Want to jump right to material and activities you can use in class right now? Filter for what interests you in the Resources or Strategies sections.
Looking for an upcoming event or presenter on the topic of AI and education? Check out upcoming events and related speakers.
Submit an online resource, assignment, or event to contribute to this collection with our input form, currently in beta.
A simple flowchart can help overwhelmed students fit voting into their schedule. Learning With AI has partnered with AppsForEducation.ai to create the How To Vote Flowchart generator, which uses generative AI to create custom voting instructions for students on your campus.
As the sudden rise of generative AI sends shock waves through society, no sector is feeling the disruption as abruptly and pervasively as education. Large Language Models like ChatGPT threaten to make the term paper obsolete, while diffusion-based text-to-image generators like DALL-E, Stable Diffusion, and Midjourney offer a way to create images that's completely outside the workflow of most illustrators and designers, whether trained in analog or digital media.
Rather than try to ban this technology from classrooms outright, the Learning With AI project asks if this moment offers an opportunity to introduce students to the ethical and economic questions wreaked by these new tools, as well as to experiment with progressive forms of pedagogy that can exploit them.
While this site includes plenty of strategies for using generative AI, we've created the acronym IMPACT RISK to remember its downsides. Download free infographics at bit.ly/impactrisk, or watch the explainer video.
Learning With AI has been featured in ABC-TV, WGME-TV, News Center Maine, and Meta News. Learn more in the official press release.
Learning With AI
was originally launched by the
New Media program
and
Center for Innovation and Teaching and Learning,
who
built this site to survey the challenges and offer resources to
any
educator or student wondering how to accommodate this disruption.
Other partners in the initiative include related units at the University of Maine and the AI in Education Google Group organized by Daniel Stanford. More on our partners.
When you type a question or request into a text generator like ChatGPT, the bot will type back an answer based on the data it has been trained on. This is typically information available on websites such as Wikipedia and news outlets. Media generators like Midjourney will attempt to show you an image they have constructed from existing images with similar caption descriptions.
Notable text generators include ChatGPT and Google's Bard. Notable image generators include DALLE-2, Stable Diffusion, and Midjourney.
Generative AI is a more versatile version of the predictive text you see when typing a message on your phone, which suggests words you are likely to type next.
Generative AI is like a disembodied, clueless brain. Researchers feed this "neural network" information from the Internet and then ask it questions. When the network guesses better answers to these questions, researchers reinforce the connections that led to those answers. With enough training by researchers, the system can improve its accuracy.
Every piece of information fed to generative AI--like the words "rain," "umbrella", and "dinosaur"--occupies a different point in an abstract mathematical space. Researchers train the system to recognize correlations among these points in existing web pages like Wikipedia. That enables the AI to measure semantic distances between points; for example, the words "umbrella" and "rain" would be closer in this space than "umbrella" and "dinosaur." When prompted by the user with a string of words, the system finds the corresponding points in space and gives back a text or image that essentially represents the average of all those points.
Each point in the space described above is a multidimensional vector. In the case of large language models like ChatGPT, the number of variables in each vector can be in the millions. These variables can be combined in various ways to calculate the "average" of any set of existing points. This average may not have been a point in the original dataset, which is why image generators like Midjourney can create simulated photographs that have never existed before based on photographs it has already categorized.
No, we created it to share information and tactics with anyone interested in how generative AI may transform education. That said, we encourage you to cite LearnWithAI.org if you've found the site helpful.
We make no claims about the validity of content on third-party sites, however.
Yes! We'd love to hear your suggestions--just contact us via the link below.
Learning With AI was convened by UMaine's New Media major and Center for Innovation in Teaching and Learning. Other partners in the initiative include the AI in Education Google Group organized by Daniel Stanford and UMaine's the College of Education and Human Development, Department of Electrical and Computer Engineering, Digital and Spatial History Lab, and UMaine AI. Learn more in the official press release.
This program is not for you; we're focused on how to cope with generative AI when learning any subject. Fortunately, the UMaine AI cluster offers degrees and certificates in Computer Science, Data Science, and related fields.
No.
Let us know via the link below.
A project of the University of Maine's Center for Innovation in Teaching and Learning and New Media major
