Prompt-Ed

Motivation

With AI becoming an increasingly prevalent and accessible part of our lives through free/freemium offerings like ChatGPT, Dall-E, Bard, Bing AI, Llama, amongst others, the importance of proactively educating our youth about the limitations and responsible usage of AI is obvious.

It would be as easy to fear and reject this new family of technologies as it would be to over-idolize them and assign an expertise or cognition to them which is not yet present. While the AI reliability question is one that has been engaged in academia for quite a while, Prompt-Ed aims to target this educational goal at a younger audience. Through engaging visuals and whimsical plot, students are guided through the process of prompt engineering, as well as issues like understanding how DALL-E is trained on existing art, the debate around attribution for data used to train large models, and bias in datasets.

What It Does

With a intuitive drag and drop interface and diverse selection of suggested style and composition modifiers, students are provided with guided experience in effectively using specific prompts to achieve their desired outcomes. In addition, the lesson plan simultaneously introduces students to concepts such as style transfer, as well as the moral and legal issues surrounding them, such as artist attribution.

How We Built It

Using Next.js and a serverless Firebase function hooked into OpenAI's Dall-E API to serve images to the student, we focused mainly on the drag-and-drop interface and lesson plans.

Our Team

Elliot Roe and Tiger Peng, the Georgia Institute of Technology; Duncan Johnson and Michelle Zakaria, Tufts University

Challenges

We didn't have quite enough time to build out the required functionality for the more advanced lessons, and what is currently hosted on the website only has the content of the first lesson as well as the playground for students to experiment.

What We're Proud Of

The drag-and-drop interface solidified early on during the design process and we believe it was a very solid design choice to enable younger learners, who tend to operate more intuitively, to perform free form experimentation, reordering the relationships between the objects in their prompt, and quickly understanding the impacts of tweaking a specific variable, such as style or camera angle.

What's next for Prompt-Ed

Prompt-Ed may be continued under the care of the educational technology non-profit BX-Coding, which is run and operated by two of our team members, Elliot Roe and Duncan Johnson.