Google's Gemini chatbot gets a memory, but only for paying AI customers

Google's AI chatbot Gemini has been given a reminder function. Data and user preferences can be stored in the memory of this artificial intelligence, so that answers appear more personal and are individually tailored. Google itself describes the AI chatbot's answers with saved reminders as "more helpful and relevant". However, this function also raises questions, for example about data protection.

More than a year ago, Google announced that it would be adding a reminder function for personalized answers to the AI chatbot, which was still called Bard at the time. Until now, however, Gemini has not been able to remember the content of conversations, only information such as language, device and location. This is now changing, so Google is now catching up with the competition. At the beginning of 2024, OpenAI introduced memory for ChatGPT and gave its AI chatbot a memory.

Gemini can now remember the user's dinner preferences, for example, so that responses to restaurant recommendations should be tailored to them weeks later. What the AI chatbot should remember can be explicitly specified, for example with phrases such as "Don't forget" or "I can only" or "Always mention". According to Google, the function can also be switched off and reminders can be deleted individually by the user.

Memory only in English and with Gemini Advanced

However, the reminder function is currently only available in English and only in the browser. The mobile Gemini apps for Android and iOS do not yet have a reminder function. In addition, only "Gemini Advanced" has this function, so a "Google One AI Premium subscription" is required. This currently costs 21.99 euros per month and offers more cloud storage space (2 TByte) in addition to the advanced AI chatbot. ChatGPT's memory function, on the other hand, has been available to all users free of charge since the beginning of September.

However, the question is also whether Google uses the stored memories of individual users to train its AI language models, for example. Google denied a corresponding request from TechCrunch, stating: "Your stored information will never be shared or used to train the model."

This is intended to ensure the user's privacy, as ChatGPT was already able to elicit training data almost a year ago using a word repetition trick. Nevertheless, there is criticism of the storage of sometimes sensitive user data by AI chatbots. This is a data protection disaster, OpenAI is really brazen with ChatGPT's memory function. This is the continuation of personalized advertising, but at least as bad, if not worse, in terms of data protection.

(fds)