Looking for scenes for Gemini AI large model Google Project Ellman project exposed

The Google team proposed an idea to use AI technology to process user photos and search engine query information to comprehensively analyze what the user’s life is like. This project is called "Project Ellmann".

According to the original idea, Project Ellmann would use LLM (such as Gemini) to extract information, find specific patterns from the user's photos, and then build a chatbot to accurately answer questions. Google claims Ellmann will be a "personal life storyteller."

Sounds interesting, but does Google want to embed Ellmann functionality into Google Photos, or into other products? It's still unknown. Google Photos has more than 1 billion users and stores 4 trillion photos or videos.

Google is currently using AI technology to optimize its product line, and Ellmann is just one of the ways it can be used. A few days ago, Google launched Gemini. In some cases, Gemini has even surpassed OpenAI GPT-4

Google intends to license Gemini to Google Cloud users, who can develop with the help of Gemini The functions you want. The most prominent advantage of Gemini is "multi-modal", which means that it can understand multiple types of information such as text, pictures, videos, and audio.

At a recent internal meeting, a Google Photos executive showed off Project Ellman. Internal documents show that the Google team believes that large language models can be used to display users’ life stories in the real world in a “bird’s eye” way.

By reading biographies, previous records and photos, Ellman was able to gain a deeper understanding of the situation. For example, by analyzing a series of clips, Ellman can determine whether the information touches on the topic of college life. One sentence in the presentation slide reads: "If you cannot look at an individual's life from a bird's-eye view, We wouldn't be able to solve the puzzles and tell the stories." Another said: "We look at your photos, look at their tags and locations, and then determine if they belong to something meaningful. moment. Your personal story will become clearer as we take a holistic view of your life."

The Google team also talked about Ellmann Chat. What is Ellmann Chat? The team described: "You might as well imagine that when you open it, it already knows everything related to your personal life. What questions would you ask it?"

For example, the user asked Ellmann Chat a question: "Do I have a dog?" AI replied that the owner has a dog, and also told the dog's name. It also said that there are two people in the family who like to be with dogs the most. What are the names of the family members.

Another problem is that the owner wants to move and asks the AI ​​to provide some towns to choose from that are similar to the existing living environment. Ellmann can answer these questions. Not only that, Ellmann also knows what the user's eating habits are.

For example, when a user wants to buy something, what they are interested in, what work and travel plans they have, Ellmann can find the answer by analyzing photos. Ellmann also knows which websites users like to visit and which apps they use. A Google spokesperson revealed: "Google Photos has always hoped to use artificial intelligence technology to help users search for photos and videos. With LLM, we can Provide a more practical experience. Currently, Ellmann is still in the early stages of internal exploration, and it will take time to determine whether it will be eventually launched. Google needs to ensure that Ellmann is truly helpful to users, while also considering user privacy and security."

There are many technology companies that want to use new technologies to create more personalized personal memories for users, and Project Ellmann is just one of them.

Whether it is Google Photos or Apple Photos, they are all analyzing photos, finding patterns, and then creating photo albums. Google Photos can automatically group similar photos together into albums for easy search by users. Apple said in June that its software would be able to identify people, dogs or cats in photos after an upgrade. Users only need to say the person's name and the software will help you find their photos.

Sounds good, but Google and Apple's technologies still have imperfections, and sometimes errors occur. In 2015, Apple and Google's artificial intelligence mistakenly identified a black person as a chimpanzee, a mistake that caused a lot of jokes. Even now, this kind of mistake is still difficult to avoid

The above is the detailed content of Looking for scenes for Gemini AI large model Google Project Ellman project exposed. For more information, please follow other related articles on the PHP Chinese website!