Episode #554: When Fluency Lies: The Knowledge Problem at the Heart of AI

Episode #554: When Fluency Lies: The Knowledge Problem at the Heart of AI

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with Larry Swanson, creator of the Knowledge Graph Insights Podcast, for their second conversation together. The two cover a wide range of interconnected topics, starting with a correction Larry makes about the true origin of the term "artificial intelligence," tracing it back to the 1956 Dartmouth Conference and its distinction from Norbert Wiener's cybernetics. From there, the conversation moves through the history and structure of knowledge graphs, ontologies, RDF (Resource Description Framework), and the W3C standards process, touching on concepts like the T-box, A-box, and C-box, as well as the 25th anniversary of the Semantic Web paper. Stewart and Larry also dig into the limitations of large language models — particularly around reasoning, confabulation, and what Larry describes as "cognitive surrender" — and why symbolic AI and knowledge engineering may hold answers that the neural network world hasn't fully embraced. The episode also ventures into consciousness, panpsychism, Michael Pollan's ideas, and Stewart's own hands-on experience vibe coding a personal chatbot to replace functionality he feels he's lost with recent changes to Claude. Larry's podcast can be found at kgi.fm [https://knowledgegraphinsights.com/]. Timestamps 00:00 - Stewart introduces Larry Swanson; Larry corrects the record on AI's origin, distinguishing it from Norbert Wiener's cybernetics at the 1956 Dartmouth conference. 05:00 - Larry discusses interviewing semantic web paper coauthors on its 25th anniversary; RDF's hidden ubiquity compared to SIM cards powering everything invisibly. 10:00 - Knowledge graphs explained through t-box terms, a-box assertions, and Dave McComb's c-box; IKEA's three-layer knowledge graph as a practical example. 15:00 - Stewart connects metadata complexity to AI needs; faceted search explained as c-box attributes driving product filtering experiences. 20:00 - RDF 1.2 reification standards discussed; W3C's rigorous recommendation process powering governments and enterprises worldwide through collaborative standards. 25:00 - Cyc project examined as influential "successful failure"; Pat Hayes bringing description logic into semantic web; LLMs lacking true reasoning capability. 30:00 - Epistemological fault lines between human and computer intelligence; cognitive surrender paper reveals no intelligence threshold protects against AI manipulation. 35:00 - Stewart's Claude regression problem drives chatbot vibe coding quest; small language models and domain-specific approaches explored as alternatives. 40:00 - Consciousness discussion through Michael Pollan's panpsychism lens; language versus cognition disconnect revealing LLMs as pure token-stitching without genuine thought. 45:00 - Context graphs as purpose-built knowledge graphs for AI; Stewart's planning agents versus coding agents architecture and ground truth verification problem. 50:00 - Docs-as-code versus code-as-docs paradigm shift; knowledge graphs as universal verifiers against validated facts; RDF 1.2 enabling provenance and degrees of certainty. 55:00 - Jessica Talisman's Knowledge Graph Academy recommended for onboarding; kgi.fm podcast shared; knowledge representation community needs better abstraction for wider adoption. Key Insights 1. The term "artificial intelligence" was not a marketing gimmick but was coined deliberately at the 1956 Dartmouth Conference to distinguish the work of John McCarthy from Norbert Wiener's cybernetics. The two camps represented genuinely different approaches, and the AI label was a form of intentional intellectual branding rather than empty promotion. 2. The semantic web, often called the most successful failure in technology history, has quietly embedded itself everywhere despite never achieving its original vision. Technologies like RDF power metadata standards inside every Adobe product and form the invisible backbone of government systems, enterprise data infrastructure, and cultural heritage organizations worldwide. 3. Knowledge graphs are best understood as an ontology combined with all the instances that populate it. The distinction between things and strings, popularized by Google in 2012, captures the core idea that knowledge representation is about concepts as distinct from the labels we give them. 4. The t-box, a-box, and c-box framework offers a practical model for understanding knowledge architecture. The t-box holds terminology and concepts, the a-box holds assertions about specific instances, and the c-box manages the attributes, taxonomies, and controlled vocabularies that sit between them and enable things like faceted search. 5. Large language models produce fluent, convincing output but lack genuine reasoning, epistemological grounding, or judgment. Research on cognitive surrender shows that even people who understand how LLMs work are still susceptible to being misled by their fluency, meaning intelligence and awareness offer no reliable protection against being deceived. 6. The gap between language and cognition matters deeply when evaluating AI. Evidence from people with aphasia shows that thinking can occur without language, which suggests LLMs, being purely language-based systems, are missing a fundamental layer of cognition that cannot be recovered through more tokens or better training. 7. Knowledge graphs and RDF-based representation are well suited to the problem of verification and grounding in AI systems. Rather than relying on vectorized embeddings of language, a knowledge graph can store validated, provenance-tracked facts with degrees of certainty, making it a natural foundation for building trustworthy AI applications.

