An arms race for the supremacy of artificial intelligence (AI), triggered by recent panic on the Chinese chatbot Deepsek, risks amplifying the existential dangers of superintelligence, according to one of the “sponsors” of the AA .
The Canadian Pioneer of Automatic Learning Yoshua Bengio, author of the first international IA security report which will be presented at an International AI Summit in Paris next week, warns uncontrolled investment in power IT for unattended AI is dangerous.
“The effort is to get started in who will win the race, rather than how to make sure that we are not going to build something that explodes us,” said Bengio.
It warns that military and economic races “lead to the reduction of ethics corners, the reduction in corners of responsibilities and security. It is inevitable ”.
Mr. Bengio worked on neural networks and automatic learning, the software architecture underlining modern AI models.
He is in London, with other Pioneers of the AI to receive the Queen Elizabeth Prize for Engineering, the most prestigious world prize for engineering, in recognition of AI and its potential.
He is enthusiastic about his advantages for the company, but the pivot of AI regulation Donald TrumpThe White House and the frantic competition between large technological companies for more powerful AI models are a worrying change.
“Surhuman systems become more powerful”
“We build increasingly powerful systems; Become superhuman in certain dimensions, ”he says.
“As these systems become more powerful, they also become extraordinarily more precious, economically speaking.
“So the extent of” wow, it will make me a lot of money “motivates a lot of people. And of course, when you want to sell products, you don’t want to talk about risks. »»
But all the “sponsors” are not so worried.
Take Yann Lecun, the chief scientist of Meta AI, also in London to share the Qeprize.
Yann Lecun, chief scientist of Meta AI
“We were deceived by thinking that the models of big languages are intelligent, but really, they are not,” he says.
“We have no machines that are almost as intelligent as a house cat, in terms of understanding the physical world.”
In the three to five years, predicted Mr. Lecun, the AI will have certain aspects of intelligence at the level of man. The robots, for example, which can perform tasks that they have not been scheduled or trained to perform.
Find out more:
What is Deepseek? The low -cost Chinese AI company which has overthrown the technological world
Bill Gates says he would be diagnosed with autism if he was young today
But, he maintains, rather than making the world less safe, models of open source AI such as In depth – A chatbot developed by a Chinese company rivating with the Best of America Big Tech with a tenth of computer power – demonstrates that no one will dominate for a long time.
“If the United States decide from Clamour with regard to AI for geopolitical reasons, or, commercial reasons, then you will have innovation elsewhere in the world. Deepseek showed it, ”he says.
The price of Queen Elizabeth for the engineering price is awarded each year to engineers whose discoveries have or promise to have the greatest impact on the world.
Previous beneficiaries include pioneers of photovoltaic cells in solar panels, wind turbines and neodymium magnets found in hard drives and electric motors.
The Minister of Sciences Lord Vallance, who presides over the Qeprize Foundation, says that he is alert the potential risks of the AI.
Organizations such as the new IA security institute in the United Kingdom are designed to predict and prevent potential damage from intelligence “similar to humans”.
Minister of Sciences Lord Vallance
But he is less concerned with a nation or a company with a monopoly on AI.
“I think what we have seen in recent weeks is that it is much more likely that we are going to have many companies in this space, and the idea of single -point domination is quite unlikely,” said -he.