ChatGPT Gets Customization With Instruction Feature | | | | Turtles AI

ChatGPT Gets Customization With Instruction Feature
DukeRem
  #OpenAI is #launching a new #feature called #custom #instructions to #help #users #shape #ChatGPT's #responses according to their #preferences. Custom instructions allow users to specify one-time preferences or requirements that ChatGPT will consider across all future conversations. The feature aims to reduce the need for users to repeat context in every conversation. For example, a teacher can specify that they teach third grade science and ChatGPT will take that into account for subsequent lesson plan requests. OpenAI indicates that custom instructions will initially be available to Plus plan users in beta, expanding to all users in the coming weeks. The company has made adaptations to its safety and privacy measures to accommodate the new feature. During the beta period, OpenAI warns that ChatGPT won't always interpret custom instructions perfectly, with the potential for overlooking or wrongly applying them. The Moderation API and model itself may also refuse instructions that violate usage policies. OpenAI notes that custom instructions can be disabled via a users' data controls and that identifiers will be removed from instructions before using them to improve the model, as with standard ChatGPT conversations. Short article: ChatGPT Gets Custom Instructions in Beta Rollout OpenAI is launching a beta of custom instructions for ChatGPT, letting users specify a context that will apply to all future responses. The feature aims to reduce the need for repetitive information in conversations. In a 3rd-grade science example, specifying "I teach 3rd grade science" led ChatGPT to outline moon topics in a pros/cons table. The custom instructions beta is initially for Plus plan users and in the coming weeks will expand to all users. OpenAI notes the potential for issues interpreting instructions during the beta period and indicates safety measures have been adapted for the feature. Highlights:
  • Custom instructions let users specify context for all future ChatGPT responses
  • The aim is to reduce repetition of context in conversations
  • Initially for Plus plan users, then rolling out to all users
  • OpenAI warns of imperfect interpretation of instructions during beta period
  • Safety and privacy measures have been adapted for the new feature