Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Google Deepmind co -founder and Chief Executive Director Demis Hasabis appears during the Mobile World Congress, the largest annual telecommunications collection in Barcelona, ​​Spain, February 26, 2024.
Barrena Pau | AFP | Gets the image
London – artificial intelligence that can fit people on any task, still some way – but this is just a matter of time before it becomes reality, according to the CEO Google Deepmind.
Speaking at a briefing at the Deepmind London offices on Monday, Demis Hasabis said he believed that in the next five or 10 years he believes that artificial general intelligence (AGI) is just as intelligent or smarter than people.
“I think that today’s systems are very passive, but there are many more things they can’t do. But I think over the next five -10 years will start to come to the fore, and we will start to move to what we call artificial general intelligence,” Hassabis said.
Hassabis has identified AGI as “a system that can show all the complex opportunities that people can.”
“We are not quite there yet. These systems are very impressive in certain things. But there are other things they cannot do yet, and we still have a lot of scientific work to go to it,” Hassabis said.
Hasabis is not one, assuming that it will take some time to get the agitate. Last year the CEO of Chinese Technology Giant Bake Robin Lee said he sees Agm ‘More than 10 years“Pushing away from the stunning forecasts of some of their peers about this breakthrough, which occurs in much shorter time.
Some time is still needed
Hasabis’s forecast is pushing the term to reach some back compared to the fact that his peers are painted in the industry.
Dario Amadeus, CEO of AI Anthropic Anthropic, told CNBC at the World Economic Forum in Switzerland in Davos, in January that he sees the form II, which is “better than almost all people on almost all tasks” that arise in “The next two years”.
Other technological leaders see that AGA is coming earlier. Cisco Jeetu Patel Products CEO believes that we will be able to see an example that will appear this year. “There are three main stages”, “the fodder told CNBC in an interview at the Mobile World Congress in Barcelona earlier this month.
“There is the main II, which we are all experiencing now. Then there is an artificial general intelligence where the cognitive opportunities meet with people. Then there is what they call the warden,” the pan said.
“I think you will see significant evidence that AGA plays in 2025. We don’t talk about years,” he added. “I think at best for a few years.”
It is expected that artificial super intelligence, or ASI, will come after the AGA and exceed human intelligence. However, “no one really knows” when such a breakthrough happens, Hasabis said on Monday.
Last year Tesla CEO Elon Muscovy predicted that AGI is probably Such a system can be developed in a “reasonable near future”.
What is needed to achieve AGA?
Hasabis noted that the main problem of achieving artificial general intelligence is that today’s AI systems to understand context from the real world.
Although it was possible to develop systems that could destroy the problems and autonomously complete the tasks in the field of games – for example, Difficult Capital Game Go Go Go Go – The introduction of such technology into the real world is stronger.
“The question is how quickly we can summarize the ideas of planning and agency, planning and reasoning, and then summarize it for work in the real world, besides how world models are models that can understand the world around us,” said Hasabis. “
“And I think we have made good progress in world models over the last couple of years,” he added. “So, now the question is what the best way to combine it with these planning algorithms?”
Hasabis and Thomas Kurian, CEO of Google Cloud Calculations, said the so-called “multi-agen” systems are technological progress, which takes a lot of craving.
Hasabis said that there is a lot of work to go to this stage. One example he referred to is the Deepmind’s work that makes AI agents to find out how to play the popular Starcraft strategic game.
“We have done a lot of work on this things such as the Starcraft game in the past where you have agents or agents, and they can fight, they can cooperate,” said the Deepmind chief.
“If you think about the agent’s agent’s connection, this is what we do too to let the agent show yourself … What are your skills? What tools are you using?” said Curian.
“These are all the elements you need to be able to ask the agent a question, and then, if you have this interface, then other agents can communicate with it,” he added.