Episode #553: The Connection Economy: What Recruiting Teaches Us About Human Value

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with client strategist Amadeus Huff to cover a wide range of topics that wind their way from the nuts and bolts of recruiting and payment models to the rapidly shifting landscape of AI adoption in business. The two dig into how AI tools are reshaping client success roles, the murky territory of recording laws and privacy in a globalized world, the geopolitical implications of oil supply chains, sanctions, and the rise of domestic tech ecosystems in countries like Russia and Argentina, and what all of this means for the future of human connection and the nation-state. Amadeus closes on an optimistic note, arguing that as AI takes over bureaucratic busywork and erodes trust online, people will increasingly hunger for genuine human relationships and third spaces. You can connect with Amadeus Huff on LinkedIn [https://www.linkedin.com/in/amadeus-huff/]. Timestamps 00:00 - Stewart introduces Amadeus Huff, diving into recruiting as building connections between job seekers and employers with minimal variance. 05:00 - Amadeus discusses AI adoption pitfalls, comparing aggressive growth strategies to Amazon's early model, questioning whether tools deliver promised results. 10:00 - Conversation shifts to AI notetaking versus human perception, exploring probabilistic interpretation differences between humans and machines. 15:00 - Recording consent laws debated across states, touching on Waymo surveillance, Uber data collection, and public versus private space definitions. 20:00 - Global privacy landscape examined, covering Swiss banking secrecy erosion, ProtonMail's departure, and RISC-V semiconductor development escaping US jurisdiction. 25:00 - Sanctions creating domestic innovation ecosystems discussed through Russia's example, paralleling Argentina's emerging commerce evolution. 29:00 - Closing reflections on AI replacing bureaucracy while preserving human purpose, optimism about meaningful work and deeper personal connections emerging. Key Insights 1. Recruiting is fundamentally about reducing variance between what job seekers want and what employers offer. The most ethical payment models in recruiting are tied to proven success, such as waiting three months to confirm a hire is working out, rather than collecting fees the moment a contract is signed. 2. Business thinking has shifted from shareholder value to stakeholder value, meaning companies now consider the wellbeing of employees, families, and communities, not just stock price. This shift is accelerating due to AI overpromising and underdelivering, making value-based measurement more important. 3. AI is most useful when it handles administrative tasks that provide no direct value to customers, such as transcribing meetings and populating CRM systems. This frees up workers to focus on meaningful relationship-building and intellectual work rather than bureaucratic busywork. 4. There is an important distinction between recorded and unrecorded conversation in professional settings. Building trust through informal off-the-record dialogue before switching on a transcription tool creates clearer boundaries and stronger relationships with clients. 5. Sanctions tend to follow a bell curve of effectiveness. Over time they force sanctioned countries to build domestic alternatives, which gain adoption and loyalty, ultimately reducing the influence of the original foreign companies once sanctions lift. 6. AI is degrading trust in online information to the point where people will increasingly crave authentic human connection, physical gathering spaces, live experiences, and real relationships rather than algorithmically generated content. 7. AI is quietly improving intergenerational relationships by removing codependency. When elderly parents learn to use AI for technical help, their calls to family members shift from problem-solving to genuine connection, which strengthens the relationship.

Episode #552: The Unbanked Advantage: How Nigeria's Financial Chaos Made It Crypto-Ready

