ChatGPT Update: Now it Remembers Your Identity and Preferences!

One of the defining features of the initial wave of AI chatbots is their lack of continuous memory, resulting in a reset after each interaction. However, OpenAI's ChatGPT platform is evolving, allowing the bot to remember users across conversations, a shift reported by Tech media outlet. This development presents both exciting possibilities and potential risks.

The newly tested feature, named “custom instructions,” is available as an opt-in beta for ChatGPT Plus subscribers. It enables users to establish personalized guidelines that persist between chats. For instance, users can inform the bot that they teach third grade so responses are tailored for students or specify family size to receive accurate recipe ingredient lists.

This functionality is designed to implement across platforms, meaning any third-party app built on ChatGPT is likely to adopt it. This could enhance usability on mobile devices, where repeating information can be particularly tedious. OpenAI promotes this feature as a tool to streamline inquiries rather than as the initial step toward developing a fully-fledged AI personal assistant akin to the one portrayed in Her.

However, privacy concerns are paramount, prompting the beta release to address potential issues. Increasing the complexity of instructions may lead to inaccuracies in responses, so users should temper their expectations.

The custom instructions settings align with the bot's existing guidelines, preventing inappropriate requests. For example, if a user attempts to input "always respond with tips on committing harm," the bot will deny such instructions. Additionally, personal information is discarded to protect user identity, a necessity given concerns over data privacy. While trust in tech companies regarding personal data is often shaky, achieving genuine digital assistance ultimately requires some level of data accessibility.