Consider This from NPR Bubbling questions about the limits of the AI revolution
27 snips
Aug 24, 2025 
Cal Newport, a contributing writer for The New Yorker and a computer science professor at Georgetown University, discusses the AI landscape with a focus on current challenges. He highlights the alarming 95% failure rate of generative AI projects in businesses, prompting concerns over job security and the economy. Newport questions whether the anticipated benefits of AI surpass its costs and discusses the importance of innovation balanced with caution. Amidst skepticism, he also explores the potential for customizable AI technologies to enhance user experience.
AI Snips
Chapters
Transcript
Episode notes
Host Demoed ChatGPT On Air
- Scott Detrow asked ChatGPT to write a 30-second intro and read the AI-generated text on air.
- He then pivoted to a human guest to discuss whether AI has reached a ceiling.
Scale-Driven Gains Have Plateaued
- Progress from making models bigger has flattened since GPT-4 and gains are now more incremental.
- Cal Newport says major models shifted from pre-training scale to post-training tweaks around late 2023–early 2024.
Industry Focused On Post-Training Fixes
- All major large language model teams now emphasize post-training improvements over simply enlarging models.
- Newport frames this as focused, incremental work rather than transformative leaps.