Crazy Wisdom
Episode #505: From Big Data to Big Meaning: Jessica Talisman on the Hidden Architecture of Knowledge
- AI & Agents
- Knowledge & Learning
- Contextually
- knowledge management
- ontology
- controlled vocabularies
- information architecture
- VibeCoding
- ETL
- ELT
This video isn't available here. Try YouTube, or listen to the audio below.
Watch on YouTubeAbout this episode
Timestamps
- 00:00Stewart Alsop welcomes Jessica Talisman to discuss Contextually, ontologies, and how controlled vocabularies ground scalable systems.
- 05:00They compare philosophy’s ontology with information science, linking meaning, categorization, and sense-making for humans and machines.
- 10:00Jessica explains why SQL and Postgres can’t capture knowledge complexity and how neuro-symbolic systems add context and interoperability.
- 15:00The talk turns to library science’s split from big data in the 1990s, metadata schemas, and the FAIR principles of findability and reuse.
- 20:00They discuss neutrality, bias in corporate vocabularies, and why “touching grass” matters for reconciling internal and external meanings.
- 25:00Conversation shifts to interpretability, cultural context, and how Western categorical thinking differs from China’s contextual knowledge.
- 30:00Jessica introduces process knowledge, documentation habits, and the danger of outsourcing how-to understanding.
- 35:00They explore knowledge as habit, the tension between break-things culture and library design thinking, and early AI experiments.
- 40:00Libraries’ strategic use of AI, metadata precision, and the emerging role of AI librarians take focus.
- 45:00Stewart connects data labeling, Surge AI, and the economics of good data with Jessica’s call for better knowledge architectures.
- 50:00They unpack content lifecycle, provenance, and user context as the backbone of knowledge ecosystems.
- 55:00The talk closes on automation limits, human-in-the-loop design, and Jessica’s vision for collaborative consulting through Contextually.
- Ontology is about meaning, not just data structure. Jessica Talisman reframes ontology from a philosophical abstraction into a practical tool for knowledge management—defining how things relate and what they mean within systems. She explains that without clear categories and shared definitions, organizations can’t scale or communicate effectively, either with people or with machines.
- Controlled vocabularies are the foundation of AI literacy. Jessica emphasizes that building a controlled vocabulary is the simplest and most powerful way to disambiguate meaning for AI. Machines, like people, need context to interpret language, and consistent terminology prevents the “hallucinations” that occur when systems lack semantic grounding.
- Library science predicted today’s knowledge crisis. Stewart and Jessica trace how, in the 1990s, tech went down the path of “big data” while librarians quietly built systems of metadata, ontologies, and standards like schema.org. Today’s AI challenges—interoperability, reliability, and information overload—mirror problems library science has been solving for decades.
- Knowledge is culturally shaped. Drawing from Patrick Lambe’s work, Jessica notes that Western knowledge systems are category-driven, while Chinese systems emphasize context. This cultural distinction explains why global AI models often miss nuance or moral voice when trained on limited datasets.
- Process knowledge is disappearing. The West has outsourced its “how-to” knowledge—what Jessica calls process knowledge—to other countries. Without documentation habits, we risk losing the embodied know-how that underpins manufacturing, engineering, and even creative work.
- Automation cannot replace critical thinking. Jessica warns against treating AI as “room service.” Automation can support, but not substitute, human judgment. Her own experience with a contract error generated by an AI tool underscores the importance of review, reflection, and accountability in human–machine collaboration.
- Collaborative consulting builds knowledge resilience. Through her consultancy, Contextually, Jessica advocates for “teaching through doing”—helping teams build their own ontologies and vocabularies rather than outsourcing them. Sustainable knowledge systems, she argues, depend on shared understanding, not just good technology.
Episode transcript
Welcome to the Crazy Wisdom Podcast. This podcast is for you. If you have an insane drive to find the truth of things, it's not the good answers that we seek, but the good questions. I interview a range of different guests from many different fields, all with the intention to uncover the simple truths that are hidden in plain sight. Most people don't want to go there. I go there, my guests go there, and you benefit. Please let me know if you enjoy these episodes and as always, subscribe on itunes, Spotify, or wherever you listen to the podcasts.
Welcome to the Crazy Wisdom Podcast. I've got Jessica Talisman here and this is the second time we've done an interview and she's running contextually as a consultancy and very excited to get into that and all things knowledge management. Welcome to the show.
Thank you, Sura. It's great to be here. I'm excited for our conversation. It's been a while.
Yeah. Last time you came in and totally changed everyone's minds on ontology ontologies, particularly within big companies. And so far, everything I've seen so far is like, there's the. There's the people like me who are Vibe coding, and it doesn't feel like I need an ontology yet, but if I were to build a bigger company and do anything with AI, it feels like ontologies are actually really important. Is that accurate? Is that really important?
And I've had interesting experiences with people that have Vibe coded apps, you know, starting with nothing, nothing between, like, you might have a postgres database or something connected, you know, depending on how you determine or decide to collect and deploy. And so what's interesting is with Vibe coding, we're seeing, you know, Vibe coded engineering or apps where, you know, people will deploy, problems will occur. Right. I mean, they're going to be data issues. And what I'm seeing most often is the need for etl, elt, or handling of transformations, for example, as something scales and builds and grows. So getting in front of that sooner than later, hired your bump. Because I've seen a lot of issues, you know, down downstream and upstream. So.
Your first full transcript is free. After that, an email opens every transcript in the index — a list of readers we can write to, not a guest book.