Image Generation

Today's Stories

1. 1

   PCE Uses Multiple AI Agents to Catch Hallucinations

   A colleague has developed the Perseverance Composition Engine (PCE), which uses Artificial Organisations—multiple AI agents assigned specific roles—to work through tasks iteratively and catch problems such as confident false claims, hallucinations, or dangerous advice. The system assigns agents like a Composer and Corroborator, where the Corroborator verifies claims against source documents. Current large AI companies attempt to reduce hallucination through better training and instruction, but research suggests hallucination may be a fundamental mathematical inevitability in language model architecture. PCE takes a different approach by building organizational structure (separation of duties, independent checks, persistent knowledge bases) to contain and correct inevitable errors rather than eliminate them at the source.

   The core research code is available for daily use. The system addresses three failure modes—hallucination, context issues (where models lose information when context windows fill up), and memory issues (where AI forgets between conversations)—by using a persistent, indexed knowledge base (the Curator agent) and enforced role-based agents that operate from a quality prior of documents rather than guessing from scratch.

2. 2

   Insurers deploy AI to model catastrophes, but face hallucination risks

   Insurance companies are adopting diffusion models (a type of generative AI) to generate tens of thousands of plausible weather events in areas where historical data is sparse. The goal is to improve the accuracy of catastrophe risk assessments. More precise risk modeling could help insurers better price policies and manage exposure. However, researchers are warning that AI-generated synthetic data may contain hallucinations—fabricated or unreliable patterns—which could compromise the reliability of these assessments.

   The tension between the appeal of synthetic data generation and the technical challenge of validating its accuracy. Sales pressure to deploy the technology may outpace due diligence on whether the models are producing trustworthy results.

3. 3

   We're rebuilding financial services with AI

   We're rebuilding financial services with AI

4. 4

   Sanders proposes U.S. sovereign wealth fund for AI

   Senator Bernie Sanders has called for the U.S. government to establish a sovereign wealth fund dedicated to artificial intelligence, positioning it as a mechanism for public benefit from AI development. A sovereign wealth fund could shift how the U.S. captures economic value from AI—potentially redirecting gains toward public investment rather than concentrating them in private hands. This reflects growing debate over how democracies should govern AI's economic impact.

   The proposal remains at the advocacy stage; there is no indication of legislative timeline or political support needed to move it forward.

5. 5

   Companies Rush to Cut AI Token Spending as Costs Spiral

   Businesses are scrambling to reduce their spending on AI token usage—the computational units charged by AI providers—as costs have become unexpectedly large. The article reports that companies are actively working to cut back on how many tokens they consume when using AI services. AI spending has grown faster and larger than many companies anticipated, creating pressure on budgets and forcing a reckoning about how efficiently organizations use these tools. For businesses relying on AI APIs (interfaces that let software communicate with AI services), managing token consumption has become a new operational priority.

   The shift signals that the economics of AI deployment—how much it costs to run AI in production—are forcing companies to make hard choices about which AI features and use cases are actually worth the expense. This pressure may reshape which AI vendors and pricing models succeed in the market.

6. 6

   Singapore leads per-capita Claude usage globally

   Singapore ranks #1 in per-capita usage of Anthropic's Claude AI, according to usage data reported by Opentools. Per-capita rankings reveal where AI adoption is most concentrated relative to population, offering insight into regional demand patterns and market penetration for AI services outside the US.

   The metric highlights Singapore's position as a technology hub and suggests strong adoption of Claude among its population, though the body does not disclose absolute usage figures or timeframes.