Show notes
Artificial Intelligence is getting smarter every month. Models can pass exams, write code, summarize documents, and even outperform humans in specific tasks. Yet according to Moritz Sudhof, one of the biggest risks in AI today has very little to do with intelligence.
Moritz is the co-founder of BigSpin.ai and a former VP of AI at BetterUp, where he helped build AI-powered coaching systems. His research focuses on a surprising problem: most AI failures are not obvious. In fact, BigSpin's research found that 79% of AI failures are invisible to users. The AI appears helpful, sounds confident, and produces convincing outputs, but users often walk away with incorrect assumptions, incomplete information, or entirely wrong conclusions without realizing it.
In this episode, we explore why AI hallucinations are only part of the problem. Moritz explains why the real challenge lies in the interaction between humans and AI. He shares how conversational failures emerge, why expert AI users actually encounter more failures than beginners, and why trust may become the defining challenge of the AI era.
📧💌📧
Tune in to get my thoughts and all episodes, don't forget to subscribe to our Newsletter: beginnersguide.nl
📧💌📧
We also discuss the seven hidden failure patterns that appear repeatedly across AI systems, including the Confidence Trap, Death Spiral, Silent Walk Away, and other interaction failures that impact AI agents, copilots, and enterprise AI deployments.
Towards the end of the conversation, we explore a fascinating question: what is the real long-term risk of AI? Moritz argues that the biggest danger may not be superintelligent machines taking over the world, but humans gradually outsourcing their judgment and decision-making to systems they trust too much.
In this episode, you'll learn:
• Why 79% of AI failures go unnoticed
• The difference between AI intelligence and AI trust
• Why hallucinations are often caused by interaction failures
• How AI agents create new risks for businesses
• The seven most common invisible AI failure modes
• Why expert users encounter more AI failures
• The role of human-in-the-loop systems
• How enterprises can improve AI reliability
• Why observability matters more than perfection
• The future of trust, verification, and AI governance
If you're building AI products, deploying AI agents, or simply trying to understand where AI is heading, this conversation provides a practical framework for thinking about AI reliability, AI trust, and the future of human-AI collaboration.
Chapters
00:00 Why AI Failures Matter
08:00 Why Hallucinations Really Happen
12:25 The 7 Invisible AI Failure Modes
19:30 Why AI Literacy Beats Better Prompting
25:25 Human-in-the-Loop and AI Trust
39:50 Claude Code, Agentic AI and Trust Problems
46:00 The Real AI Risk: Dependence vs Judgment
Top Three Quotes
• "79% of failures in AI conversations are invisible."
• "The real thing AI is shipping is not a model. It's an interaction."
• "The negative future is people abdicating their own judgment."
🌐 Where to Find Moritz Sudhof
🔹 BigSpin AI
https://bigspin.ai
Learn more about BigSpin's research on AI reliability, invisible failures, and human-AI interaction.
🔹 Personal Website
https://msudhof.com
Moritz shares his latest writing, research, and publications on AI, language, and human-centered technology.
🔹 LinkedIn
https://linkedin.com/in/sudhof
About Dietmar Fischer:
Dietmar is a podcaster and AI marketer from Berlin. If you want to know how to get your AI or your digital marketing going, just contact him at argoberlin.com
Hosted on Acast. See acast.com/privacy for more information.
Our TOPPODCAST Picks 
Stay Connected
