OpenAI, the pioneering artificial intelligence research organization, announced on Thursday a significant improvement to its ChatGPT platform. Users will soon be able to tailor the AI chatbot’s responses to their preferences without having to repeat those needs every time.
This new feature, called “custom instructions”, encourages users to more clearly define the parameters for ChatGPT’s responses. While currently exclusive to Plus subscribers, OpenAI CEO Sam Altman hinted in a tweet that a wider rollout is imminent.
The introduction of custom instructions is a step towards a more individualized AI experience, but it comes at a cost: users will have to choose between convenience and preserving their privacy.
Simplify your conversations
OpenAI has provided “thought starters” to guide users on optimizing this feature. These prompts include, for example, aspects such as the user’s location, professional background, hobbies, and aspirations. Once this data is entered, users can further refine parameters such as the desired formality level of the response, the way ChatGPT addresses the user, and the degree of subjectivity in the responses.
In its tweet, OpenAI shared some use cases from early adopters who have been trying out the new feature:
- Users can share their level of expertise in a specific area to avoid unnecessary explanation.
- Users interested in learning a new language or improving their existing knowledge can practice their conversations with grammar correction.
- Users looking for more legal context may receive responses that are governed by the specific laws of their country.
- Users can improve their novel writing by using “character sheets” to help ChatGPT maintain a consistent understanding of story characters in ongoing interactions.
- Users can instruct ChatGPT to perform code updates consistently in a unified format.
- Users who want to maintain a consistent voice can apply the same voice and writing style as in emails for all future email writing requests.
Once set up, ChatGPT will consistently take these guidelines into account in all subsequent interactions, eliminating the need for users to repeat their preferences in every conversation.
Convenience always trumps privacy
With these innovative features, there are still challenges regarding the potential for integrating sensitive information into ChatGPT for training purposes. Tech giants such as Apple and Samsung have already placed restrictions on the use of ChatGPT among their employees, citing privacy concerns and possible intellectual property infringements.
Additionally, an advisory memo instructed US House of Representatives staff to exercise caution when using ChatGPT. It recommended the exclusive use of the Plus version and emphasized the importance of not entering private data blocks into the chatbot.
In response to these concerns, OpenAI unveiled a series of privacy improvements to ChatGPT in April. These updates gave users the option to opt out of using their chat history for training OpenAI’s AI models, but that automatically disables a user from using the “custom instructions” feature.
With most of today’s consumer technology, there’s always the question of how much of your personal information you’re willing to give up just so you can use that new gadget (or smart home assistant).
As the field of artificial intelligence continues to evolve, the balance between customization and privacy remains an important point of discussion. With OpenAI’s latest update, the conversation about user-centric AI experiences and the associated challenges will increase.