Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Get the latest international news and world events from around the world.

Log in for authorized contributors

View a PDF of the paper titled When Models Manipulate Manifolds: The Geometry of a Counting Task, by Wes Gurnee and 6 other authors

When you look at text, you subconsciously track how much space remains on each line. If you’re writing “Happy Birthday” and “Birthday” won’t fit, your brain automatically moves it to the next line. You don’t calculate this—you *see* it. But AI models don’t have eyes. They receive only sequences of numbers (tokens) and must somehow develop a sense of visual space from scratch.

Inside your brain, “place cells” help you navigate physical space by firing when you’re in specific locations. Remarkably, Claude develops something strikingly similar. The researchers found that the model represents character counts using low-dimensional curved manifolds—mathematical shapes that are discretized by sparse feature families, much like how biological place cells divide space into discrete firing zones.

The researchers validated their findings through causal interventions—essentially “knocking out” specific neurons to see if the model’s counting ability broke in predictable ways. They even discovered visual illusions—carefully crafted character sequences that trick the model’s counting mechanism, much like optical illusions fool human vision.

2. Attention mechanisms are geometric engines: The “attention heads” that power modern AI don’t just connect related words—they perform sophisticated geometric transformations on internal representations.

1. What other “sensory” capabilities have models developed implicitly? Can AI develop senses we don’t have names for?

Language models can perceive visual properties of text despite receiving only sequences of tokens-we mechanistically investigate how Claude 3.5 Haiku accomplishes one such task: linebreaking in fixed-width text. We find that character counts are represented on low-dimensional curved manifolds discretized by sparse feature families, analogous to biological place cells. Accurate predictions emerge from a sequence of geometric transformations: token lengths are accumulated into character count manifolds, attention heads twist these manifolds to estimate distance to the line boundary, and the decision to break the line is enabled by arranging estimates orthogonally to create a linear decision boundary. We validate our findings through causal interventions and discover visual illusions—character sequences that hijack the counting mechanism.

Multi-omic analysis of guided and unguided forebrain organoids reveals differences in cellular composition and metabolic profiles

The differences in guided and unguided forebrain organoids.

The differences arising from guided or unguided differentiation of human forebrain organoids is not well understood.

The researchers perform a multiomic analysis of forebrain organoids generated by these two key methods.

The researchers demonstrate that guided forebrain organoids contained a larger proportion of neurons, including GABAergic interneurons, whereas the unguided organoids contained significantly more choroid plexus, radial glia, and astrocytes at later stages.

They also show increased levels of oxidative phosphorylation and fatty acid β-oxidation in the unguided forebrain organoids and a higher reliance on glycolysis in the guided forebrain organoids. sciencenewshighlights ScienceMission https://sciencemission.com/guided-and-unguided-forebrain-organoids

Øhlenschlæger et al. perform a multi-omic analysis of forebrain organoids generated by two key methods, guided and unguided differentiation. They document significant differences in the cell type composition and metabolic profiles of the two forebrain organoid types, providing a resource and methodological guide for the neural organoid field.

Continue reading “Multi-omic analysis of guided and unguided forebrain organoids reveals differences in cellular composition and metabolic profiles” | >

Humidity-resistant hydrogen sensor can improve safety in large-scale clean energy

Wherever hydrogen is present, safety sensors are required to detect leaks and prevent the formation of flammable oxyhydrogen gas when hydrogen is mixed with air. It is therefore a challenge that today’s sensors do not work optimally in humid environments—because where there is hydrogen, there is very often humidity. Now, researchers at Chalmers University of Technology, Sweden, are presenting a new sensor that is well suited to humid environments—and actually performs better the more humid it gets.

“The performance of a hydrogen gas sensor can vary dramatically from environment to environment, and humidity is an important factor. An issue today is that many sensors become slower or perform less effectively in humid environments. When we tested our new sensor concept, we discovered that the more we increased the humidity, the stronger the response to hydrogen became. It took us a while to really understand how this could be possible,” says Chalmers doctoral student Athanasios Theodoridis, who is the lead author of the article published in the journal ACS Sensors.

Hydrogen is an increasingly important energy carrier in the transport sector and is used as a raw material in the chemical industry or for green steel manufacturing. In addition to water being constantly present in ambient air, it is also formed when hydrogen reacts with oxygen to generate energy, for example, in a fuel cell that can be used in hydrogen-powered vehicles and ships. Furthermore, fuel cells themselves require water to prevent the membranes that separate oxygen and hydrogen inside them from drying out.

How scientists are trying to use AI to unlock the human mind

Compared with conventional psychological models, which use simple math equations, Centaur did a far better job of predicting behavior. Accurate predictions of how humans respond in psychology experiments are valuable in and of themselves: For example, scientists could use Centaur to pilot their experiments on a computer before recruiting, and paying, human participants. In their paper, however, the researchers propose that Centaur could be more than just a prediction machine. By interrogating the mechanisms that allow Centaur to effectively replicate human behavior, they argue, scientists could develop new theories about the inner workings of the mind.

