If your jaw dropped as you watched the latest AI-generated video, your bank balance was saved from criminals by a fraud detection system, or your day was made a little easier because you were able to dictate a text message on the run, you have many scientists, mathematicians and engineers to thank.

But two names stand out for foundational contributions to the deep learning technology that makes those experiences possible: Princeton University physicist John Hopfield and University of Toronto computer scientist Geoffrey Hinton.

The two researchers were awarded the Nobel Prize in physics on Oct. 8, 2024, for their pioneering work in the field of artificial neural networks. Though artificial neural networks are modeled on biological neural networks, both researchers’ work drew on statistical physics, hence the prize in physics.

How a neuron computes

Artificial neural networks owe their origins to studies of biological neurons in living brains. In 1943, neurophysiologist Warren McCulloch and logician Walter Pitts proposed a simple model of how a neuron works. In the McCulloch-Pitts model, a neuron is connected to its neighboring neurons and can receive signals from them. It can then combine those signals to send signals to other neurons.

But there is a twist: It can weigh signals coming from different neighbors differently. Imagine that you are trying to decide whether to buy a new bestselling phone. You talk to your friends and ask them for their recommendations. A simple strategy is to collect all friend recommendations and decide to go along with whatever the majority says. For example, you ask three friends, Alice, Bob and Charlie, and they say yay, yay and nay, respectively. This leads you to a decision to buy the phone because you have two yays and one nay.

However, you might trust some friends more because they have in-depth knowledge of technical gadgets. So you might decide to give more weight to their recommendations. For example, if Charlie is very knowledgeable, you might count his nay three times and now your decision is to not buy the phone—two yays and three nays. If you’re unfortunate to have a friend whom you completely distrust in technical gadget matters, you might even assign them a negative weight. So their yay counts as a nay and their nay counts as a yay.

Once you’ve made your own decision about whether the new phone is a good choice, other friends can ask you for your recommendation. Similarly, in artificial and biological neural networks, neurons can aggregate signals from their neighbors and send a signal to other neurons. This capability leads to a key distinction: Is there a cycle in the network? For example, if I ask Alice, Bob and Charlie today, and tomorrow Alice asks me for my recommendation, then there is a cycle: from Alice to me, and from me back to Alice.

If the connections between neurons do not have a cycle, then computer scientists call it a feedforward neural network. The neurons in a feedforward network can be arranged in layers. The first layer consists of the inputs. The second layer receives its signals from the first layer and so on. The last layer represents the outputs of the network.

However, if there is a cycle in the network, computer scientists call it a recurrent neural network, and the arrangements of neurons can be more complicated than in feedforward neural networks.

Hopfield network

The initial inspiration for artificial neural networks came from biology, but soon other fields started to shape their development. These included logic, mathematics and physics. The physicist John Hopfield used ideas from physics to study a particular type of recurrent neural network, now called the Hopfield network. In particular, he studied their dynamics: What happens to the network over time?

Such dynamics are also important when information spreads through social networks. Everyone’s aware of memes going viral and echo chambers forming in online social networks. These are all collective phenomena that ultimately arise from simple information exchanges between people in the network.

Hopfield was a pioneer in using models from physics, especially those developed to study magnetism, to understand the dynamics of recurrent neural networks. He also showed that their dynamics can give such neural networks a form of memory.

Boltzmann machines and backpropagation

During the 1980s, Geoffrey Hinton, computational neurobiologist Terrence Sejnowski and others extended Hopfield’s ideas to create a new class of models called Boltzmann machines, named for the 19th-century physicist Ludwig Boltzmann. As the name implies, the design of these models is rooted in the statistical physics pioneered by Boltzmann. Unlike Hopfield networks that could store patterns and correct errors in patterns—like a spellchecker does—Boltzmann machines could generate new patterns, thereby planting the seeds of the modern generative AI revolution.

