When you ask a question to an AI chatbot like ChatGPT, you’re using a piece of software loosely inspired by the structure of our brains, an artificial “neural network.” Nobel laureates in physics contributed to the development of such networks, and to their theoretical understanding, in the 1980s – decades before they became truly useful. The researchers used concepts and techniques from physics. In turn, neural networks have made progress in physics in recent years, according to the Nobel Prize jury, for example in analyzing the results of experiments on elementary particles.
John Hopfield (91 years old) is affiliated with Princeton University in New Jersey, United States. He created a type of neural network known as a Hopfield network in the early 1980s that can recognize information such as images from incomplete or distorted data. He used mathematical techniques from statistical physics, which were originally developed to understand some magnetic materials. You can see such a substance as a network of atoms, each acting like little magnets and influencing each other. Hopfield saw the analogy with the neural network of brain cells (mimics) and their interconnections.
Risks
Geoffrey Hinton (76 years old), born in London, affiliated with the University of Toronto in Canada. He built on Hopfield’s work and developed a “Boltzmann machine,” another type of neural network, that can classify images or create new examples of the type of pattern it has been trained on. Boltzmann networks were theoretically interesting, but turned out to be of no practical use. However, Hinton later also contributed much to the development and spread of “Back propagation“, a practical algorithm for training neural networks. Later he played an important role in the development of Deep learningModern neural network technology. Sometimes it becomes like “Godfather of artificial intelligence” he described. In 2018, he won the Turing Prize, an award that is seen as the Nobel Prize for computer science — one of the many academic disciplines that does not have a true Nobel Prize of its own.
From 2013 to 2023, Hinton worked at Google. He left that company because, he said, he wanted to be free to speak out about the dangers of artificial intelligence. He warns that in about twenty years, artificial intelligence could become smarter than humans and take over, possibly with disastrous consequences.
On the occasion of his resignation from Google, Hinton also stated that he regretted his life’s work. He confirmed this (via video link) at the press conference in which the Nobel Prize was announced. “In the same circumstances I would probably do the same.” But I worry about the ultimate consequences if systems smarter than us end up taking control. He predicts that artificial intelligence will eventually become superior to us in terms of intellectual ability. “And we have no experience working with anything smarter than ourselves.”
When asked if he uses AI himself, Hinton replied: “I use GPT-4 a lot, although I don’t trust it completely. It’s an expert in everything, but it’s not a good expert.” He added that it is very useful. Hinton says he had no idea he was nominated for the Nobel Prize in Physics.
I’m amazed. “I had no idea this could happen,” he responded to the news. “It was a bolt from the blue.” His genuine astonishment is evident from the fact that he was planning something else: “I went to the hospital today for an MRI – I might have to cancel that.”
