What AI Won’t Tell You When It’s Wrong | Jonathan Schaeffer

AI can sound confident, friendly, and brilliant, and still fabricate an answer.In this episode of An Hour of Innovation podcast, Vit Lyoshin speaks with Jonathan Schaeffer, founder and CEO of Synsira and creator of Kind, about AI trust, hallucinations, privacy, local AI, and why today’s most impressive tools should still be treated as tools rather than intelligent authorities. Jonathan explains why large language models do not truly understand the questions they answer, why “hallucination” is a misleadingly soft word for fabrication and error, how users can reduce risk by grounding answers in curated data, and why private files may be safer when AI runs locally on the user’s own computer.You’ll Learn* Why calling AI tools “intelligent” can create dangerous misunderstandings* Why hallucinations are better understood as fabrications, errors, or lies* How ChatGPT changed public expectations after November 30, 2022* Why students and everyday users can overtrust confident AI answers* What Jonathan’s incorrect AI-generated biography revealed about band-aid fixes* How curated datasets and source citations can make AI answers easier to verify* Why local AI can analyze personal files without moving them to cloud AI providers* What Kind does for documents, presentations, images, videos, and personal knowledge collections* Why cloud AI creates privacy risk for books, intellectual property, company strategy, customer documents, and personal data* Why 95 or 99 percent accuracy may still be unacceptable when users cannot tell which answers are wrong* Why Jonathan prefers “augmented intelligence” over artificial intelligence as a product philosophy* What basic AI education every user should have before trusting these toolsJonathan Schaeffer is an AI researcher, founder and CEO of Synsira, and creator of Kind. He spent 40 years at the University of Alberta, where he worked in artificial intelligence, research, teaching, and public speaking before leaving the university to build a local AI product focused on privacy, verifiable answers, and user-controlled data. His background also includes pioneering work in game-playing AI, including Chinook and computer checkers, and recognition as an AAAI Fellow and Royal Society of Canada Fellow.Timestamps00:00 Introduction01:18 Understanding AI Misconceptions04:55 The Trust Dilemma in AI13:19 Hallucinations vs Accuracy in AI20:20 Introducing Kind: A Local AI Solutions29:39 Curating AI Models for Optimal Performance34:14 Understanding AI’s Practical Applications37:20 AI as Augmented Intelligence42:08 The Limitations of AI and Public Perception48:06 Machine vs Human Learnings52:58 Educating Users on AI Interactions58:00 Innovation Q&AConnect with Jonathan* Website: https://kind.synsira.com/ * LinkedIn: https://www.linkedin.com/in/jonathan-schaeffer-phd-frsc-aaai-fellow-3318015/ * Other: https://en.wikipedia.org/wiki/Jonathan_Schaeffer Connect with Vit* Substack: https://anhourofinnovation.substack.com/* LinkedIn: https://www.linkedin.com/in/vit-lyoshin/ * X: https://x.com/vitlyoshin * Website: https://vitlyoshin.com/ Support the PodcastIf you enjoy the podcast, you can support it by exploring the tools below.These are affiliate links, meaning the show earns a small commission at no extra cost to you.MeetGeek: Record, transcribe, summarize, and share insights from every meeting. https://get.meetgeek.ai/yjteozr4m6ln Amplemarket: Step into the future of sales: Human + AI. Empower reps, uncover opportunities, and grow revenue.https://grow.amplemarket.com/04e068165684-affiliateDatabox: Turn business performance data into clear answers your team can understand, explain, and act on – instantly.https://join.databox.com/vhw0wbr9qg9o For inquiries about sponsoring An Hour of Innovation, email iris@anhourofinnovation.com