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with software engineer and entrepreneur Arowolo Muritadhor for a wide-ranging conversation that moves from agriculture and manufacturing in Nigeria to the evolving role of crypto in the country's economy. They touch on how hyperinflation, particularly the naira's dramatic drop in 2023, pushed Nigerians toward stablecoins as a practical savings tool, and how informal kiosk networks have stepped in where traditional banking infrastructure falls short. The conversation also covers the tension between government regulation and the permissionless nature of blockchain technology, comparisons between the decline of the Roman Empire and current shifts in US economic dominance, the role of mobile payments in Africa, language learning, and whether AI agents have any real utility in crypto infrastructure yet. You can connect with Arowolo on LinkedIn [https://www.linkedin.com/in/arowolomuritadhor/] and X at @armolas_06 [https://x.com/armolas_06]. Timestamps 00:00 - Host welcomes Arowolo Muritadhor, introducing topics of software engineering and animal food production in Nigeria. 05:00 - Discussion shifts to manufacturing, components assembly, and China's dominance in low-cost production globally. 10:00 - Conversation explores crypto adoption in Nigeria as a network state phenomenon, separating informed users from mainstream population. 15:00 - Mobile payments and kiosk ATM replacements emerge as critical financial infrastructure bridging unbanked Nigerians. 20:00 - Roman Empire parallels drawn to modern crypto taxation, government control, and inevitable death-and-taxes reality. 25:00 - Bitcoin and Ethereum permissionless nature debated against government wallet-level censorship vulnerabilities. 30:00 - AI agents examined as crypto infrastructure tools, revealing mostly trading bots rather than foundational builders. 35:00 - Nigeria's 2023 naira collapse compared to Argentina's hyperinflation, driving citizens toward stablecoin dollar savings. 40:00 - US Treasury history unpacked through FDR gold confiscation and Nixon ending convertibility, paralleling empire decline. 45:00 - Crypto reframed as anti-bank rather than purely anti-government, enabling freedom through immutable accountability. 50:00 - Transparent blockchain ledgers discussed as potential government accountability tools across democracy, republic, and oligarchy structures. Key Insights 1. Nigeria has a significant divide between its northern and southern regions in terms of economic activity. The north, centered around Abuja, is more agricultural with substantial cattle production, while Lagos in the south functions as a dense urban and commercial hub. This geographic and economic split shapes how different financial tools and technologies are adopted across the country. 2. China's dominance in low-cost manufacturing has made it nearly impossible for countries like Nigeria, the United States, or Argentina to compete on price alone. The more realistic path for developing economies is to import components and focus on local assembly and creativity, which is where meaningful economic participation becomes possible. 3. Crypto adoption in Nigeria accelerated dramatically around 2023 when the naira experienced a sharp devaluation against the US dollar. Before that point, saving in dollars was difficult for many Nigerians, especially those without formal bank accounts, making stablecoins like USDT an attractive and practical alternative for preserving wealth. 4. Informal kiosk operators in Nigeria have organically become a substitute for ATMs, giving communities access to basic financial services where traditional banking infrastructure does not reach. This grassroots financial layer is now a key entry point for integrating crypto and stablecoin payments into everyday commerce. 5. Governments are increasingly trying to regulate crypto at the wallet and centralized exchange level, using tax compliance as a primary mechanism. While Bitcoin and Ethereum remain largely permissionless, the practical chokepoints for most users remain centralized platforms where identity and transactions can be monitored. 6. The historical parallel between the fall of the Roman Empire and current shifts in US economic and geopolitical power offers a useful frame for understanding why crypto matters. Just as Rome debased its currency and struggled to sustain imperial costs, the US faces mounting debt and a financialized economy that may accelerate dollar instability and push more people toward alternative stores of value. 7. One genuinely constructive use case for blockchain beyond speculation is immutable accountability, particularly for public institutions and prediction markets. A transparent ledger that governments or officials voluntarily adopt could create verifiable records of decisions and promises, reducing corruption and increasing trust in ways that traditional governance structures have struggled to achieve.

Episode #551: From Trash to Tools: The Open Hardware Revolution Powering Solarpunk Science

