Generative Artificial Intelligence (GAI) and machine-generated content have become prominent in educational discussions. Amidst technical jargon and concerns about the impact of traditional learning, writing, and other facets, understanding what these tools are and what they can do can be overwhelming. This toolbox guide provides insights into some commonly used generative AI tools and explains how they are changing the landscape of higher education.
What is Generative AI?
CATL created a short video presentation in Fall 2023 that provides instructors with an introduction to generative AI tools. The video and the linked PowerPoint slides below can help you understand how generative AI tools work, their capabilities, and their limitations. Please note, minor parts of the tool identification in the video have been corrected below in the ‘Common Generative AI Tools’ section.
Introduction to Generative AI – CATL Presentation Slides (PDF)
Microsoft Copilot – UWGB Supported GAI Tool
Microsoft Copilot is the recommended tool for UWGB instructors and students for safety, equity, and GBIT technical support. Using Microsoft Copilot with your UWGB account will bypass the need for individuals to create personal accounts which require providing personal information in the sign-up process. Learn more about Copilot below.
- Microsoft has created its own AI called Copilot using a customized version of OpenAI’s large language model and many of the features of ChatGPT. Users can interact with the AI through a chatbot, compose feature, or the with Microsoft Edge search engine. Microsoft is also rolling out Copilot-powered features in many of its Office 365 products, but these features are currently only available for an additional subscription fee.
- Faculty, staff, and students can access Copilot (which uses both ChatGPT 4.0 and Bing Chat) with their UWGB account. Visit www.copilot.microsoft.com to try out Copilot or watch our short video on how to log in using a different browser. By logging in with UWGB credentials, a green shield and “protected” should appear on the screen. The specifics of what is/is not protected can be complicated, but this Microsoft document is intended to provide guidance. Regardless of potential protections, FERPA and HIPPA-protected information (student or employee) should not be entered.
Common Generative AI Tools
Since OpenAI released ChatGPT in November 2022, various companies have developed their own generative AI applications based on or in direct competition with OpenAI’s framework. Learn more about a few common, browser-based generative AI tools below.
- ChatGPT is an AI-powered chatbot created by OpenAI. The "GPT" in "ChatGPT" stands for Generative Pre-trained Transformer.
- ChatGPT previously required users to sign up for an account and verify with a phone number, but it can now be used without an account. Users can use the chatbot features of ChatGPT both with or without an account (currently version ChatGPT 3.5) or access more advanced models and features with a paid account (currently version ChatGPT 4.0). For more information or to try it yourself, visit chatgpt.com.
- Google has created their own AI tool called Gemini (formerly Google Bard). Similar to ChatGPT and Copilot, Gemini can generate content based on users’ inputs. Outputs may also include sources fetched from Google.
- Using Gemini requires a free Google account. If you have a personal Google account, you can try out Gemini at gemini.google.com.
Note that we are also learning more about potential access to Adobe Express and Firefly (including their image generation features) with UWGB login credentials, at least for employees. Watch this space for additional details as they become available.
What Can Generative AI Tools Do?
The generative AI tools we’ve discussed so far are all trained on large datasets that produce outputs based on patterns in that dataset. User prompts and feedback can be used to improve their outputs and models, so these tools are constantly evolving. Explore below to learn about some use cases and limitations of text-based generative AI tools.
Generative AI tools can be used in a multitude of ways. Some common use cases for text-based generative AI tools include:
- Language generation: Users can ask the AI to write essays, poems, emails, research papers, and Powerpoint presentations, or code snippets on a given topic.
- Information retrieval: Users can ask the AI simple questions like “explain the rules of football to me” or “what is the correct way to use a semicolon?”.
- Language translation: Users can use the AI to translate words or phrases into different languages.
- Text summarization: Users can ask them to condense long texts, including lecture notes or entire books, into shorter summaries.
- Idea generation: Users can use the AI to brainstorm and generate ideas for a story, research outline, email, or cover letter.
