Google has added a memory feature to its Gemini chatbot, enabling it to recall details about users' preferences, work, and personal interests.
This functionality, similar to ChatGPT's memory feature, allows Gemini to enhance conversations by using stored information for context. For example, if a user asks the chatbot to remember their favorite foods, it can later use this data to provide personalized restaurant recommendations.
Currently, the memory feature is available only to subscribers of Google’s $20-per-month Google One AI Premium plan and is accessible through the web version of Gemini. It has not yet been introduced to Gemini's iOS and Android apps.
The Gemini interface includes examples of how users might utilize memory, such as asking the chatbot to use simple language, focus on JavaScript when writing code, or include cost estimates during trip planning. At present, the feature supports English-language prompts and can be turned off at any time. Memories are stored until manually deleted by the user.
Google has clarified that user data saved in Gemini's memory is not shared or used to train the AI model. This ensures privacy, though security remains a consideration. For instance, earlier this year, a researcher demonstrated how malicious actors could plant false memories in similar AI tools, potentially compromising user data.
This update demonstrates Google's efforts to make its generative AI tools more personalized and practical, while balancing privacy and security concerns.
