Google is preparing artificial intelligence that can understand your life through Google Photos

According to CNBC newspaper, Google is preparing artificial intelligence (AI) that can understand the context of the recordings saved in Google Photos. The model, called “Project Ellman,” will have the ability to understand the content of an image based on the faces and people featured in the images.

The model will be able to extract context using biographies, old or new records to understand what is happening in each photo and explain the scenario.

In practice, Project Ellman can tell if a photo is a meeting of old college friends, a romantic encounter, the age of those involved, and deduce the stories behind a photo. Even screenshots can be evaluated by AI that tries to understand the reason behind the recording.

According to the CNBC source, Google must have realized that multimodal models are ideal for such vehicles. The company will present the new feature as a chatbot that knows everything about your life, like a true personal virtual assistant.

In one of the slides presented, the chatbot was able to detect the user’s food preferences, booked trips, and interests based on media saved on the mobile phone. The chatbot can even identify favorite websites, apps, and social networks through prints.

Project Ellman is being tested at Google

For now, the Ellman Project It will be subjected to internal testing by the Google Photos team.

“Google Photos has always used AI to help people search for their photos and videos, and we’re excited about the potential for Masters to unlock even more useful experiences. This was an early discovery internally, and as always, if new If we decide to publish photos, a spokesperson told the site: “We will take the time necessary to ensure these features are useful to people and are designed to protect users’ privacy and security as our top priority.”

Given that Google has adopted a slower pace in launching AI tools, it is possible that development of “Project Ellman” will take some time to complete. In any case, this description reveals a very interesting potential for multimodal models.