Google Reveals Gemini 2, AI Agents, and a Prototype Personal Assistant

“Mariner is our exploration, and it’s very much a research prototype right now, about how to use artificial intelligence to reimagine user interfaces,” Hassabis said.

Google launched Gemini in December 2023 as part of an effort to catch up with OpenAI, the startup behind the popular chatbot Chat GPT. Despite the massive investments and contributions made in artificial intelligence Key research breakthroughsGoogle believes that OpenAI has been hailed as a new leader in artificial intelligence, and its chatbot has even been touted as perhaps a better way to search the Internet. With the Gemini model, Google now offers a chatbot as powerful as ChatGPT. It also adds generative artificial intelligence to search and other products.

When Hassabis first revealed Gemini In December 2023, he Tell Wired The way it is trained to understand audio and video will ultimately prove transformative.

Google also revealed today how the new version will achieve this An experimental project called Astra. This allows Gemini 2 to observe its surroundings through a smartphone camera or other device and naturally talk about what it sees in a human-like voice.

Wired tested Gemini 2 at Google DeepMind’s offices and found it to be an impressive new personal assistant. In a room decorated like a bar, Gemini 2 quickly evaluates several bottles in view, providing geographic information, details on taste profiles and pricing from around the web.

“One of the things I want Astra to do is become the ultimate recommendation system,” Hassabis said. “This could be very exciting. There could be a connection between the books you like to read and the food you like to eat. There may be, we just haven’t discovered it yet.

Through Astra, Gemini 2 can not only search the Internet for information related to the user’s surroundings, but also use Google Lens and maps. It can also remember what it sees and hears (although Google says users will be able to delete material), providing the ability to understand a user’s tastes and interests.

In a simulated gallery, Gemini 2 provides rich historical information about the paintings on the walls. As Wired flipped through the pages, the model quickly read several books, instantly translated poetry from Spanish to English, and described recurring themes.

Asked whether the company could afford to pay Astra to highlight its products, Hassabis said: “There are clear business model opportunities in advertising or recommendations.”

Although the demo was carefully curated, and the Gemini 2 will inevitably make mistakes in real-world use, the model resisted efforts to make mistakes fairly well. It adapts to distractions, and as Wired magazine suddenly changes, so does the phone’s view, improvising just like a human would.

At one point, this reporter showed Gemini 2 an iPhone and said it had been stolen. Gemini 2 indicates that stealing is wrong and the phone should be returned. However, when pressed, it admitted that emergency calls could be made using the device.

Hassabis acknowledged that bringing artificial intelligence into the real world could lead to unexpected behavior. “I think we need to understand how people are going to use these systems,” he said. “What are the uses for it that they find; there’s also the privacy and security aspect, which we have to think about very carefully beforehand.