OpenAI has introduced a new feature called customised instructions for its ChatGPT model to enhance the quality of responses. Although the feature is currently in beta for Plus users, the company plans to roll it out to all users soon. Unfortunately, users in the UK and EU are not able to access this feature yet.
With customised instructions, users can edit or delete instructions as per their preference for new conversations. Notably, these instructions are not shared with shared link viewers, ensuring privacy and security. However, OpenAI mentions that if users utilize third-party plug-ins, these plug-in developers may receive relevant information from users’ instructions. To mitigate concerns, the company advises users to employ trustworthy plug-ins and refrain from sharing any information they do not want to reveal to third parties.
The information provided through customised instructions will also be used to enhance the model’s performance by teaching it how to adapt its responses without going overboard. Users accessing ChatGPT on iOS can find the feature under ChatGPT Account Settings > Custom Instructions. On the web, users can access it by clicking on their name and selecting Custom Instructions. By entering instructions into both fields and clicking Show tips, users can gain insights into what they should write. Finally, selecting Save saves the custom instructions.
OpenAI emphasizes that upon deleting their OpenAI accounts, any custom instructions associated with those accounts will be deleted within 30 days as part of the account termination process. This ensures the user’s preferences are respected and their data is not retained unnecessarily.
The previous month, OpenAI made updates to its ChatGPT application on iOS, offering Bing integration for Plus plan users. Additionally, the company improved the history search functionality with the update.
In conclusion, OpenAI’s introduction of customised instructions for ChatGPT aims to enhance user experience and make the model’s responses more tailored. Users can exert control over the instructions, delete them if necessary, and have the assurance that their information is not shared with shared link viewers. OpenAI suggests using trustworthy plug-ins and reassures users that their instructions will be utilized to improve model performance.
