After weeks of collaboration with the Bok Center we are finally on the eve of our partnered installation neatly titled "Prompts and Play". The process of setting up this event both logistically and technically has been quite an undertaking, with hours being put in towards each. Personally, I've been tasked with creating and demonstrating a station that showcases the potential of Generative AI through its APIs. To do this I plan on showcasing the different potential types of calls you can make in isolation (text-to-text, text-to-image, voice-to-text, text-to-voice, etc.) and in combination. The latter, doing so in combination, is also the central concern of my capstone project for T127 titled MagicMountain.
Built on the back of multiple interlinking API calls to platforms like OpenAI, Anthropic, Stability, and ElevenLabs, MagicMountain acts a perfect culmination of my work this semester. The function of the project is to generate a picture book based off of real time conversations held between people in a space. This project was inspired by my work early in my career in education spent working with kindergarten to first age children. Though a great portion of the curriculum is purposefully interactive to keep students engaged, the times where lecture was given students had a hard time remembering exactly what was said. This shorter attention span, defined both by student's age and early tech adoption, disrupted the effective learning of material through the lecture format, reducing its effectiveness during instruction. What this project seeks to provide for students is a secondary representation of the material presented during lecture in an age-appropriate format through the creation of picture books generated by the lectures themselves. This fully leverages the generative function of these AI tools allowing for teachers to present information in one mode and have supplementary representational material produced simultaneously and seamlessly with their instruction.
Now, with this project laid out all nice and tidy, I do not want to present it as faultless, as there certainly are issues that I would like to turn into areas for improvement. Foremost is the question of the program's relationship to classroom privacy. With teachers already under pressure to teach to specific standards and not present information off the cuff, I fear this technology could be coopted as a form of surveillance, assuming the data from the product would be available to the district who will likely be the ones footing the bill for such a service. On top of this, with the recording of classroom audio, teachers and students alike may have experiences akin to the Hawthrone effect and not feel able to express their authentic selves while the recording is taking place. Pursuing solutions to these questions will be essential future work to be done alongside any technical optimization.
Though the future of this project will face these hurdles, its present task is simply being prepared for the upcoming installation. Having the opportunity to share the power of GenAi platforms APIs with my fellow educators and having MagicMountain receive its first round of engagement and user feedback is an exciting affair to look forward to.
Comments