Google has recently rebranded its AI chatbot Bard as Gemini and made it available as an app for Android users and as a feature in the Google app for iOS users. Gemini is a generative chatbot that can help users with various tasks, such as preparing for interviews, debugging code, brainstorming ideas, or creating images.
However, users should be aware of the privacy implications of using Gemini. According to a support document by Google (via TechCrunch), some of the conversations and prompts that users have with Gemini are reviewed by human annotators, who do not have access to users’ personal information. The purpose of this review is to enhance the quality and performance of the service, Google says. Users can delete their conversations and prompts from the Gemini Apps Activity screen (more on that later).
Users also have the option to turn off Gemini Apps Activity, which stops Google from collecting and storing their conversations. However, even if this option is turned off, Google still keeps the conversations for up to 72 hours to “maintain the safety and security of Gemini apps and improve Gemini apps.”
Google also warns users not to share any confidential or sensitive information or data with Gemini, as it may be seen by reviewers or used by Google to improve its products, services, and machine learning technologies. Google says it stores the conversations and prompts for up to three years by default, but users can adjust the settings to retain the data for 3 or 36 months.
To turn off or delete Gemini Apps Activity, users can visit myactivity.google.com on their mobile or desktop browser. They can also delete their previous queries made with Bard and Gemini and toggle off Gemini Apps Activity.
Google explains that human review is a necessary part of the model improvement process, as it helps to refine and optimize the generative machine learning models that power Gemini. Google also says that it uses the conversations, feedback, and related data from Gemini Apps users to improve Google products so they can be safer, more helpful, and work better for all users.
Google has recently launched Gemini Advanced, a paid version of Gemini that gives users access to Ultra 1.0, Google’s largest and most capable AI model. Ultra 1.0 is the first model to outperform human experts on MMLU (massive multitask language understanding), a test that measures knowledge and problem-solving abilities across 57 subjects, such as math, physics, history, law, medicine, and ethics. Gemini Advanced with Ultra 1.0 can handle more complex tasks, such as coding, logical reasoning, following nuanced instructions, and collaborating on creative projects. Gemini Advanced is available in more than 150 countries and territories in English and will be expanded to more languages over time.