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop interviews Joshua Pearce, the John Thompson Chair in Innovation at the Department of Electrical and Computer Engineering and Ivey Business School at Western University, about the revolution in open source hardware for scientific research. They discuss how three-dimensional printing, Arduino controllers, and open source designs are dramatically reducing research costs—often by 85-95%—while democratizing access to lab equipment worldwide. Pearce shares stories from his 2013 book "Open Source Lab" and explains how the movement has exploded since then, covering everything from filter wheel changers and ball mills to metal three-dimensional printers and battery research equipment. The conversation explores recycle bots that turn plastic waste into filament, the role of AI in accelerating hardware development, and how open source licensing creates a global knowledge management system where improvements are shared across the scientific community. For those interested in learning more, Pearce recommends checking out the journal HardwareX [https://www.hardware-x.com/], repositories like Thingiverse [https://www.thingiverse.com/] and My Mini Factory [https://www.myminifactory.com/], and appropedia.org [https://www.appropedia.org/Welcome_to_Appropedia] for open source scientific tools and appropriate technology designs. Timestamps 00:00 Welcome and introduction to Joshua Pearce, discussing his work on open source lab equipment and the evolution since publishing his book in 2013 05:00 Early development of open source hardware including the breakthrough filter wheel changer project built by a high school student that saved thousands of dollars 10:00 Discussion of how Arduino and RepRap three-d printers enabled the democratization of scientific tools, making complex equipment accessible to anyone 15:00 Economic impact showing average tool savings of 85 percent, with Arduino and three-d printing combinations reaching mid-90s percent cost reduction 20:00 Case study of PhD student Mariam building complete battery research tool chain from scratch using open source designs and three-d printed components 25:00 Recycle bots enabling transformation of waste plastic into three-d printer filament for pennies, revolutionizing material costs and sustainability 30:00 Collaboration between universities and open source companies creating fluid handlers and acquisition systems, accelerating research capabilities globally 35:00 Large language models assisting code translation and research planning, though hallucinations require careful verification and domain expertise 40:00 Importance of fundamental knowledge when using AI tools, comparing vibe coding acceleration with necessity for understanding underlying principles 45:00 Testing standards and calibration methods for open source equipment, balancing precision requirements against cost-effectiveness for specific applications 50:00 Metal and ceramic three-d printing developments including MIG welding techniques and sintering processes for creating functional parts 55:00 Knowledge management through open source licenses, repositories like Thingiverse and Apropedia enabling global collaboration and continuous improvement Key Insights 1. Open source hardware has evolved dramatically since Joshua Pearce wrote his book in 2012-2013, to the point where he can no longer keep up with all the developments in the field. What started as a collection where every single example could fit in one book has exploded into an entire ecosystem with dedicated journals and thousands of researchers contributing. The vision was that scientific papers would eventually include hyperlinks to equipment designs that anyone could download and replicate, and that future is largely here today. There are now so many open source hardware articles being published that no single person can read them all, which represents a massive success for the movement. 2. The fundamental breakthrough enabling open source scientific hardware came from combining several key technologies, particularly the RepRap three-d printer project and Arduino microcontrollers. Pearce's introduction to the field came when he needed a sixty-five dollar plastic part for a solar laptop project and discovered Adrian's open-sourced rapid prototyper that could make its own parts. This led to building equipment like a filter wheel changer for testing solar panels with a high school student in about a week, replacing a device that would have cost two thousand five hundred dollars with five months lead time. The democratization of tools like three-d printing and Arduino, combined with extensive code libraries and shared designs, means that even high school students can now create sophisticated scientific equipment. 3. Open source scientific hardware delivers massive economic benefits, with the average tool saving scientists around eighty-five percent compared to commercial equipment, and savings reaching the mid-nineties when using Arduino and three-d printing. The economics are so compelling that the tax paid on a normal scientific tool can cover the cost of an open source alternative. A thousand dollar three-d printer can manufacture scientific tools worth more than a thousand dollars in a single Saturday. This dramatic cost reduction makes sophisticated research accessible to laboratories around the world regardless of their funding levels, fundamentally democratizing scientific capability. 4. The knowledge management approach enabled by open source licenses creates a powerful collaborative improvement cycle where thousands of people worldwide contribute to evolving designs. When researchers publish equipment designs with strong reciprocal licenses, anyone can use, modify, or even sell the designs, but improvements must be shared back with the community. This creates a dispersed international engineering effort where equipment continuously improves through contributions from researchers across different institutions and countries. The RepRap three-d printer exemplifies this process, starting as barely functional prototypes but evolving through community contributions to surpass commercial alternatives in speed, resolution, and material capabilities. 5. The integration of large language models and AI tools has significantly accelerated open source hardware development, though with important caveats about their limitations. LLMs excel at translating code between languages, suggesting experimental approaches, and helping researchers navigate unfamiliar fields by quickly synthesizing information from scientific literature. However, they suffer from hallucination problems and cannot be trusted for writing scientific articles or conducting complete literature reviews without verification. The key to effective use is having enough foundational knowledge to ask the right questions and verify outputs, using AI as a powerful acceleration tool rather than a replacement for expertise. 6. Material science capabilities in open source hardware have expanded far beyond plastic three-d printing to include metals, ceramics, semiconductors, and composites through innovative adaptations of basic equipment. Pearce's lab has developed methods for metal three-d printing using modified MIG welding for as little as twelve hundred dollars, created slot-die coating systems for seventeen nanometer semiconductor layers using converted three-d printers, and developed techniques for ceramic printing through various material mixing approaches. The recycle bot technology enables converting waste plastic into high-quality filament for twenty-five cents instead of twenty-five dollars per roll, dramatically reducing material costs while enabling circular manufacturing practices. 7. The infrastructure for sharing and discovering open source hardware designs has matured into a robust ecosystem spanning academic journals, commercial repositories, and specialized communities. Hardware X and the Journal of Open Hardware publish peer-reviewed designs alongside traditional scientific journals increasingly incorporating open hardware sections. Repositories like Thingiverse recently returned to hardcore open source principles after ownership changes and contains millions of designs, while Appropedia serves as a wiki for appropriate technology with thousands of open source designs. The GOSH community hosts annual conferences bringing together university researchers, companies, and independent hardware hackers, while field-specific communities have formed around technologies like the OpenFlexure microscope, creating networks where knowledge accumulates and never gets lost.

