Nobel Prize in Physics for Geoffrey Hinton, the godfather of artificial intelligence who warns of its dangers

These findings laid the foundation for the creation of today’s major language models, such as ChatGPT and other chatbots that have changed the world. But Hinton, whose company Google bought, left Silicon Valley last year to warn of the dangers of the technology he helped create. has stated many times that the day may come when machines become so intelligent that we are incapable of controlling them, an idea that he has repeated again after learning, “very surprised”, of his award.

“Learning is a fascinating ability of the human mind. We put millions of neurons to work together, giving them neurological abilities. Machine learning is inspired by this and uses fundamental concepts from statistical physics,” he said. said Ellen Moons, president of the Nobel Committee in Physics, after the announcement of the winners. Their findings “have enormous benefits that affect our future and it is our responsibility to use this technology ethically, for the greatest benefit of humanity,” he said.

Hopfield (Princeton University, USA) and Hinton (University of Toronto, Canada) have been developing their work for decades. The American created an associative memory that can store and reconstruct images and other types of patterns in data. For his part, the British used the Hopfield network as a basis to create a new one with a different method: the Botzmann machine, which could learn to recognize characteristic elements in a certain type of data. The machine is trained by feeding it examples that are most likely to arise when it is run. Thus, it is able to analyze thousands of photos and identify common things, such as dogs and cars, very efficiently. Google bought the company from him, which became the basis for developing increasingly powerful technologies, including new chatbots such as ChatGPT and Google Bard.

Threat to humanity

But last year, Hinton quit his job in Silicon Valley to warn of the dangers posed by these chatbots. He has come to point them out as a “threat to humanity” for its ability to “eliminate millions of jobs, compromise elections over fake videos, launch much more effective cyberattacks, and develop very dangerous biological warfare viruses.”

After learning about the award, Hinton, who was also awarded the BBVA Foundation’s Frontiers of Knowledge in 2017 and the Princess of Asturias Award for Scientific and Technical Research in 2022, said he felt “overwhelmed.” “I don’t know what happened. I’m very surprised,” he stated briefly in a telephone connection with the academy. “I’m in a cheap hotel in California that doesn’t have a good internet or phone connection,” he explained. “I was going to have an MRI today, but I’ll have to cancel it,” he said with humor.

When asked if he regrets his discovery and its consequences, he stated that “there are two types of regrets: those you have because you feel guilty for having done something and those you have for what you could have done, but you don’t do.” “In my case it is the second. I am worried about the consequences of machines becoming so intelligent that we have no control over them.”

Potential risks

That a Nobel Prize winner regrets the consequences of his findings is not usually the most common. Regarding these potential dangers, Moons has pointed out that it is “good that concerns are expressed”, since they “contribute to knowledge.” In his opinion, “it is important that as many people as possible have access to machine learning, that it is not only in the hands of a few individuals.” In this sense, Hans Ellegren, secretary general of the Royal Swedish Academy of Sciences, has indicated: “Many discoveries potentially have risks, but it is our responsibility to regulate it and that also includes artificial intelligence.”

Despite the possible threats, “the work of the winners has already been very useful. In physics, we use artificial neural networks in a wide range of areas, such as in the development of new materials with specific properties,” defended Moons. .

Santiago Mazuelas, researcher at the Basque Center for Applied Mathematics (BCAM), considers the new Nobel Prize in Physics “very deserved.” The artificial neural networks on which the winners have worked for decades “have been very important in the field of machine learning and have made technologies such as ChatGPT or text and image generating applications possible.”

He does not share Hinton’s fears: “I believe that AI is just another type of technology like the Internet or telephone, and like any other technology with great potential, it can be used for perverse purposes. But, in general, it has a positive effect on society”. According to Mazuelas, beneficiary of a Leonardo scholarship from the BBVA Foundation in 2018 for a machine learning project, AI “only reproduces the behavior observed in many examples and learns patterns that humans are not capable of observing. However, “Human intelligence is capable of logical deduction and taking into account ethical considerations, among many other capabilities.”

In addition to recognition, the winners receive a sum that exceeds 800,000 euros. Last year, the Royal Swedish Academy of Sciences awarded Anne L’Huillier, Pierre Agostini and Ferenc Krausz for having created, through experiments, flashes of light short enough to capture snapshots of the extremely fast movements of electrons.

As explained on the awards website, the Nobel Prize in Physics has been awarded 117 times to 224 experts between 1901 and 2023. John Bardeen is the only winner to have received it twice, in 1956 and 1972.

On Monday, the Nobel Prize in Medicine 2024 was awarded to Victor Ambros and Gary Ruvkun for a fundamental discovery about how cells function: microRNAs, small RNA molecules that play a key role in gene regulation.