Hinton was also part of another breakthrough that happened in the 1980s: backpropagation. If you want artificial neural networks to do interesting tasks, you have to somehow choose the right weights for the connections between artificial neurons. Backpropagation is a key algorithm that makes it possible to select weights based on the performance of the network on a training dataset. However, it remained challenging to train artificial neural networks with many layers.

In the 2000s, Hinton and his co-workers cleverly used Boltzmann machines to train multilayer networks by first pretraining the network layer by layer and then using another fine-tuning algorithm on top of the pretrained network to further adjust the weights. Multilayered networks were rechristened deep networks, and the deep learning revolution had begun.

AI pays it back to physics

The Nobel Prize in physics shows how ideas from physics contributed to the rise of deep learning. Now deep learning has begun to pay its due back to physics by enabling accurate and fast simulations of systems ranging from molecules and materials all the way to the entire Earth’s climate.

By awarding the Nobel Prize in physics to Hopfield and Hinton, the prize committee has signaled its hope in humanity’s potential to use these advances to promote human well-being and to build a sustainable world.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

In the year 946 CE, the Changbaishan-Tianchi volcano, on the border between China and North Korea, erupted ferociously. The eruption released dozens of cubic kilometers of magma and triggered a massive flood from the lake atop the volcano’s summit, known today as Heaven Lake. Evidence of the flood can still be seen in the form of boulders and smaller rocks that washed down from the upper reaches of the volcano.

Changbaishan-Tianchi, known as Baekdu in Korean, could erupt again, so volcanologists want to understand the risks it poses. To investigate the catastrophic flood that followed the 946 eruption, Qin and team dug into the layered deposits from the volcano. Their work suggests that at least 1 cubic kilometer of water spilled from the volcano’s caldera, causing sediment to erode at rates as high as 34 meters per hour over about 3 hours. The paper is published in the journal Water Resources Research.

The researchers also concluded that the eruption consisted of two phases, with the flood occurring between the two. Other scientists have hypothesized that the flood gushed out in one instantaneous outburst after the eruption cracked the volcano’s rim, but this study’s authors found that scenario unrealistic because the sediment is not as widely spread as would be expected from one sudden burst.

The researchers suggest three alternative scenarios. In the first, the water simply overflowed the edge of the caldera in response to magma rising from below it. In the second, the volcano triggered an earthquake that collapsed the inner wall of the caldera into the lake, causing it to overflow. And in the third, precipitation prior to the event filled the caldera to capacity and weakened the crater rim, allowing the water to flow out.

Understanding ancient floods like the 946 CE event may help vulnerable populations prepare for future natural disasters, not just at Changbaishan-Tianchi but also at volcanoes around the world, the researchers wrote.

This story is republished courtesy of Eos, hosted by the American Geophysical Union. Read the original storyhere.

Gender biases around male and female roles and under-representation of female characters appeared in textbooks from around the world, with male-coded words appearing twice as often as female-coded words on average, according to a study published October 9, 2024 in the open-access journal PLOS ONE by Lee Crawfurd from the Center for Global Development, United Kingdom, and colleagues.

School textbooks play an important role in shaping norms and attitudes in students—one reason why controversy over textbook content is high in many countries today. In this study, Crawfurd and colleagues investigated how gender norms are depicted in textbooks around the world.

The authors used a particularly large corpus of textbooks to conduct their analysis: 1,255 publicly available online English-language school textbooks spanning subjects and grade levels from grades 4–13 from 34 countries downloaded over 2020–2022.

They compared textbook content with predefined lists of gendered nouns and pronouns (e.g. “Auntie/she/her/woman”) and investigated how often these gendered words were associated with key words used in previous studies relating to achievement, appearance, family, home, and work (e.g. “powerful/gorgeous/household/executive”) within the textbook. Finally, the authors compared their text analysis results with other measures of gender equality at the country level.

