The Next Big Idea Daily Cade Metz and Kevin Roose on the Rise of AI
31 snips
Feb 27, 2026 
Kevin Roose, NYT columnist and author of Futureproof, offers rules for thriving alongside AI. Cade Metz, NYT tech reporter and author of Genius Makers, traces neural networks and their rise. They discuss neural networks’ 2012 revival, real-world uses like vision and language, risks like bias and disinformation, debates over AGI timelines, and what work humans should keep to stay valuable.
AI Snips
Chapters
Books
Transcript
Episode notes
Neural Networks Sparked Modern AI
- Neural networks drove the recent AI surge by learning tasks from vast data instead of hand-coded rules.
- Jeff Hinton's 2012 breakthrough in image recognition showed neural nets could outperform previous methods and attracted big tech investment.
AI Learned By Experience Not Rules
- Engineers shifted from coding behavior to training systems through massive experience on data.
- GPT-3 trained on internet text learned language patterns and could be repurposed for diverse tasks like writing and programming.
AI Mirrors Human Biases In Training Data
- Neural nets absorb human flaws because they learn from internet data, producing bias and hate speech.
- Retraining from scratch can't fully remove harmful content due to the enormous scale of training corpora.
