Something You Should Know

What AI Is Really Good At & That Feeling You Get When You Don’t Fit In

35 snips
Apr 30, 2026
Claude Steele, Stanford social psychologist who studies identity and belonging, and Christopher Mims, tech columnist focused on practical AI uses, dive into what draws us in and what AI can actually do today. They explore processing fluency and attractiveness, where current AI shines and fails, personalization and tool integration, and small mindset shifts that ease social tension.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
ANECDOTE

Transcription Example Revealed AI Hallucination

  • Mims used Otter to transcribe interviews and saw the AI produce a confident but incorrect summary of its own transcript.
  • That example illustrated how hallucinations can be close to the truth yet confidently wrong, a persistent issue.
ADVICE

Stick With The AI That Knows You

  • Stick with the AI you already use because accumulated context and memory matter more than marginal capability differences.
  • Mims compares switching chatbots to switching platforms: high switching costs often outweigh small gains.
INSIGHT

Next Leap Needs New AI Architecture

  • A future breakthrough may require a new architecture beyond transformers to enable abstract reasoning and one‑shot learning.
  • Mims notes humans and octopuses learn from single examples, a capability current models lack and that would be transformative.
Get the Snipd Podcast app to discover more snips from this episode
Get the app