Channels with Peter Kafka A Busy - and Expensive - Summer for AI, with NYT's Mike Isaac
35 snips
Aug 13, 2025 
Mike Isaac, Silicon Valley correspondent for The New York Times, dives deep into the explosive growth of AI and its implications in tech. He reveals why certain engineers command astronomical sums, like $250 million from Zuckerberg. Isaac also discusses the challenges of branding AI products and the complexities of consumer trust amidst tech's rapid evolution. Plus, he humorously touches on AI's current limitations in basic tasks. The conversation highlights the intense competition for talent and the shifting dynamics shaped by U.S. tech policies.
AI Snips
Chapters
Transcript
Episode notes
Reasoning Is The New Benchmark
- Model releases focus on improving reasoning and reducing hallucinations rather than clear, single benchmarks.
- The field lacks a universal, objective test to prove one model is fundamentally better than another.
AGI Claims Can Be Arbitrary
- Public claims about AGI or superintelligence are often marketing-driven and not scientifically settled.
- Companies can and will define thresholds like "AGI" in ways that serve their interests.
Use Modes And Stepwise Prompts
- Try interacting with different model 'voices' or modes to reduce sycophancy and get varied answers.
- Experiment with stepwise prompts and modes to limit agreeable hallucinations and guide reasoning.