OpenAI, the trailblazing AI research organization, has set a new milestone in the realm of chatbots with its latest beta feature for ChatGPT.
The game-changing addition, known as “Custom Instructions,” empowers users to tailor their interactions with the chatbot like never before. No longer bound by repetitive instructions, users can now craft a unique conversational experience that suits their individual preferences and needs. Although initially exclusive to ChatGPT Plus subscribers, OpenAI plans to roll out this feature to all users in the coming weeks, excluding the UK and EU regions for now.
With the Custom Instructions feature, users gain the power to dictate their preferred conversational approach to ChatGPT. This means that the chatbot can retain specific details and preferences across multiple interactions, eliminating the need to start afresh with every chat session.
Imagine a teacher preparing a lesson plan; instead of repeatedly stating they teach 3rd-grade science, they can input this information once, and ChatGPT will remember it for future reference. For developers who favor coding in a language other than Python, they need only specify it once, and the AI model will consider this preference in subsequent interactions. Even grocery shopping for a big family becomes more convenient, as ChatGPT takes into account the requirement for 6 servings in the grocery list.
To access this cutting-edge feature, users need to venture into the “Settings” menu and navigate to the “Beta Features” section, where they can activate “Custom Instructions.” For iOS users, a simple trip to “Settings,” then “New Features,” and toggling “Custom Instructions” on will do the trick. However, the feature cannot reach that specific account region if the option does not appear in the settings.
Once the Custom Instructions feature is activated, users are met with two text input fields. The first field collects essential contextual information about the user, including their profession, preferences, or any details that could enhance the model’s responses. For example, a user specializing in science education for third-grade students could input relevant information about their role.
The second input field focuses on the user’s desired response style or format. This allows users to shape the AI model’s output to suit their preferences, such as requesting information to be presented in a table format or providing pros and cons for better decision-making.
Initial testing showed that Custom Instructions significantly influenced the chatbot’s responses, much like providing the same information in a prompt. While the full impact on the overall context window, currently limited to 4,096 tokens, is yet to be clarified by OpenAI, the feature promises to be an exciting prompt extension.
OpenAI has also ensured that Custom Instructions work harmoniously with plugins, offering further personalization. For instance, by specifying the city they live in, a user utilizing a restaurant reservation plugin might find the AI model considering the city information while making a reservation. However, it is essential to note that custom instructions, like prompts, are shared with plugin providers.
OpenAI acknowledges that Custom Instructions may not be flawless during the beta testing phase. There might be instances where ChatGPT overlooks specific instructions or applies them in unintended contexts. OpenAI welcomes user feedback and collaboration during the beta period to refine and enhance this revolutionary feature.
The future of AI conversations has arrived, and ChatGPT is paving the way to a more personalized and dynamic chatbot experience!