A Google team has suggested leveraging mobile phone data, including images and searches, to generate a “bird’s-eye” perspective of users’ life through artificial intelligence.
The plan was to use LLMs like Gemini to process search results, identify patterns in a user’s photos, build a chatbot, and “answer previously impossible questions.” The project was named “Project Ellmann,” after biographer and literary critic Richard David Ellmann. A copy of the presentation was seen by CNBC. According to the statement, Ellmann wants to be “Your Life Story Teller.”
Whether the business intends to develop these features for Google Photos or any other product is unknown. According to a business blog post, Google Photos has four trillion photos and videos and more than one billion users.
Google is utilising artificial intelligence (AI) in a number of ways, Project Ellman being just one of them. In certain instances, Google’s Gemini AI model topped OpenAI’s GPT-4 when it was released on Wednesday as the “most competent” and sophisticated model to date. Through Google Cloud, the business intends to licence Gemini to a broad spectrum of clients, allowing them to utilise it in their own applications. One of Gemini’s most notable qualities is its multimodality—the ability to receive and comprehend data that goes beyond text, such as images, audio, and video.
At a recent corporate meeting, Project Ellman was presented by a Google Photos product manager alongside Gemini teams, according to papers obtained by CNBC. They stated that over the course of the previous few months, the teams came to the conclusion that massive language models are the best technology available for enabling an aerial view of a person’s entire story.
In order to characterise a user’s photos more fully than “simply pixels with labels and metadata,” Ellmann may incorporate context through biographies, earlier events, and later images, according to the presentation. It suggests that a number of moments, such as college years, Bay Area years, and parent years, should be allowed to be recognised.
Source (CNBC)
