The Naked Scientists Podcast Is AI changing the way we think?
Sep 2, 2025
In this engaging discussion, Mike Waldrich, a Professor of AI Foundations at Oxford University, dives into how artificial intelligence is reshaping our thought processes. He highlights the risks of misinformation from large language models and the rise of predatory publishing that compromises research integrity. The conversation also addresses alarming mental health issues linked to AI, like 'AI psychosis.' Waldrich emphasizes the need for balanced use of AI, urging caution in how we rely on technology for knowledge and decision-making.
AI Snips
Chapters
Transcript
Episode notes
How LLMs Learn From The Web
- Large language models train on enormous swathes of web text to predict the likeliest next words.
- This scale makes discrimination between high- and low-quality sources difficult and harms output quality when training data is poor.
AI Content Could Pollute Future Data
- AI-generated content is proliferating and risks becoming the dominant internet material.
- Training future models on AI-made content can degrade quality through iterative decline.
Why AI Hallucinations Seem Convincing
- LLMs 'fill gaps' when information is missing and do so in highly plausible ways.
- That plausibility makes their factual errors hard for readers to detect.