Top scientists are alerted by the control AI
Yoshua Bengio (L) and Max Tegmark (R) discusses the development of artificial general intelligence in the Live Podcast of CNBC in the Swiss city of Davos in January 2025.
CNBC
Artificial intelligence, built as “agents”, can be dangerous because their creators can lose control of the system, two of the world’s most famous AI scientists told CNBC.
In the latest episode of CNBC “Out of the valley” On Tuesday, Max Tegmark, Massachusetts Institute of Technology and Institute of Life and Yoshua Bengio, Yoshua Bengio and Yoshua Bengio, “EU thinking” and the EU thinking “and a professor in University de Montréal. concerns about artificial general intelligence or AGI. The term provides extensive information to AI systems, which are smarter than humans.
Their fears are caused by the largest companies in the world, “AI agents” or “Agentic AI” or “Agentic AI” – will require any companies to behave as auxiliary or agent and help in the work and daily life. Industrial calculations vary when AGI has formed.
According to this concept, according to Bengio, some of the AI ​​systems are “agency” and the idea that it can be their own thoughts.
“Researchers in the EU inspired the human intelligence to establish intelligence and understanding the world as pure intelligence, but also to understand the world as an agentic behavior … There is a mess to use your knowledge to achieve your goals.” Bengio said CNBC’s “outside the valley.”
“Currently, this is AGI: It is a very dangerous offer that understands many things about the world and then.”
Bengio would be “Create a new kind of new kind of new kind of” and “to create a new wise body on this planet” and “would not know how to deal with our needs.”
“Therefore, we can think, what are the scenarios in which the work worsened and all trusts the agency? In other words, because this is because the problem has its own goals.”
Bengio said that the idea of ​​protecting himself can also begin.
“Do we want to compete with the bodies smarter than us? We must understand how a very consolation gamble, which means protection can act as a purpose in the AI.”
Key AI tools
The keyword for MIT’s Tegmark is located in the systems created for a certain, narrow defined purpose – a narrow defined purpose, but you do not need to be an agent.
Tegmark will be able to handle how can be a means that cancer be treated and can be a means that can be a really reliable means of being a really high level of cancer . “
“I think we can do almost everything we are excited in a optimistic note here, we can do almost everything … if people have some basic security standards before they sold their strong AI systems,” he said.
“They must demonstrate them if we can control them. Then the industry will innovate the industry rapidly to know what this will do.”
The future of Tegkark in 2023 called for a break to the development of AI systems that can compete with human level intelligence. When this did not happen, Tegmark said that people talked about the subject and now it’s time to move the guards to manage AGI.
“So, at least a lot of people now talk about conversations. We have to see if we can get them to walk.”
“It is clear madness to build something smarter than us before we understand how we will manage.”
AGI’s future has a few ideas, partly controlled by different definitions.
Openai CEO SAM Altman said that the company knows the AGI and lower the influence of technology, he said that people would come sooner than they think.
“I’m guessing, we will hit the majority of most people in the world and we will give it less attention to it,” he said gossip In December.
English