMs, such as ChatGPT, are sometimes called "chatbots" because you interact with them like an online chat.
  • Some LLMs work with more than just text as input and output. Those tools are considered "multi-modal" because they can process and generate images, speech, video, music, and other media.

Please note that LLMs use a lot of computational power and run on physical servers in data centers. It is estimated that one LLM produces approx. 660,000 pounds of carbon dioxide emissions via its use of electricity and water. Data centers have reached roughly the same carbon footprint as the aviation industry and will likely surpass it soon (Crawford 2021).

How do LLMs work?

Large Language Models create their output by making predictions about which words, sentences and paragraphs are most likely to follow others. The model uses statistical probabilities of one word relating to another (patterns) to produce new text in response to questions and prompts. 

When you supply the LLM chatbot with a prompt (i.e. enter text), it responds with a set of words which it calculates are the best fit to be an answer to the pattern of words set by your prompt.

This can make it seem like you're chatting with something that thinks: but in reality, the program doesn't understand what's being said. It is instead generating sequences of words based on patterns, probability, and a few other factors that can differ depending on which “foundation model” the chatbot is using.

This is sometimes called "auto-complete on steroids" - which is not incorrect!

What is a foundation model?

Examples of LLM foundation models are:

  • OpenAI’s ChatGPT-3.5 and ChatGPT-4o
  • Google’s Gemini
  • Anthropic’s Claude 3
  • and many more

A foundation model is another term for an LLM, but one which implies a broader scope. As an LLM, it is a pre-gathered and pre-trained map of data. The data can be text, images, audio, and/or video. The probable relationships within it are built using machine-learning techniques. A chatbot or other interface can use this model to generate its "own" responses to prompts.

The term "foundation model" in this context was introduced in 2022 by Stanford University, with the launch of their Center for Research on Foundation Models (opens in new tab). The term emphasizes that these models serve as foundations for specific contracted uses, e.g. AI tools for businesses, government, or research.

 

Here is an interactive online article (8-12 minutes) that does an excellent job of showing, as well as telling, how and why AI tools work.

Here is a list of common concerns for educators with suggestions for navigating anticipated issues. Faculty can also refer to this list of pro-active approaches to AI. 

Here are a few suggestions for crafting effective AI syllabus statements: 

  • Check to see if you wrote your AI syllabus statement similarly to your plagiarism statement. The two can inform each other. 
  • See if you clearly defined what you mean by using AI and/or ChatGPT, so that students understand what it is and what it means to use it. Can it be used for some assignments, but not others?  
  • If you feel that using AI for course assignments is detrimental to learning, see if you clearly explain why.  
  • If you haven’t already, offer opportunities to discuss AI with students in class, during office hours, or via email.  
  • If you'd like your students to use AI for some assignments, let them know how you've marked those assignments in the syllabus or elsewhere. 
  • Make sure that you have let students know how you'd like them to document and/or cite AI usage. 
  • Show it in addition to saying it: A short video or a synchronous/live class demo and activity will help students ask clarifying questions and demonstrate their understanding. 

 

There are more than 8,000 data centers globally, but it’s not nearly enough to keep up with the power needs of generative AI:

  • One ChatGPT query takes about 10 times as much energy as a typical Google search.
  • Training one large language model can produce as much CO2 as the entire lifetime of five gas-powered cars and use as much water as a small country.
  • It is estimated that one Large Language Model produces approx. 660,000 pounds of carbon dioxide emissions via its use of electricity and water.
  • Data centers have reached roughly the same carbon footprint as the aviation industry and will likely surpass it soon (Crawford 2021).

Even if we generate enough power, our aging power grid is increasingly unable to handle transmitting electricity to where it’s used. Some companies are building closer to where power is generated, while others invest in alternate energy sources and improvements to the grid.

The amount of information about AI is enormous as well as full of rabbit holes and speculations. It can be helpful to scan a collection of headlines and limit deep dives to the most promising topics. Here are two newsletters that provide small chunks of information on a regular basis:   

This guide to staying informed provides tips for learning together with colleagues and a sampling of recent headlines, research articles, and product announcements.  

GenAI tools like ChatGPT have capabilities that can make them suitable for learning activities and assignments. The ideas in this AI assignment guide can provide a starting point for developing course materials that integrate AI and teach students foundational skills.  The guide includes tips for implementation and further resources. You can also reference this AI discussion guide for instructors as well as this overview of pro-active approaches to AI.

If you have trouble accessing any of the links above, you can self-enroll in this AI Resource shell on Canvas first.