Episode #550: From Armies to Algorithms: Why the Biggest Player No Longer Wins

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with returning guest Ekue Kpodar for their third conversation together, covering a wide range of topics at the intersection of technology, geopolitics, and the evolving information age. They dig into Ekue's unconventional setup of running local AI models across roughly 15 computers, the growing case for open source models over closed ones from companies like OpenAI and Anthropic, and how Chinese open source models may be positioned to outcompete Western alternatives on a global scale. The conversation also touches on vibe coding and the democratization of software development, the strategic use of small models for IoT and enterprise applications, the role of Israel and China as dominant players in the information age, and how smaller nations and even individuals may wield outsized power as AI continues to collapse the cost of knowledge work. You can find Ekue Kpodar on X @ekpodar [https://x.com/ekpodar] and LinkedIn [https://www.linkedin.com/in/ekue-kpodar/overlay/photo/]. Timestamps 00:00 Stewart welcomes Ekue for their third episode, diving into vibe coding and AI-driven development changes. 05:00 Ekue explains using Claude on Chrome to auto-reply on Skool, burning tokens through screenshots, and Playwright as a more efficient alternative. 10:00 Stewart describes his Claude-dependent planning and coding agent system breaking after a model update, prompting him to build his own chatbot. 15:00 Small models discussed as critical for IoT, defense, and privacy-focused enterprises building internal APIs instead of routing traffic to OpenAI. 20:00 Open source versus closed source debated, with Chinese models gaining global traction while US foundational labs remain expensive and restrictive. 25:00 SaaS apocalypse explored as AI commoditizes knowledge work, with Linux and Terraform cited as proof open source still generates wealth. 30:00 OpenAI's sci-fi terminator fears explained as the reason they stayed closed source, ultimately handing China a strategic open source advantage. 35:00 China's economic dumping strategy applied to AI, potentially displacing US model dominance globally the same way manufacturing was disrupted. 40:00 Israel's signals intelligence dominance discussed alongside asymmetric warfare, drones defeating tanks, and information control replacing military muscle. 45:00 Global information age rankings debated, Israel leading, US and China tied, France and Poland emerging as sovereign tech players. 50:00 Qatar, NVIDIA, and Iran cited as proof that rare resources and technology matter more than population size in the 21st century power landscape. Key Insights 1. Running local AI models on a network of affordable computers can be more cost-effective than relying entirely on third-party APIs. By using compressed or smaller open source models locally, developers can handle repetitive or lower-stakes tasks without burning through expensive tokens from providers like Anthropic or OpenAI. 2. Small AI models are becoming increasingly important for IoT, defense applications, and companies that do not want to send sensitive data to external providers. Organizations can download open source models, run them on internal servers, and build proprietary APIs around them, creating something like an intranet of specialized small models. 3. The value created by AI tools is being redistributed away from traditional SaaS companies toward foundational model providers and individual builders. People are canceling subscriptions to software they once paid hundreds per month for, because AI now allows a single person to build comparable tools themselves. 4. Open source technology does not eliminate the ability to profit. Linux and Terraform are both open source yet made their creators wealthy. People will still pay for installation, setup, troubleshooting, and customization even when the underlying software is free. 5. China is applying its longstanding manufacturing dumping strategy to artificial intelligence by releasing cheap open source models globally, which threatens to erode US dominance in AI the same way Chinese manufacturing undercut other countries for decades. 6. In the information age, the size of a country or institution matters far less than its access to rare resources or advanced technology. Qatar, Israel, and NVIDIA each demonstrate that small populations or headcounts can wield enormous global negotiating power through concentrated technological or resource advantages. 7. Asymmetric warfare is redefining military power, with inexpensive drones defeating tanks that cost millions to build. This shifts the advantage toward nations that excel at signals intelligence and information management rather than those with the largest conventional military forces.