They found that on average, across the full sample of textbooks, there were more than twice as many occurrences of male words (178,142) as female words (82,113), though there was considerable variation between countries. After adjusting for book length, grade, and subject, the countries with the lowest representation of women and girls were Afghanistan, Pakistan, Sri Lanka, and South Sudan, where fewer than one in three gendered words were female.

Across all countries, the adjectives most likely to describe only female and not male characters included “married,” “beautiful,” “aged,” and “quiet.” Verbs for only female characters included “bake,” “cook,” and “sang.” The adjectives most likely to describe male and not female characters included “powerful,” “rich,” “wise,” “certain,” and “unable.” Verbs for only male characters included “rule,” “guide,” “sign,” and “order.”

Almost all of the individual achievement- and work-themed words showed a stronger association with male words than female words, and the individual appearance- and home-themed words showed a stronger association with female words than male words.

The authors note that countries with textbooks containing a greater number of female characters also had stronger GDPs and more legal rights for women compared to countries with less female representation, though this is only correlation and cannot speak to causation.

The authors also note there are several limitations to this work—their tool was not able to assess non-text items (such as images) and was not always correct at parsing names (though the authors used manual validation where possible), and the analysis reflects a binary view of gender illustrated in the textbooks.

Furthermore, this analysis is restricted to English language literature and therefore may not be generalizable to languages beyond English. However, the results suggest that combating gender biases in textbooks could potentially lead to real-world effects.

The authors add, “Our findings reveal a troubling reality: school books are perpetuating outdated gender stereotypes. Schools should broaden horizons not limit children’s potential. It’s crucial for policymakers and educators to address these disparities.”

Physicists share a common interest in understanding how the physical world works. For example, when a particle physicist breaks apart a particle into smaller pieces, they ask themselves: are those the smallest pieces we can find in nature?

For years, theoretical physicists have been limited to using classical computers, that is, computers that process information in 1s and 0s, as they explore the interactions between these small particles. Thanks to the power of quantum computers, which can encode numerous possible combinations of 1 and 0 simultaneously, physicists can create larger models that will potentially solve some of the most compelling puzzles in theoretical physics.

However, theoretical researchers need access to novel computing technologies and a strong expert network for this new scientific endeavor. That’s where the Quantum Systems Accelerator‘s catalyzing energy comes in.

Since 2020, the Quantum Systems Accelerator (QSA), a U.S. Department of Energy (DOE) National Quantum Information Science (QIS) Research Center, has developed advanced quantum prototypes across major technologies to facilitate groundbreaking research in fundamental physics into the quantum world.

Led by Lawrence Berkeley National Laboratory (Berkeley Lab) with Sandia National Laboratories as the lead partner, QSA brings together an ecosystem of 15 institutions in North America. With over 60 principal investigators, 130 staff, 91 postdocs, and 139 students, QSA advances national particle physics research by co-designing across institutions.

Two teams of QSA researchers across partner institutions have recently made headway toward understanding more about the framework of the subatomic quantum world with quantum devices.

One collaboration between Berkeley Lab and UC Berkeley, led by researchers Anthony Ciavarella and Christian W. Bauer, examined the dynamics between quarks and gluons, the tiny particles that comprise every atom’s core.

Their published results build on previous studies supported by the QSA. The work is published on the arXiv preprint server.

They created a model to understand how these particles were held together (their nuclear force). They also wondered if a quantum computer could help them understand what would happen if quarks and gluons collided and broke into more pairs of particles.

Using a theoretical understanding of these subatomic relationships, researchers built a model that mapped gluons onto a lattice for a quantum computer, marking the first time that anyone in this field has created such a model.

Ciavarella and Bauer’s model will allow experimentalists to run simulations and compare these results to their experimental data. It will also help theorists take experimental results and, working backward, create a more robust theoretical underpinning for nuclear models.

“If you want to search for new physics events where you produce Higgs bosons,” Ciavarella explains, “what you measure in [a particle accelerator] looks very similar to an event where you just have quarks and gluons radiating.”

