![\"Nobel](\"http:\/\/seebeyond.shop\/wp-content\/uploads\/2024\/10\/20241008_news_icon.jpg\"\/){kind=icon}
<\/p>\n
The Nobel Prize medal. [Image: Photo by Cl\u00e9ment Morin; \u00a9 Nobel Prize Outreach]<\/p>\n<\/div>\n
The Royal Swedish Academy of Science has awarded the 2024 Nobel Prize in Physics to John Hopfield of Princeton University, USA, and Geoffrey Hinton of the University of Toronto, Canada, for \u201cfoundational discoveries and inventions that enable machine learning with artificial neural networks.\u201d<\/p>\n
In awarding the prize to Hopfield and Hinton, the academy highlighted work from the mid-1980s that is having an almost unimaginable\u2014and still ultimately uncertain\u2014impact some 40 years later. In the field of optics alone, contemporary applications of machine learning, built on the foundational computer-science work of the two laureates, show up in areas as diverse as device design, the search for new laser materials, new ideas on diffractive analog networks for processing data with light<\/a>, image analysis and a host of other areas. And the burgeoning data demands of AI computation are spurring new interest in energy-efficient optical approaches<\/a> to data transmission and even neural-net processing itself.<\/p>\n Soon after the advent of electronic computers, beginning in the 1950s, researchers wondered whether the machines might be pressed into service in various forms of pattern recognition, a fundamental AI task. They also began to speculate on whether approaches adapted from modeling what goes on in the human brain\u2014so-called artificial neural networks (ANNs)\u2014might do the trick.<\/p>\n Early efforts to realize that promise, such as the feedforward network for image interpretation devised by Frank Rosenblatt in 1957, unearthed some of the core concepts. But the outputs of this early work ultimately proved disappointing, leading to the \u201cAI winter\u201d in funding and research opportunities that started in the early 1970s.<\/p>\n John J. Hopfield (top) and Geoffrey E. Hinton. [Image: Illustration by Niklas Elmehed; \u00a9 Nobel Prize Outreach]<\/p>\n<\/div>\n A thaw of sorts began in the 1980s, with breakthroughs, including those of the two 2024 Nobel physics laureates, that rekindled interest in AI and machine learning. In 1982, one of the pair, John Hopfield\u2014a theoretical physicist who was already a significant figure in biological physics\u2014extended that background by proposing a simple model for associative memory in a recurrent neural network, which came to be called a Hopfield network.<\/p>\n Hopfield wondered if a large enough network of this type could ultimately enable \u201cemergent\u201d computational abilities analogous to those of the brain. He followed up his initial work with subsequent studies fleshing out how the model might be applied in pattern completion, error correction and various kinds of optimization problems.<\/p>\n Geoffrey Hinton, meanwhile, working with a number of colleagues, used the concepts of statistical mechanics to extend Hopfield\u2019s initial network, creating a so-called Boltzman machine. The application of statistical techniques was a crucial advance. That\u2019s because, unlike the Hopfield model, the Boltzman machine was a generative model, capable of drawing inferences and generating new data instances based on probability distributions.<\/p>\n Even more consequential, perhaps, was Hinton\u2019s development (with David Rummelhart and Ronald Williams) of a workable feedforward neural network, enabled through so-called backpropagation algorithms. These algorithms, which involve a \u201chidden\u201d layer in the neural network that handles error correction through a loss function and selective weightings, have become a core tool in training so-called deep neural networks, and underlie most of today\u2019s headline-grabbing ANNs such as GPT4.<\/p>\n It would be difficult to overstate the importance and impact of these findings from four decades ago on the AI advances that are now reshaping science and society. That impact is apparent in high-profile tools ranging from Google Translate to ChatGPT; in face-recognition applications (for both good and ill); in efforts to apply machine intelligence to the interpretation of medical images; and in much more.<\/p>\n In science, meanwhile, AI techniques based on the 2024 laureates\u2019 foundational work have energized discovery in a variety of data-intensive undertakings. These capabilities have proved particularly useful in high-resolution climate modeling, particle physics, and astrophysics. The eye-catching images<\/a> of the black hole at the center of the Milky Way from the Event Horizon Telescope, for example, would likely have been impossible without the data-processing contribution of ANNs.<\/p>\n
<\/p>\n It\u2019s difficult to identify an area of optical research that has not been touched by the work of Hopfield and Hinton.<\/p>\n
\n<\/p><\/blockquote>\n AI tools are also spawning radically new ways of doing research. A case in point: A recent study in Science<\/em> documented the use of so-called self-driving labs\u2014which involve multi-institution, distributed laboratories combining AI and robotics\u2014to uncover scores of new candidates for laser materials<\/a> in record time. Elsewhere in optics and photonics, researchers are increasingly turning to libraries of AI software tools to supercharge device design<\/a>, to refine lens setups<\/a> for augmented-reality applications, to identify tiny flaws in solar cells<\/a>, to develop new tools for image interpretation and noise reduction in communications, and for a raft of other undertakings. Indeed, it\u2019s difficult to identify an area of optical research that has not<\/em> been touched by the work of Hopfield and Hinton.<\/p>\n Reached by phone during the press conference announcing the prize, one of the laureates, Geoffrey Hinton\u2014who admitted that he was \u201cflabbergasted\u201d by the news\u2014acknowledged the prospects of the AI technology his work has advanced for improving human society.<\/p>\n \u201cI think it will have a huge influence; it will be comparable with the industrial revolution,\u201d he said. \u201cBut instead of exceeding people in physical strength, it\u2019s going to exceed people in intellectual ability.\u201d That capability, he said, should prove \u201cwonderful in many respects,\u201d in areas such as health care, industrial efficiency and economic productivity.<\/p>\n
<\/p>\n Hinton said that the impact of AI should prove \u201cwonderful in many respects\u201d\u2014but, interestingly, he also alluded to some of the dark auguries surrounding AI that have become a staple of debate.<\/p>\n
\n<\/p><\/blockquote>\n Yet, interestingly, Hinton also alluded to some of the dark auguries surrounding AI that have become a staple of political debate, popular art and science fiction. \u201cWe have no experience of what it\u2019s like to have things smarter than us,\u201d he said, adding that \u201cwe have to worry about a number of possible bad consequences, particularly the threat of these things getting out of control.\u201d<\/p>\nLandmark work in pattern recognition<\/h2>\n
![\"Portraits](\"http:\/\/seebeyond.shop\/wp-content\/uploads\/2024\/10\/20241008_news_image02.jpg\"\/)
<\/p>\nEnormous influence<\/h2>\n
A nod toward AI\u2019s benefits\u2014and threats<\/h2>\n