Google DeepMind, the founder and executive director Demis Hassabis, speaks during the mobile world conference, the largest annual communication industry gathering, in Barcelona, Spain, February 26, 2024.
Barrena Pau | AFP | Gety pictures
London – Artificial Intelligence that can fit with humans on any task that still exists – but it is just a matter of time before it becomes a reality, according to the CEO of Google Deepmind.
Speaking at a press conference at Dibmand’s offices in London on Monday, Dimis Hasabis said he believed the AGI – and it is an intelligent or more intelligent people of humans – will begin to appear in the next five or ten years.
“I think today’s systems are very negative, but there are still a lot of things they cannot do. But I think over the next five to ten years, many of these capabilities will start in the lead and we will start moving towards what we call artificial public intelligence,” Hasabis said.
Hassabis defined AGI as “a system capable of showing all the complex capabilities that humans can.”
“We are not there yet. These systems are very great in some things. But there are other things they cannot do yet, and we still have a lot of research work to go before that,” Hasabis said.
Hassabis is not alone indicating that it will take some time until AGI appears. Last year, CEO of a Chinese technology company Bido Robin told me that he sees Agi.More than 10 years“Return to exciting predictions of some of his peers about this penetration that occurs in a much shorter time frame.
Some time to go yet
Hassabis expectations pushing the timeline to reach AGI in a direction compared to what his peers were in the industry.
Dario Ameudi, CEO of AI Startup Anthropic, told CNBC at the World Economic Forum in Davos, Switzerland in January, he sees a form of artificial intelligence “better than almost all people in almost all tasks” emerging in “two or thirty” years.
Other technology leaders AGI see their access to a closer time. Jetu Patel, CIECO chief product official, believes that there is an opportunity because we can see an example of AGI that appears as soon as this year. “There are three main phases” for Amnesty International, Patel told CNBC in an interview at the World Consms Mobile event in Barcelona earlier this month.
“There is the main artificial intelligence that we all experience at the present time. Then there is artificial general intelligence, as cognitive capabilities meet the rules of human beings,” Patel said.
He added: “I think you will see useful evidence that Agi is playing in 2025. We are not talking about years.” “I think SuperInte Britivity, at best, is a few years outside.”
It is expected that the super artificial intelligence, or ASI, after AGI and bypass human intelligence. However, “No one really knows” when this penetration will happen, the account said on Monday.
Last year, Elon Musk, CEO of Tesla, expected that AGI would be available by 2026, while Openai SAM Altman said. Such a system can be developed in a “reasonable proximity”.
What is required to reach AGI?
A account said that the main challenge in achieving artificial general intelligence is to obtain artificial intelligence systems today to the point of understanding the context of the real world.
Although it was possible to develop systems that could break problems and complete tasks independently in the world of games – such as Complex strategy board game goes Bringing such technology to the real world proves to be more difficult.
Hasabis said: “The question is, to what extent we can generalize the ideas of planning and a kind of behaviors, planning and thinking, then generalizing this to work in the real world, in addition to things like the world models – models capable of understanding the world around us.”
“I think we have made good progress with the world’s models over the past two years,” he added. “Now the question is, what is the best way to combine this with these planning algorithms?”
The so -called “multi -agent” systems of artificial intelligence are technological progress that acquires a lot of traction behind the scenes.
The account of a lot of work is being done to reach this stage. One of the examples he referred to is DeepMind to get artificial intelligence agents to know how to play the common strategy “Starcraft”.
“We have made a lot of work on it with things like Starcraft in the past, where you have an association of agents, or in the agents association, and they may compete, they can cooperate,” said DeepMind.
“When you think about the agent in the agent’s connection, this is what we also do to allow the agent to express himself … What are your skills? What kind of tools do you use?” Corian said.
He added: “These are all elements that need to be able to ask a question to one of the factors, and then once you get that interface, other agents can communicate with it.”
https://image.cnbcfm.com/api/v1/image/107421857-1717071123152-gettyimages-2033216655-AFP_34KA2N6.jpeg?v=1742211497&w=1920&h=1080
Source link