Why LLMs Can't Be Trusted With High-Stakes Decisions | Scott Cohen on GlobalEdgeTalk

Ask a large language model the same question twice, and you can get two different answers. That's not a prompt bug. That's what probability does. Jaxon.AI founder and CEO Scott Cohen sat down with Alex Romanovich on GlobalEdgeTalk to walk through why large language models are token predictors returning the highest-probability answer rather than the correct one, why retrieval-augmented generation and LLM-as-judge are probabilities stacked on probabilities, and where that breaks down in insurance, financial services, defense, and life sciences. The alternative Scott describes is symbolic reasoning: formal rules where a thing either is or it isn't, the way a geometry proof either holds at every step or fails. Jaxon turns policy requirements into executable DSAIL® (Domain-Specific AI Language) logic. Relevant claims are extracted from the input and evaluated against formal rules using symbolic reasoning, producing traceable outcomes that show which rules passed, failed, or could not be determined. Also in the conversation: the Air Force project that started all of this, what culinary school taught him about running a company, and the first hire he'd take back. CHAPTERS 0:00 Intro: Scott Cohen, founder and CEO of Jaxon.AI 1:57 The Air Force project, and what "deterministic" actually means 3:46 Sweetheart Cup, the flexi-straw, and a family of inventors 4:30 Culinary school, mise en place, and running a company 5:49 "AI for AI" and betting on provable answers 6:40 Token predictors: why AI output sounds right 8:12 Hallucinations, lawsuits, and why companies fly blind on the risk 9:43 RAG, LLM-as-judge, knowledge graphs: probabilities on probabilities 11:20 What defense work demands that most enterprises never ask for 13:02 The most expensive hire he ever made 14:22 What he'd tell a younger founder 15:22 Fatherhood, risk, and legacy 16:56 Integrity, tenacity, and Samisms Recorded on GlobalEdgeTalk, hosted by Alex Romanovich of GlobalEdgeMarkets. Full episode page and transcript: https://globaledgemarkets.com/globale... #jaxonai #trustworthyai #aigovernance