How AI Relies on Structured Knowledge

On this episode of the Crazy Wisdom Podcast, host Stewart Alsop welcomes Jessica Talisman, a senior information architect deeply immersed in the worlds of taxonomy, ontology, and knowledge management. The conversation spans the evolution of libraries, the shifting nature of public and private access to knowledge, and the role of institutions like the Internet Archive in preserving digital history. They also explore the fragility of information in the digital age, the ongoing battle over access to knowledge, and how AI is shaping—and being shaped by—structured data and knowledge graphs. To connect with Jessica Talisman, you can reach her via LinkedIn. 

Check out this GPT we trained on the conversation!

Timestamps

00:05 – Libraries, Democracy, Public vs. Private Knowledge
Jessica explains how libraries have historically shifted between public and private control, shaping access to knowledge and democracy.

00:10 – Internet Archive, Cyberattacks, Digital Preservation
Stewart describes visiting the Internet Archive post-cyberattack, sparking a discussion on threats to digital preservation and free information.

00:15 – AI, Structured Data, Ontologies, NIH, PubMed
Jessica breaks down how AI trains on structured data from sources like NIH and PubMed but often lacks alignment with authoritative knowledge.

00:20 – Linked Data, Knowledge Graphs, Semantic Web, Tim Berners-Lee
They explore how linked data enables machines to understand connections between knowledge, referencing the vision behind the semantic web.

00:25 – Entity Management, Cataloging, Provenance, Authority
Jessica explains how libraries are transitioning from cataloging books to managing entities, ensuring provenance and verifiable knowledge.

00:30 – Digital Dark Ages, Knowledge Loss, Corporate Control
Stewart compares today’s deletion of digital content to historical knowledge loss, warning about the fragility of digital memory.

00:35 – War on Truth, Book Bans, Algorithmic Bias, Censorship
They discuss how knowledge suppression—from book bans to algorithmic censorship—threatens free access to information.

00:40 – AI, Search Engines, Metadata, Schema.org, RDF
Jessica highlights how AI and search engines depend on structured metadata but often fail to prioritize authoritative sources.

00:45 – Power Over Knowledge, Open vs. Closed Systems, AI Ethics
They debate the battle between corporations, governments, and open-source efforts to control how knowledge is structured and accessed.

00:50 – Librarians, AI Misinformation, Knowledge Organization
Jessica emphasizes that librarians and structured knowledge systems are essential in combating misinformation in AI.

00:55 – Future of Digital Memory, AI, Ethics, Information Access
They reflect on whether AI and linked data will expand knowledge access or accelerate digital decay and misinformation.

Key Insights

1. The Evolution of Libraries Reflects Power Struggles Over Knowledge: Libraries have historically oscillated between being public and private institutions, reflecting broader societal shifts in who controls access to knowledge. Jessica Talisman highlights how figures like Andrew Carnegie helped establish the modern public library system, reinforcing libraries as democratic spaces where information is accessible to all. However, she also notes that as knowledge becomes digitized, new battles emerge over who owns and controls digital information​​.
2. The Internet Archive Faces Systematic Attacks on Knowledge: Stewart Alsop shares his firsthand experience visiting the Internet Archive just after it had suffered a major cyberattack. This incident is part of a larger trend in which libraries and knowledge repositories worldwide, including those in Canada, have been targeted. The conversation raises concerns that these attacks are not random but part of a broader, well-funded effort to undermine access to information​​.
3. AI and Knowledge Graphs Are Deeply Intertwined: AI systems, particularly large language models (LLMs), rely on structured data sources such as knowledge graphs, ontologies, and linked data. Talisman explains how institutions like the NIH and PubMed provide openly available, structured knowledge that AI systems train on. Yet, she points out a critical gap—AI often lacks alignment with real-world, authoritative sources, which leads to inaccuracies in machine-generated knowledge​​.
4. Libraries Are Moving From Cataloging to Entity Management: Traditional library systems were built around cataloging books and documents, but modern libraries are transitioning toward entity management, which organizes knowledge in a way that allows for more dynamic connections. Linked data and knowledge graphs enable this shift, making it easier to navigate vast repositories of information while maintaining provenance and authority​​.
5. The War on Truth and Information Is Accelerating: The episode touches on the increasing threats to truth and reliable information, from book bans to algorithmic suppression of knowledge. Talisman underscores the crucial role librarians play in preserving access to primary sources and maintaining records of historical truth. As AI becomes more prominent in knowledge dissemination, the need for robust, verifiable sources becomes even more urgent​​.
6. Linked Data is the Foundation of Digital Knowledge: The conversation explores how linked data protocols, such as those championed by Tim Berners-Lee, allow machines and AI to interpret and connect information across the web. Talisman explains that institutions like NIH publish their taxonomies in RDF format, making them accessible as structured, authoritative sources. However, many organizations fail to leverage this interconnected data, leading to inefficiencies in knowledge management​​.
7. Preserving Digital Memory is a Civilization-Defining Challenge: In the digital age, the loss of information is more severe than ever. Alsop compares the current state of digital impermanence to the Dark Ages, where crucial knowledge risks disappearing due to corporate decisions, cyberattacks, and lack of preservation infrastructure. Talisman agrees, emphasizing that digital archives like the Internet Archive, WorldCat, and Wikimedia are foundational to maintaining a collective human memory​​.