Gartner ThinkCast Understanding AI Hallucinations: Inside the Machine Mind
51 snips
Jul 29, 2025 Chris Howard, Gartner's Chief of Research, dives into the intriguing world of AI hallucinations. He explains what these hallucinations reveal about machine cognition and their implications for business decisions. The discussion highlights how multiagent systems can revolutionize healthcare diagnostics by enabling collaborative solutions. Howard also explores physics-informed neural networks, showcasing their potential to align AI predictions with real-world laws. Surprisingly, he suggests that hallucinations could actually offer innovative perspectives to problem-solving.
AI Snips
Chapters
Transcript
Episode notes
AI Made False Obituaries
- ChatGPT generated obituaries for living Gartner analysts when asked for biographies.
- This happened because many biographies it was trained on were of deceased people.
Why AI Hallucinates
- Hallucinations happen when AI predicts data to fill gaps where it lacks training.
- Narrowing the training data reduces hallucinations by constraining AI's knowledge base.
Multi-Agent Systems Mimic Experts
- Multi-agent AI systems mimic expert panels debating to refine decisions.
- Agents bring diverse expert views to collectively improve accuracy and reduce hallucinations.