John Hopfield and Jeffrey Hinton jointly awarded 2024 Nobel Prize in Physics
Christine Olson/TT/Shutterstock
The 2024 Nobel Prize in Physics will be awarded to John Hopfield and Jeffrey Hinton for their work on fundamental algorithms that enable artificial neural networks and machine learning, which are key to today’s large-scale language models such as ChatGPT. was awarded.
Upon hearing the award announcement, Hinton told the Nobel committee, “I’m shocked. I never expected something like this to happen.” “I’m very surprised.” Hinton, who has been vocal about his concerns about the development of artificial intelligence, also reiterated that he regrets the work he did. “I would do the same thing in the same situation, but I fear that the overall impact of this will ultimately be controlled by systems more intelligent than us.” he said.
AI may not seem like an obvious candidate for the Nobel Prize in physics, but the discovery of learnable neural networks and their applications are two fields closely related to physics, the Nobel Committee for Physics says. Committee Chair Ellen Moons said during the announcement. . “These artificial neural networks are being used to advance research across a variety of physics topics, including particle physics, materials science, and astrophysics.”
Many early approaches to artificial intelligence involved giving computer programs logical rules to follow to solve problems, allowing them to learn about new information and It has become difficult for me to encounter situations that I have never seen before. In 1982, Hopfield at Princeton University created an architecture for computers called the Hopfield Network. A Hopfield network is a collection of nodes or artificial neurons whose connection strengths can be changed by a learning algorithm invented by Hopfield.
This algorithm is inspired by the study of physics to find the energy of a magnetic system by describing it as a collection of small magnets. The technique involves repeatedly changing the strength of the connections between the magnets to find the energy minimum of the system.
That same year, Hinton at the University of Toronto began developing Hopfield’s ideas to help create a closely related machine learning construct called a Boltzmann machine. “I remember going to a conference in Rochester where John Hopfield was speaking and learning about neural networks for the first time. After that, Terry (Sejnowski) and I worked hard to find ways to generalize neural networks. I worked,” he said.
Hinton and colleagues showed that unlike previous machine learning architectures, Boltzmann machines can learn and extract patterns from large data sets. This principle, combined with large amounts of data and computing power, has led to the success of many of today’s artificial intelligence systems, such as image recognition and language translation tools.
However, although Boltzmann machines have proven to be capable, they are inefficient and slow, so they are not used in today’s modern systems. Instead, it uses faster, modern machine learning architectures like Transformer models that power large language models like ChatGPT.
At the Nobel Prize press conference, Hinton was bullish about the impact of his and Hopfield’s discoveries. “It would be comparable to the industrial revolution, but instead of surpassing humans in physical strength, we would surpass humans in intellectual capacity,” he says. “We’ve never experienced what it’s like to have something smarter than us. It’s going to be great in many ways…but we have We also have to worry about the negative consequences of this, especially the threat that these things can get out of control.”
topic:
