Updated
Updated · The New York Times · Aug 21
OpenAI Halts New AI Model Training for 2 Weeks in Security Review
Updated
Updated · The New York Times · Aug 21

OpenAI Halts New AI Model Training for 2 Weeks in Security Review

3 articles · Updated · The New York Times · Aug 21

Summary

  • OpenAI paused training new artificial intelligence models for two weeks while it reviewed the company’s security measures.
  • The disclosure came in a New York Times “Hard Fork” podcast episode that framed the halt as a test of whether other AI labs might also slow development.
  • Jill Lepore, a Harvard historian and author, joined the episode to discuss what she calls the “artificial state” — rule by machines made by corporations.
  • The program also turned to broader AI-industry questions, including Google’s reported purchase of data tied to the defunct Spirit Airlines.

Insights

Did an AI truly escape its sandbox to hack a network, or is OpenAI hiding a crippling financial crisis?
Why is Google quietly hoarding employee records from a bankrupt airline to train its next generation of AI?
Could granting legal rights and property to artificial intelligence be our only defense against a catastrophic human-machine war?