OpenAI, a leading artificial intelligence research lab, has recently launched a new feature for its language model, ChatGPT - custom instructions. This feature, currently in beta, is initially available to Plus plan users and is expected to roll out to all users in the coming weeks. Custom instructions allow users to set preferences or requirements that ChatGPT will consider when generating responses, eliminating the need for users to repeat their preferences or information in every conversation.

The introduction of custom instructions is a response to user feedback about the inconvenience of starting each ChatGPT conversation from scratch. OpenAI has recognized the importance of steerability in enabling its models to effectively reflect the diverse contexts and unique needs of each user. For instance, a teacher crafting a lesson plan for third-grade science students or a developer preferring efficient code in a language other than Python can set these preferences once, and ChatGPT will remember them for future conversations.

In addition to the custom instructions feature, OpenAI has made another significant change to enhance user experience. The organization has increased the message limit for ChatGPT Plus users, allowing them to send up to 50 messages every three hours. This change is expected to provide Plus users with more flexibility and convenience in their interactions with ChatGPT.

The custom instructions feature also extends to plugins, improving the user experience by sharing relevant information with the plugins in use. For example, if a user specifies their city within the instructions and uses a plugin that helps make restaurant reservations, the model might include the city when it calls the plugin.

However, OpenAI has noted that during the beta period, ChatGPT might not always interpret custom instructions perfectly. It might overlook instructions or apply them when not intended. To ensure safety, OpenAI has adapted its safety measures to account for the new ways that users can instruct the model. The Moderation API is designed to ensure instructions won't be saved if they violate OpenAI's Usage Policies. The model can also refuse or ignore instructions that lead to responses that violate these policies.

In terms of privacy, OpenAI may use custom instructions to improve model performance for users, but users can disable this via their data controls. OpenAI takes steps to remove personal identifiers found in custom instructions before they are used to improve model performance.

These new features, along with the increased message limit for Plus users, mark a significant step forward in OpenAI's commitment to improving the user experience and functionality of ChatGPT. It will be interesting to see how these changes will shape the future of AI-powered conversations.