One day, simulations of Ciavarella and Bauer’s model can be used by experimentalists to reduce the size of background noise in their experiments and interpret results more clearly.

Thanks to QSA, Ciavarella and Bauer can soon run their model on a quantum computer. While the researchers worked on their paper, QSA staff helped them connect with researchers at Harvard University and Massachusetts Institute of Technology’s Center for Ultracold Atoms. The Harvard group has been building the type of hardware capable of running Ciavarella and Bauer’s proposed simulations.

Ciavarella and Bauer are currently working with their colleagues to start running simulations this fall. “Once we can start doing these full simulations,” Ciavarella explains, “that should open up a whole new set of experimental observables that we’ve never really worked with because we’ve never been able to make predictions for them.”

Another multi-institutional QSA team has been investigating how one might use quantum processors to investigate models inspired by quantum chromodynamics (QCD). Using quantum processors to study the theories of the Standard Model in particle physics is a long-term goal of quantum simulation.

While current quantum processors have not yet reached this capability, researchers have begun considering the required steps. Elizabeth Bennewitz, a researcher and graduate student at the University of Maryland, worked with other team members to develop a protocol for understanding the interactions between two mesons composed of tightly bound quarks.

Bennewitz, the lead author of a second article, released as a preprint on the arXiv server, is a DOE Computing Sciences Graduate (CSGF) Fellow and was formerly a Berkeley Lab intern. As a theorist, Bennewitz is interested in exploring what happens when mesons collide.

However, running her simulations on a classical computer limits the number of interacting particles that theorists can study in a model. She worked with her collaborators to create a protocol that lists the necessary steps to build a model that would simulate the movement of mesons and what happens when those mesons collide.

Their protocol will help experimentalists fill in the gaps as they try to create working models they can implement on quantum devices. Their work showed the types of meson scattering that might show up in a quantum device. By recognizing the byproducts of meson scattering, experimentalists can better interpret their results, distinguishing between irrelevant data (noise) and essential data (signal).

This study represents one step towards using quantum processors as powerful tools to study quantum phenomena in particle physics and potentially extend the standard model of physics.

“Probing physics in these new ways will hopefully lead to discoveries in the physics side of things but can also influence other areas like material sciences and chemistry,” said Bennewitz.

Like Ciavarella and Bauer, Bennewitz’s team benefited from QSA’s support. Bennewitz has been struck by the interdisciplinary nature of QSA’s partners and its ability to connect researchers across different areas with relevant collaborators.

“Being able to work really closely with experts in quantum simulation, high-energy and nuclear physics, and experimental quantum devices was one of the reasons this work was possible,” said Bennewitz.

Both QSA teams voiced how working with new colleagues helped them consider new dimensions of planning an experiment.

“Something that might be obvious to them when they’re just looking at the experiments up in the lab might not be obvious to me because I’ve never worked directly in the lab on a device myself and vice versa, with how to map problems on their computers. It’s a way to learn a lot about interesting things,” said Ciavarella.

The work of these QSA researchers has underscored the importance of quantum simulators as a tool for investigating the almost invisible subatomic world of particle physics.

By bringing together experimentalists and theorists, QSA continues to facilitate research by streamlining the design of quantum devices and engineering solutions. Moreover, QSA provides early career researchers like Ciavarella and Bennewitz with the infrastructure and support to conduct their studies.

“There’s a lot of evidence that studying these models on quantum computers and quantum simulators should be more powerful for certain types of problems. With quantum simulations, we can unveil and probe particle physics in ways that we haven’t been able to before and hopefully find new physics,” said Bennewitz.

QSA director Bert de Jong believes that, in the coming decade, even more researchers like Bennewitz will embrace quantum computers as an investigative tool.

“We have a high-energy physics team in the QSA team that would like to build better theories to understand the foundational building blocks of the universe,” concluded de Jong.