Content

Signals, Systems, and Speech — Rethinking Meaning in the Age of AI

What happens when language models meet logic? Larry Swanson and I explore neurosymbolic AI, voice tech, and the human meaning behind machine-readable content.

Rob Hoeijmakers

11 Apr 2025 • 5 min read

Having a deep conversation with Larry Swanson.

Part 2 of a conversation with Larry Swanson

In part one, we explored the world of podcasting. In this second conversation, the format opens up: we dive into knowledge infrastructure, AI's impact on identity and work, and the emerging architectures of meaning in a digitised world.

Larry had just landed a new role as Community Growth Manager at a knowledge graph and AI company. The way he got there says a lot about how modern careers are evolving.

Listening as a Career Move

Larry joined the Never Search Alone programme, a methodical approach to career reflection and outreach. He talked to thirty people. Not recruiters, but peers.

The goal: identify resonance.

What energises you? What patterns emerge? Through these conversations, Larry uncovered a clear fit: he lights up when he talks about common infrastructure, semantics, community. Within weeks, he had a job offer.

His new role blends community building, AI evangelism, and developer outreach. He describes it as a cross between developer relations, content strategy, and live representation. As he put it, he's giving a human face to a deeply technical field. And it suits him.

Probabilistic Meets Deterministic

Our conversation moved quickly from job titles to the deeper shifts behind them. Larry works at the intersection of knowledge graphs and LLMs.

He sees how neuro-symbolic loops combine structured, ontology-driven data with pattern-seeking models. These hybrid systems are already powering enterprise-level search and reasoning.

We explored how deterministic knowledge systems (like Prolog, OWL, or RDF) give shape to the otherwise fluid outputs of LLMs. I shared my experience with companies building RAG pipelines (Weaviate, Zeta Alpha) that layer domain-specific knowledge over general-purpose AI. Without this structure, large language models hallucinate. With it, they inform.

💡

What are neurosymbolic loops? Neural systems like LLMs are flexible and generative, but prone to hallucination. Symbolic systems like knowledge graphs are structured and logical, but rigid. A neurosymbolic loop combines both: LLMs generate or interpret content, symbolic systems validate and refine it, and the cycle continues. This interplay grounds creativity in fact and brings structure to fluidity.

Speaking is the New Writing

One theme that surfaced was the return to orality. I’ve dictated thousands of prompts, and I see voice as not just an interface, but a form of data generation. Whisper-level ASR combined with LLMs is not just transcription—it's transformation.

That shift isn’t just technical. It’s cognitive. Spoken prompts invite a different kind of thinking. There's improvisation, rhythm, a tempo of thought. And that, in turn, changes the kind of data we generate. It changes how AI listens.

💡

Transformer-based ASR models like Whisper process audio in parallel, allowing for accurate live captioning and translation. More background in The Silent Revolution: How AI is Transforming Speech Recognition.

Websites for Machines

We also talked about SEO. I’ve worked in SEO for over a decade, but now I see its endgame. Optimising for Google is no longer about humans reading pages. It's about machines parsing meaning. Structured data, schema.org, entity graphs — these are the new foundations. Human visitors are rare. The web, increasingly, is for robots.

I recently came across a sharp distinction: Google is optimised for keyword intent; LLMs for conversation intent.

That nails the transition. Traditional SEO was about matching search terms. Today's systems respond to how people think aloud. They interpret narrative intent, not just lexical signals.

That doesn’t mean content is obsolete. It means we publish differently. We publish as signals, as identities, as connected nodes. And ironically, the human part becomes more important: voice, conviction, presence.

Where Are We Going?

Larry and I share concern about the social impacts of these shifts. LLMs are not neutral tools. Their integration into education, healthcare, and identity systems raises deep questions. We agreed: it’s not the tech that changes society. It’s the way we relate to it.

For Larry, that means bringing clarity to complex systems. For me, it means articulating purpose in a time of overload. We both believe in infrastructure with meaning.

Larry has also been exploring these questions more deeply in the TruthCollapse series with Noz Urbina, where they discuss the blurred lines between fact, fiction, and generated content. It’s a fitting name: the collapse isn’t just technical, but epistemological. What we believe—and why—is being reshaped by systems we barely understand. Their conversations add another layer to this dialogue: one of responsibility, narrative, and trust.

This was part two of our ongoing dialogue. More will follow once Larry is settled in to his new job (and life).