The head of the Google International Organization says that the capabilities of the new Gemini indicate AGI

Dimis Hasabis, CEO to Google DeepMindHe says that access to artificial general intelligence or AGI-is a mysterious term that is usually used to describe machines with human-like intelligence-it means sharpening some of the emerging capabilities in the leading Google twin Models.

Google has announced a large number of artificial intelligence promotions and new products on the annual I/o happened. Today in Mountain View, California. The search giant revealed the upgraded versions of Gemini Flash and Gemini Pro, the fastest and most capable Google, in a row. Hassabis said that Gemini Pro is outperforming other models on LMARNA, a widely used standard to measure the capabilities of artificial intelligence models.

Hassabis showed some experimental artificial intelligence shows that reflect a vision artificial intelligence This exceeds the chat window. “The way we ended up working with Chatbots today is, I think, a transitional period,” he told Hasabis Wire before today’s event.

Hasabis says that the potential of the emerging Gemini, the agent, and the models of the world can enable the most capable and pre -emptive personal assistants, and the really useful human robots, and eventually intelligence like anyone.

In I/O, Google has unveiled deep thinking, a more advanced type of emulator thinking of the Pro model. The latest artificial intelligence models can destroy problems and deal on them in a way that is very similar to human thinking than instinctive output of standard large language models. Tulsee Doshi says, many innovations that have not been revealed to improve this trick, says Tulsee Doshi, many innovations that have not been revealed to improve this trick.

Today, Google has revealed new products that depend on Gemini’s ability to think and take action. This includes Mariner, a chrome agent that can explode and do household chores such as shopping when giving an order. Mariner will be presented as a “research inspection” through a new subscription plan called Google Ai Ultra, costing $ 249.99 per month.

Google also offered a more capable version than Google ASTRA experimental assistantWhich can see the world and hear it through a smartphone or a pair of smart glasses.

In addition to talking about the surrounding world, ASTRA can now play a smartphone when needed, for example using applications or searching in the web to find useful information. Google showed a scene in which the user gets ATRA help to find spare parts needed to repair bikes.

Doshi adds that Gemini is trained to better understand how the user needs are prohibited, starting with launching a web search when it is useful. Future assistants will need to be pre -emptive without being annoying, as both Duch and Hasabis say.

ASTRA’s capabilities depend on Gemini’s modeling of the material world to understand how he works, which Hassabis says it is it is very important to biological intelligence. Artificial intelligence will need to refine his thinking, agency and creativity as well. “There are missing capabilities.”

Before AGI arrives well, artificial intelligence is to present the way people search on the web, which may affect the basic Google works deeply.

The company announced new efforts to adapt the research to the era of artificial intelligence in I/O (see WAID’s I/O Liveblog For everything announced today). Google will launch an AI version of Search Search called Mode Mode to everyone in the United States and will provide a AI shopping tool that allows users to download a picture to see how you will look like a component of clothes on it. The company will also provide an artificial intelligence overview, a service that summarizes the results of Google users, which are available in more countries and languages.

Converting time tables

Some researchers and critics argue from artificial intelligence that AGI may be just a few years away – or even here depends on how to define the term. Hassabis says it may take from five to 10 years so that the machines can master everything that a person can do. “This is still imminent in the big plan for things,” Hassan says. “But it is not tomorrow or next year.”

Hassabis says that thinking, agency and global modeling should not only be able to assist like ASTRA, but also give Human robots The minds they need to work reliably in the true, chaotic world.