We explored 20 original AI innovation concepts for O’Reilly’s online platform. From this bank of initial proposals, two were selected for further design refinement and development as fully realized digital products.
Create novel educational tools tailored to an individual’s learning behaviors by experimenting with latest advancements in the AI landscape.
HONORING AN ARCHIVE OF EXTENSIVE LEARNING MATERIAL
As a leading voice in digital technology literacy, O’Reilly was uniquely positioned to create AI functionality in an educational context. Its expansive pool of resources spanning books, tutorials, and training programs presented an opportunity to innovate new AI-driven learning experiences that centered their learners’ personal journey while staying true to their roots and values.
APPROACH
In the first phase, we delivered nearly 20 distinct proposals, of which 2 were finalized in collaboration with the client team. Each of the two final digital experiences is a novel representation of dynamic learning formats. Using state-of-the-art AI models, we addressed gaps in dissemination of educational content.
“Unlike static design tools, prototyping with code helps us address design issues in real time. This approach allows us to be more user-centric.”
INTERFACE 1: PERSONALIZED GOALS
Instead of stitching together resources from across dense books and lengthy courses, Personalized Goals compiles discrete bits of information using AI to create more intuitive and holistic learning paths. Learners can state their objectives in natural language with the rest being taken care of by the interface.
CREATING VALUE
Giving users the opportunity to create a personalized learning plan centered around their particular needs can help break down information into more accessible, digestible chunks.
Dynamic inputs for learning objectives and example learning goals relieves the user of the burden of crafting complete text prompts.
A modified version of the users’ prompt is sent to the Miso Ask API to suggest related resources from the entire O’Reilly library.
A selection of learning resources are handpicked for the user who can begin, complete, or revisit their self-paced journey whenever they like.
Change your goal? Change your prompt. Get an updated personalized plan. Progress can be tracked over time by saving plans to your profile.
“Generative products break traditional waterfall processes. Designing directly in code helps us frame expectations in the face of uncertainties a lot faster.”
INTERFACE 2: DISTILLED VIDEOS
Distilled Videos make searching for video content more precise and personable. The tool extracts snippets from longer video courses that often contain several layers of intertwined content. This format is especially beneficial for visual learners who respond better to shorter content lengths or are interested in staying abreast with different topics in an approachable way.
BRIDGING THE GAP
Video courses often run for hours and can be intimidating to keep up with. Tearing down this one-way instruction channel by reinstating a learner-led pattern of discovery makes it much easier to move from top-level topics into specific content one might be looking for.
When a learner prompts a video, it is embedded with an Embeddings model by OpenAI, which is also used to chunk video transcripts into paragraphs with timecodes.
A vector search is performed on the transcript embeddings (each in a 1536-dimensional vector space) in a AWS RDS PostgreSQL Database.
The search results are a list of clips to be featured in the distilled video course. These are used to construct an on-the-fly video that can be played in the video player.
The 'Extend Clip' functionality adds a minute onto the video timeline in case the learner wishes to stay with a particular clip for longer.