- Editorial assistance: Users can input their own writing and then ask the AI to provide feedback or rewrite it to make it more concise or formal.
- Code generation: Users can ask the AI to generate code snippets, scripts, or even full programs in various programming languages based on specific requirements or prompts.
- Image generation: Users can ask the AI to create images or visual content from text descriptions, including illustrations, designs or conceptual art.
These tools are constantly evolving and improving, but in their current state, many have the following limitations:
- False or hallucinated responses: Most AI-powered text generators produce responses that they deem are likely answers based on complex algorithms and probability, which is not always the correct answer. As a result, AI may produce outputs that are misleading or incorrect. When asking AI complex questions, it may also generate an output that is grammatically correct but logically nonsensical or contradictory. These incorrect responses are sometimes called AI "hallucinations."
- Limited frame of reference: Outputs are generated based on the user's input and the data that the AI has been trained on. When asking an AI about current events or information not widely circulated on the internet, it may produce outputs that are not accurate, relevant, or current because its frame of reference is limited to data that it has been trained on.
- Citation: Although the idea behind generative AI is to generate unique responses, there have been documented cases in which an AI has produced outputs containing unchanged, copyrighted content from its dataset. Even when an AI produces a unique response, some are unable to verify the accuracy of their outputs or provide sources supporting their claims. Additionally, AI tools have been known to produce inaccurate information, citations, and can even hallucinate citations
- Machine learning bias: AI tools may produce outputs that are discriminatory or harmful due to pre-existing bias in the data it has been trained.
The potential for GAI tools seems almost endless — writing complete essays, creating poetry, summarizing books and large texts, creating games, translating languages, analyzing data, and more. GAI tools can interpret and analyze language, similar to how human beings can. These tools have become more conversational and adaptive with each update, making it difficult to discern between what is generated by an AI and what is produced by a human, and the machine-learning models they are based upon imitate the way humans learn, so their accuracy and utility will only continue to improve over time.
What Does This Mean for Educators?
The existence of this technology raises questions about which tasks will be completed all or in part by machines in the future and what that means for our learning outcomes, assessments, and even disciplines. Some experts are discussing to what extent it should become part of the educational enterprise to teach students how to write effective AI prompts and use GAI tools to produce work that balances quality with efficiency. Other instructors are considering integrating lessons on AI ethics or information literacy into their teaching. Meanwhile, organizations like Inside Higher Ed have rushed to conduct research and surveys on current and prospective AI usage in higher ed to offer some benefits and challenges of using generative AI for leaders in higher education looking to make informed decisions about AI guidance and policy.
Next Steps for UWGB Instructors
The Universities of Wisconsin have issued official guidance on the use of generative AI, but the extent to which courses will engage with this technology is largely left up to the individual instructor. Instructors may wish to mitigate, support, or even elevate students’ use of generative AI depending on their discipline and courses.
Those interested in using these tools in the classroom should familiarize themselves with these considerations for using generative AI, especially regarding a tool’s accuracy, privacy, and security. As with any tool we incorporate into our teaching, we must be thoughtful about how and when to use AI and then provide students with proper scaffolding, framing, and guardrails to encourage responsible and effective usage.
Still, even for those who don’t want to incorporate this technology into their courses right now, we can’t ignore its existence either. All instructors, regardless of their philosophy on AI, are highly encouraged to consider how generative AI will impact their assessments, incorporate explicit guidance on AI tool usage in their syllabi, and continue to engage in conversations around these topics with their colleagues, chairs, and deans.
Learn More
Explore even more CATL resources related to AI in education:
- Considerations for Using Generative AI Tools
- How Might Generative AI Impact Your Course? (GAI Checklist)
If you have questions, concerns, or ideas specific to generative AI tools in education and the classroom, please email us at catl@uwgb.edu or set up a consultation!
A very relevant topic at the moment. Thank you!