Toggle light / dark theme

Optimized magnetic pulses could cut memory switching energy by several orders of magnitude

Information and communication technologies (ICTs) driven by artificial intelligence (AI) are generating data at an unprecedented rate. Every internet search, AI-generated image, recommendation, scientific simulation and large language model creates and processes enormous amounts of information that must be stored, transferred and analyzed. As AI continues to expand across every sector of society, global demand for data storage and computing is rising dramatically.

This rapid growth comes at a significant cost: energy consumption. Data centers already consume vast amounts of electricity, and demand is expected to increase sharply over the coming decades. Without major technological advances, ICTs could account for a substantial fraction of global electricity use and carbon emissions, making energy-efficient computing one of the defining scientific challenges of our time.

Google’s Westinghouse Bet

In the article “Google’s Westinghouse Bet” published on Asimov’s Addendum, Tim O’Reilly presents an alternate thesis regarding the organizational shifts, talent departures, and strategic shakeups at Google DeepMind.

While industry reports (such as a analysis by SemiAnalysis) interpreted the DeepMind restructuring as evidence that Google is falling behind in the “frontier model race” against rivals like OpenAI and Anthropic, O’Reilly argues that Google is not losing the race—it is choosing to run a different one.


SemiAnalysis thinks the DeepMind shakeup means Google is losing the AI race. It might be that Google is choosing to run a different race.

Technological Convergence — the DARPA Lift Challenge

The DARPA Lift Challenge aims to shatter the heavy-lift bottleneck by seeking novel drone designs capable of carrying payloads more than four times their weight. This would revolutionize how we use drones across all sectors.

As military missions become more complicated, warfighters need more capable drones to use across diverse scenarios. The same applies to infrastructure inspection, package delivery, disaster response, and other civilian applications.

Current multirotor drones, also known as unmanned aircraft systems (UAS), are simple, affordable, and easy to operate. But their payload-to-weight ratio is low, typically 1:1 or less.

DARPA Challenges accelerate technological breakthroughs by focusing the ingenuity of teams of innovators and giving wild ideas a place to thrive.

Follow along at www.darpaliftchallenge.com

The Great Progression: Peter Leyden on AI and the Next 25 Years

80 years, America tears itself down and rebuilds. 1776. 1865. 1945. And now, according to Peter Leyden, 2025.

A couple of months ago I sat down with Peter for SingularityFM. He’s the guy who co-wrote the 1997 “Long Boom” cover story for WIRED, and he’s spent three decades trying to map where this civilization is actually headed. His new book, The Great Progression (HarperCollins), makes an audacious claim:

AI, clean energy, and bioengineering are converging into a change on the scale of the Enlightenment. And Trump, in his framing, isn’t the future. He’s the wrecking ball that clears the ground for whatever comes next.

I didn’t let him off easy. I pushed him on the human cost every one of these reinventions has demanded in the past. I pushed him on the myth of golden ages, using Ada Palmer’s work on the Renaissance. I pushed him on Denmark and Norway ripping devices back out of classrooms after being early adopters. He had an answer for all three. Whether they hold up is for you to decide.

What we get into:

The email from Kevin Kelly that pulled Peter into WIRED’s founding years What the Long Boom thesis got right, and what it missed Why he calls Trump a wrecking ball, not a builder The coming “abundance politics” coalition The strange pessimism about AI unique to English-speaking countries How writing this book with AI made him 2 to 3x more productive.

User awareness in frontier models

Despite no longer “thinking out loud” about who the user is, the underlying behavioral shifts remained just as strong. This means user-aware adaptations are becoming increasingly covert and difficult to detect simply by monitoring an AI’s internal reasoning logs.


Our results suggest that user awareness is a meaningful and understudied form of situational awareness in frontier language models. The effect is significant and robust, hard to detect, and can persist even without reasoning. To be clear, these effects say nothing about the individuals named: we find no evidence that any of them sought this differential treatment, and the behavior almost certainly emerged as an unintended artifact of training rather than by anyone’s design.

The most immediate implication of our findings is for alignment evaluations. Models can already recognize particular people and organizations and behave differently for them, so results built on synthetic names and companies may not transfer to deployments involving real, high-stakes identities. The behaviors we observe today are relatively benign, but they may be precursors of more concerning conditional behaviors.

An important limitation of this work is that we are mostly only measuring fixed-prompt propensities here rather than actual performance in critical tasks. We hope to broaden and automate the investigation to a larger degree with our ongoing efforts.

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.

Negative imaginary theory moves from math niche to robots, aircraft and nanodevices

Over the past two decades, a powerful but highly specialized branch of control engineering—known as negative imaginary (NI) systems theory—has quietly evolved into a key tool for stabilizing complex, vibration-prone systems, from flexible structures to advanced robotics.

Now, a new study reveals how fast and in what direction this field is growing. By analyzing more than 400 scientific publications from 2004 to 2024 across the world’s leading academic databases, the researchers uncovered a clear trend: NI systems are no longer just a theoretical concept. The work is published in the International Journal of Systems Science.

The field is expanding rapidly, with increasing global participation and a strong shift toward real-world applications such as multi-robot coordination, aerospace systems and nanotechnology.