OpenAI Updates the Rules That Guide How Its AI Behaves
OpenAI has updated the Model Spec, the rulebook that guides how its AI models should behave. The new version is built around five core values, including staying true to what humans intend and letting people correct the AI. Models should also use conversation history to work out what users really want and state their own assumptions.
Quick glossaryOpenAI, Model Spec, correctability
- OpenAI
- The San Francisco company behind ChatGPT, one of the most widely used AI chatbots in the world.
- Model Spec
- The written set of rules that tells an AI model how it should think, respond, and behave.
- correctability
- The ability for people to step in, adjust, or switch off an AI system when something goes wrong.
Burning Potato score: 78%