Google has unveiled a new AI initiative called "Project Ellmann" that aims to transform users' stored memories into comprehensive narratives enriched with contextual information. The concept is to offer a "bird's eye" perspective into users' lives through this project, utilizing advanced language models like Gemini. As reported by CNBC, the AI model, drawing from the capabilities of models like Gemini, seeks to address previously unanswerable questions. It can gather insights from users' search results, recognize patterns in their photos, and incorporate a chatbot for interactive engagements. Project Ellmann is purported to extract information from biographies, past photographed events, and recent occurrences to craft a cohesive story.
During a presentation, a Google product manager mentioned that Project Ellmann could discern a child's date of birth, outline details about their parents, and identify if they are an only child. The "Ellmann Chat" aspect of the software envisions a scenario where a user engages with a ChatGPT-like interface that is already well-versed in their life details. Leveraging stored memories, the AI can respond to queries like "When did I last see my sibling?" or "Do I have a pet?" In the case of the latter, Google asserts that the AI can provide specific details, including the pet's name and the individuals they were most frequently photographed with. Despite the potential invasiveness, Google emphasizes its commitment to privacy, stating that this was an early internal exploration and any new features would prioritize user privacy and safety.
Project Ellmann is designed to comprehend users' eating habits, favorite foods, preferred websites, and apps. As Google expresses interest in creating an AI serving as a "Your Life Story Teller," the deployment plan for Project Ellmann remains uncertain. However, insights from a Google Photos product manager suggest potential integration with Google Photos, a platform already familiar with AI features such as the "Memories" view and the AI video editor.
It's noteworthy that Google has been intensifying its AI endeavors, introducing models like Gemini, which consists of three variants—Nano, Pro, and Ultra—tailored for diverse applications. While Pro excels at a wide array of tasks and Ultra handles complex ones, Nano specializes in on-device tasks. If Project Ellmann aligns with Google Photos, speculation leans towards it utilizing Nano, especially considering the recent introduction of Gemini Nano to the Pixel 8 series in Google's December 2023 feature drop, enabling applications like Recorder summaries and "Smart Reply in Gboard."
