Google Deepmind co-founder and director DEEPMIND DEEPMIND HASSABIS speaks during the Mobile Congress Mobile, the largest annual gathering of the telecommunications industry, in Barcelona, Spain, February 26, 2024.
Pao Barena | AFP | Getty Images
London – an artificial intelligence that can be aligned with people on any task is still in some way – but only a matter of time before it becomes reality, according to the director of the Executive Director Google Deepmind.
Speaking at the Brifing in London offices Deepmind Monday, Demis Hassabis said he thought that artificial general intelligence (AGI) – which is smart or smarter than people – will start appearing in the next five or 10 years.
“I think today’s systems, very passive, but there are still a lot of things I can’t do. But I think during the next five to 10 years a lot of these opportunities will start to come to the fore and we will start to move toward what we call artificial general intelligence,” Hassabis said.
Hassabis defined agi as “a system that can show all the complicated opportunities that people can.”
“We are not there yet. These systems are very impressive in certain things. But there are other things I can’t do yet, and before that we still have a lot of research works,” Hassabis said.
Hassabis is not alone suggesting that it will take some time for agi to appear. Last year, executive director of the Chinese Technology Gigant Baid Robin did he say he sees that he was an agi “Over 10 years“Pushing to the exciting predictions of some of your peers about this breakthrough that takes place in a much shorter time frame.
For a while to go
Hassabis’s forecast pushes the time tape to reach AGI in some way back compared to what his peers in the industry sketched.
Dario Amodi, AS AI Startup Anthropic CEO, told CNBC at the World Economic Forum in Davos in January in Switzerland to see the form of AI, which is “better than almost all people on almost all tasks” that appear in “the next two or three years.”
Other technological leaders see that agi arrive before. CISC’s product director Jeeta Patel thinks there is a chance to see an example of AGI to show up this year. “There are three main stages” for AI, Patel told CNBC in an interview at the Mobile World Congress event in Barcelona earlier this month.
“There is a basic AI that we are all currently experiencing. Then there is artificial general intelligence, where cognitive abilities are filled with those people. Then they call superintelligence here,” Patel said.
“I think you will see significant evidence that AGI is in the 2025 game. We’re not talking about years away,” he added. “I think superintelligence is at best for several years.”
Artificial super intelligence, or Asi, is expected to arrive after Agi and surpass human intelligence. However, “no one really knows” when such a breakthrough will happen, Hassabis said on Monday.
Last year, Tesla CEO Elon Musk predicted that Agi is likely to be available until 2026, while Openi CEO I Altman Sam said Such a system could develop in a “reasonable near future”.
What is it necessary to get to AGI?
Hassabis said that the main challenge is by achieving artificial general intelligence today’s AI systems to the point of understanding of the context from the real world.
Although it is possible to develop systems that can break down problems and perform tasks autonomously in the Empire of the Games – such as Complex strategy on board games go – bringing such technology into the real world proves more difficult.
“The question is, how quickly we can generalize with planning ideas and agent types of behavior, planning and resonate, and then generalize it to work in the real world, with things like a worldwide model – models that are able to understand the world around us,” Hassabis said. “
“And I think we have progressed with world models in recent years,” he added. “So the question is, what is the best way to combine this with these planning algorithms?”
Hassabis and Thomas Kurian, Executive Director of Google’s Cloud Computer Science Department, said the so-called “multi-agent” AI systems are technological progress that gets a lot of attractions behind the scenes.
Hassabis said that a lot of work was done to get to this stage. One of the examples he mentioned was Deepmind’s work, getting AI agents to make sense of how to play the popular Starcraft Strategic Game.
“In the past, we have done a lot of work to do with things like Starcraft Game, where you have agents or a league company, and they could compete, they could cooperate,” said Chief Deepmind.
“When you think about agents to communicate with agents, this is what we also do to allow the agent to express himself … What are your skills? What kind of tools are you using?” Said Kurian.
“These are all the elements you need to be able to ask an agent question, and then after you acquire that interface, other agents can communicate with it,” he added.
