OpenAI Built an AI to Attack Itself: GPT-Red Exposed Flaws Humans Missed - Tech Times

  • OpenAI developed GPT-Red to test its own AI models.
  • GPT-Red identified flaws that human testers missed.
  • The experiment highlights the importance of self-testing in AI development.
6 similar stories from other sources

OpenAI's First Device Will Reportedly Be A 'Humanlike' Rechargeable Speaker - Engadget

  • OpenAI plans to release a 'humanlike' rechargeable speaker.
  • The device is expected to be their first hardware product.
  • Details about the speaker's capabilities and features are yet to be revealed.
2 similar stories from other sources

Inside our approach to the Model Spec - OpenAI

  • OpenAI discusses its approach to Model Spec.
  • Focuses on the technical aspects of model development.
  • No specific product launch or update mentioned.

OpenAI Says GPT-5.6 Sol Is 54% More Token Efficient for Coding - Memeburn

  • OpenAI claims that GPT-5.6 Sol is 54% more efficient for coding.
  • The model's efficiency is highlighted in a new release.
  • The improvement is expected to enhance coding productivity.
1 similar story from other sources