But some psychologists doubt whether Centaur can tell us much about the mind at all. Sure, it’s better than conventional psychological models at predicting how humans behave—but it also has a billion times more parameters. And just because a model behaves like a human on the outside doesn’t mean that it functions like one on the inside. Olivia Guest, an assistant professor of computational cognitive science at Radboud University in the Netherlands, compares Centaur to a calculator, which can effectively predict the response a math whiz will give when asked to add two numbers. “I don’t know what you would learn about human addition by studying a calculator,” she says.

Even if Centaur does capture something important about human psychology, scientists may struggle to extract any insight from the model’s millions of neurons. Though AI researchers are working hard to figure out how large language models work, they’ve barely managed to crack open the black box. Understanding an enormous neural-network model of the human mind may not prove much easier than understanding the thing itself.

Water molecules actively reshape chiral catalyst structure, research shows

Researchers have analyzed the stepwise hydration of prolinol, a molecule widely used as a catalyst and as a building block in chemical synthesis. The study shows that just a few water molecules can completely change the preferred structure of prolinol. The research is published in the Journal of the American Chemical Society.

Physical chemistry applies the principles and concepts of physics to understand the basics of chemistry and explain how and why transformations of matter take place on a molecular level. One of the branches of this field focuses on understanding how molecules change in the course of a chemical reaction or process.

Understanding the interactions of chiral molecules with water is crucial, given the central role that water plays in chemical and biological processes. Chiral molecules are those that, despite comprising the same atoms, cannot be superimposed on their mirror image in a way similar to what happens with right and left hands or a pair of shoes.

Quantum Teleportation Was Performed Over The Internet For The First Time

Scientists achieved the ‘impossible’ in 2024, teleporting a quantum state through more than 30 kilometers amid a torrent of internet traffic.

In 2024, a quantum state of light was successfully teleported through more than 30 kilometers (around 18 miles) of fiber optic cable amid a torrent of internet traffic – a feat of engineering once considered impossible.

The impressive demonstration by researchers in the US may not help you beam to work to beat the morning traffic, or download your favorite cat videos faster.

However, the ability to teleport quantum states through existing infrastructure represents a monumental step towards achieving a quantum-connected computing network, enhanced encryption, or powerful new methods of sensing.

Polar weather on Jupiter and Saturn hints at the planets’ interior details

Over the years, passing spacecraft have observed mystifying weather patterns at the poles of Jupiter and Saturn. The two planets host very different types of polar vortices, which are huge atmospheric whirlpools that rotate over a planet’s polar region. On Saturn, a single massive polar vortex appears to cap the north pole in a curiously hexagonal shape, while on Jupiter, a central polar vortex is surrounded by eight smaller vortices, like a pan of swirling cinnamon rolls.

Given that both planets are similar in many ways—they are roughly the same size and made from the same gaseous elements—the stark difference in their polar weather patterns has been a longstanding mystery.

Now, MIT scientists have identified a possible explanation for how the two different systems may have evolved. Their findings could help scientists understand not only the planets’ surface weather patterns, but also what might lie beneath the clouds, deep within their interiors.

Dyson Strawberry Farming: 5,127 Prototypes to 250% Yields

When James Dyson built his 5,127th prototype of a bagless vacuum cleaner, he had no idea that the same relentless engineering philosophy would one day transform him into Britain’s largest farmer. Today, Dyson strawberry farming represents one of the most ambitious applications of high-tech innovation to agriculture ever attempted in the United Kingdom.

The numbers tell an extraordinary story. After spending five years and creating over five thousand prototypes to perfect a single vacuum cleaner design, Dyson has now invested £140 million into a farming operation spanning 36,000 acres across five English counties. At the heart of this agricultural empire sits a 26-acre glasshouse in Lincolnshire, home to 1.25 million strawberry plants and technology that has increased yields by 250% compared to traditional farming methods.

This isn’t farming as your grandparents would recognize it. Inside Dyson’s facility, massive 5.5-meter “ferris wheel” structures rotate strawberry plants through optimal sunlight positions. Sixteen robotic arms delicately harvest ripe fruit using computer vision. UV-emitting robots patrol the aisles at night, destroying mould without chemicals. And all of it runs on renewable energy generated from an adjacent anaerobic digester.

Lung Sparing Surgery Is Possible for Patients With COPD

Patients with COPD may be especially hesitant about lung surgery when facing lung cancer, but surgery remains the most effective treatment for non–small cell lung cancer when feasible, according to experts. For those who are not surgical candidates, less invasive options exist, with stereotactic body radiation therapy (SBRT) the standard noninvasive treatment for early-stage disease, delivering targeted high-dose radiation over three to five sessions while sparing healthy tissue.

Some individuals with COPD and lung cancer may be eligible for less invasive options.

/* */