Summary
Large Language Models (LLMs) are to language as spreadsheets are to numbers: tools for modeling, exploration, and development. Among their many capabilities, LLMs can alleviate chores related to the design and implementation of information architectures. But doing so requires venturing beyond chat-based interfaces. In this brief demonstration, we'll see how to use OpenAI's API and a few open source command line tools to re-categorize content in a 1,000+ page website. The techniques demonstrated can be extended to other common content organization tasks.
Key Insights
-
•
The discoverability of older content on blogs is often a challenge, making accurate tagging essential.
-
•
LLMs like GPT-4 can significantly reduce the manual workload involved in organizing content.
-
•
A structured approach to task automation can be effectively implemented using a four-step framework.
-
•
Having a clean, obvious taxonomy is crucial for successful AI interaction and content tagging.
-
•
Reviewing AI-generated changes before implementation is necessary to ensure accuracy and relevance.
-
•
AI can provide valuable insights, potentially improving existing classification systems.
-
•
LLMs can function as both a tagging assistant and a surveyor of content relevance.
-
•
Automation can save considerable time and effort, exemplifying the potential of AI in daily tasks.
-
•
This process can be scaled to various content management systems beyond static site generators.
-
•
Engaging with the community for feedback can enhance AI implementation experiences.
Notable Quotes
"I'm an information architect, which means that I organize websites and apps to make stuff easier to find and understand."
"This is a case study on one such toil that I've been removing using LLMs."
"The more recent stuff gets more attention and the older stuff has discoverability problems."
"The first step is gathering the information and the second step is reviewing the changes that the LLM is proposing to make."
"I had to clean this taxonomy up, and I eliminated terms like TAOI."
"GPT actually functioned as an assistant, not just in retagging things, but also in the taxonomy itself."
"It took about three hours from beginning to end to do this, and I was only involved in about two of those hours."
"Be open to contributions; the LLM might help improve your taxonomy."
"I expect that this process is scalable across other systems."
"Good luck as you venture into this wild new world."
More Videos
"Design is a poorly designed word."
John MaedaMaking Sense of Enterprise UX
June 9, 2016
"There are a lot of variability in how people approach their intent or purpose."
Indi YoungThinking styles: Mend hidden cracks in your market
January 8, 2025
"Data should be presented in ways that empower users to make immediate decisions."
Chris ChapoData Science and Design: A Tale of Two Tribes
May 13, 2015
"Material design is a framework for creating web and mobile experiences."
Elizabeth ChurchillExploring Cadence: You, Your Team, and Your Enterprise
June 8, 2017
"We faced challenges as individuals like human connections and networking."
Rusha SopariwalaRemote, Together: Craft and Collaboration Across Disciplines, Borders, Time Zones, and a Design Org of 170+
June 9, 2022
"The most common way people give up their power is by thinking they don't have any."
George AyeThat Quiet Little Voice: When Design and Ethics Collide
November 16, 2022
"I learned that being in the red for one person was crying and for another, it was complete silence."
Ariba JahanTeam Resiliency Through a Pandemic
January 8, 2024
"The best outcomes come from shared goals and alignment between teams."
Julie Gitlin Esther RaiceDesign as an Agent of Digital Transformation at JPMC
June 9, 2021
"Ethical decisions will change over time; they're not static."
Matteo GrattonCan Data and Ethics Live Together?
October 1, 2021
