A new accelerating wave equation links changing light speeds to relativity, momentum conservation and the arrow of time.
Category: information science
LLMs show language does not describe reality
We tend to think of language, perception and thought as representing the world, pointing towards and mapping reality. But AI’s large language models suggest that this isn’t how language works, argues cognitive scientist Elan Barenholtz. These models prove that language and imagery can produce coherent continuations without ever grounding themselves in external reality. This “autogenerative” capacity was always present in language, awaiting discovery. And, Barenholtz argues, it is now the best explanation we have of how language and perception work in humans. Meaning arises not by establishing facts about the world, but rather through language’s generative role in producing further language, imagery, and, ultimately, coordinated human action.
Imagine that archaeologists unearth clay tablets from an ancient civilization, long lost to the world. There are no bilingual texts, no known descendants of the civilization, nothing to anchor a translation of the tablets. They seem to display nothing more than rows of arbitrary squiggles. Now imagine that someone claims to have decoded the squiggles. “These patterns,” they assure us, “are self-predicting. The sequence of symbols in one part of the tablet is mathematically sufficient to derive what will appear in another part.” And, indeed, they produce an algorithm which correctly predicts the text on the right side of each tablet based on the text on the left side.
The finding that the symbols contain this predictive structure would be an extraordinary insight. But we still might ask, “what do the symbols mean?” Now replace the tablets with the digital corpus we call the internet, and the algorithm with a large language model. The civilization is ours. And the question of meaning is ours too.
Physics In History (@PhysInHistory) on X
The Longest Equation in Physics The model Lagrangian is a mathematical expression that summarizes the Standard Model of particle physics, which is the most successful theory of the fundamental interactions between elementary particles. It is composed of four different parts, each describing a different aspect of the Standard Model. The model Lagrangian is written in a compact notation that uses symbols and operators from quantum field theory, such as covariant derivatives, field strength tensors, Dirac matrices, and gauge group generators. It also uses various constants and parameters that are determined by experiments, such as coupling constants, masses, and mixing angles. It is one of the longest equations in physics because it contains many terms and factors that account for all the possible interactions and symmetries of the Standard Model. It was transcribed by Thomas Gutierrez who derived it from Martinus Veltman’s Diagrammatica: The Path to Feynman Diagrams.
Molecular orbitals imaged in 3D, opening path to femtosecond videos
One of the most famous and intriguing results of quantum mechanics is the finding that fundamental particles, such as electrons, cannot be pinned down to one single location. Instead, a particle is described by its “wavefunction,” which allows researchers to derive probability distributions—a sort of mathematical map that shows the possibilities—of fundamental properties such as its position and momentum. In particular, the electron wavefunctions within a molecule, known as “molecular orbitals,” carry information about how the molecule interacts with its surroundings. For example, they show how it may absorb light or how a chemical reaction might take place.
As a consequence, knowledge of the complete three-dimensional wavefunction is highly desirable, but imaging the wavefunction has proven to be a major experimental challenge. An interdisciplinary research team at the University of Göttingen has now managed to image the three-dimensional wavefunction of a nanometer-sized organic molecule. They overcame the limitations by combining state-of-the-art photoelectron spectroscopy with powerful mathematical algorithms. The results are published in Nature Communications.
Noninvasive AI-based system translates brain signals into written text
Some physical injuries and neurological conditions can temporarily or permanently impair movement, leaving some people unable to speak, type on keyboards or use electronic devices. Brain-computer interfaces (BCIs), systems that can decode brain activity patterns and convert them into computer commands or written text, could be of great value for paralyzed patients.
Despite their potential, most of the best-performing BCIs developed to date require patients to undergo invasive surgical procedures. These systems typically rely on small sensors that need to be implanted on or within the brain and can detect electrical signals associated with neural activity.
Researchers at Meta artificial intelligence (AI), Université PSL and Hospital Foundation Adolphe de Rothschild recently introduced a noninvasive brain activity-to-text approach that does not require surgical procedures. Their proposed approach, presented in Nature Neuroscience, combines a new deep learning algorithm with electroencephalography (EEG) or magnetoencephalography (MEG) recordings.
New quantum encryption method prevents ciphertext from being cloned
Digital security currently relies on difficult equations to protect data. For example, when you use a credit card online, the information is locked inside a math problem that would take a modern computer thousands of years to solve. However, if someone builds a powerful enough computer, that security breaks.
To make systems safer, there is a major shift toward quantum security. This is where the unbreakable laws of quantum physics can be used to protect data instead.
Hidden prompts can plant false memories in AI agents, researchers warn
Large language models (LLMs), the computational algorithms underpinning ChatGPT, Gemini and other artificial intelligence (AI)-powered conversational platforms, are now widely used worldwide. These models can rapidly answer questions, source information online, assist users with specific tasks and produce text tailored for specific purposes.
Over the past few years, computer scientists have introduced a wide range of LLMs, some of which can also complete tasks autonomously and take actions on a user’s behalf, for instance, answering messages, scheduling online appointments or updating programming code. More recently, they have also provided many of these models with memory, as this allows them to tailor responses around a user’s typical preferences or earlier requests without repeatedly receiving the same information.
While memory-enhanced AI agents have notable advantages and could better meet the needs of individual users, they also pose new security risks. Researchers at New Mexico State University recently described a cyberattack that secretly poisons an AI agent’s long-term memory, which they dubbed GhostWriter. Their paper, published on the preprint arXivserver, introduces two promising strategies that could help prevent or limit the risk of this attack without adversely affecting an AI agent’s performance.
Game-engine forests train drone AI to count trees with far less labeling
A drone swoops low over an alpine forest. It climbs suddenly to follow the contours of the sharply rising landscape. Pulses from its lidar—a laser mapping instrument—rapidly scan the trees below.
The forest, however, isn’t real. In fact, the entire landscape is a synthetic rendering created by University of Cambridge researchers to teach algorithms how to see trees.
The ability to recognize an individual tree in the forest canopy is essential for calculating how forests grow, how they respond to climate change and how much carbon they store. Until now, researchers developing forest vision systems would painstakingly trace the outlines of thousands of trees to provide the system with sufficient training data, a process that can take weeks.