Posts

Platonic representation hypothesis

"The team’s hypothesis gets its name from a 2,400-year-old allegory by the Greek philosopher Plato. In it, prisoners trapped inside a cave perceive the world only through shadows cast by outside objects. Plato maintained that we’re all like those unfortunate prisoners. The objects we encounter in everyday life, in his view, are pale shadows of ideal forms  that reside in some transcendent realm beyond the reach of the senses. "The Platonic representation hypothesis is less abstract. In this version of the metaphor, what’s outside the cave is the real world, and it casts machine-readable shadows in the form of streams of data. AI models are the prisoners. The MIT team’s claim is that very different models, exposed only to the data streams, are beginning to converge on a shared Platonic representation  of the world behind the data. "'Why do the language model and the vision model align? Because they’re both shadows of the same world,' said Phillip Isola, the senior...
Post a Comment
Read more

Digital asbestos removal

"I wouldn't bank on getting work in a fanciful trade that literally can't be imagined because our consciousnesses haven't so altered by AI that they've acquired the capacity to conceptualize of these new modes of work. "But if you are looking for a job that AI will definitely create, by the millions, I have a suggestion: digital asbestos removal. "For if AI code —written at 10,000 times the speed of any human coder, designed to work well, but not to fail gracefully —is the digital asbestos we're filling our walls with, then our descendants will spend generations digging that asbestos out of the walls.  "There will be plenty of work fixing the things that we broke thanks to the most dangerous AI psychosis of all —the hallucinatory belief that writing code  is the same thing as software engineering . "At the rate we're going, we'll have full employment for generations of [digital] asbestos removers."
Post a Comment
Read more

Agent or spy?

Image
 
Post a Comment
Read more

Huang betting on bots

"Nvidia is touting its latest robotics advancements at CES 2026 as the broader tech industry begins a large-scale effort to bring humanoid robots to life. "During the company's keynote on Monday, CEO Jensen Huang revealed that firms ranging from Boston Dynamics and Caterpillar to LG Electronics and NEURA Robotics are using Nvidia's robotics technologies to develop and power their various bots. "Nvidia has claimed that physical AI could revolutionize the $50 trillion manufacturing and logistics industries, and the company wants to be at the center of it all. "During CES, Nvidia revealed a variety of new AI models to help train robots to interact with the world around them, as well as the hardware necessary to power their digital brains."
Post a Comment
Read more

Immanent by 2034

"A leading artificial intelligence expert has rolled back his timeline for AI doom, saying it will take longer than he initially predicted for AI systems to be able to code autonomously and thus speed their own development toward superintelligence. "Daniel Kokotajlo, a former employee of OpenAI, sparked an energetic debate in April by releasing AI 2027 , a scenario that envisions unchecked AI development leading to the creation of a superintelligence, which —after outfoxing world leaders —destroys humanity.    "The scenario rapidly won admirers and detractors. The US vice-president, JD Vance, appeared to reference AI 2027 in an interview last May when discussing the US’s artificial intelligence arms race with China. "Gary Marcus, an emeritus professor of neuroscience at New York University, called the piece a 'work of fiction' and various of its conclusions 'pure science fiction mumbo jumbo.' "In their update, Kokotajlo and his co-authors revise...
Post a Comment
Read more

Vulnerabilities

"AI systems embed cultural and developmental assumptions at every stage of their lifecycle. Training data reflects dominant languages, economic conditions, social norms, and historical records.  Design choices encode expectations about infrastructure, behavior, and values. "These assumptions affect system accuracy and safety.  Language models perform best in widely represented languages and lose reliability in under-resourced ones. Vision and decision systems trained in industrialized environments misread behavior in regions with different traffic patterns, social customs, or public infrastructure.  "These gaps increase error rates and create uneven exposure to harm . "From a cybersecurity perspective, these weaknesses resemble systemic vulnerabilities . They widen the attack surface by producing predictable failure modes across regions and user groups."
Post a Comment
Read more

Verbal insight

"In insight problem solving solutions with AHA! experience have been assumed to be the consequence of restructuring of a problem which usually takes place shortly before the solution. "However, evidence from priming studies suggests that solutions with AHA! are not spontaneously generated during the solution process but already relate to prior subliminal processing.  "We test this hypothesis by conducting an fMRI study using a modified compound remote associates paradigm which incorporates semantic priming. We observe stronger brain activity in bilateral anterior insulae already shortly after trial onset in problems that were later solved with than without AHA!.  "This early activity was independent of semantic priming but may be related to other lexical properties of attended words helping to reduce the amount of solutions to look for.  "In contrast, there was more brain activity in bilateral anterior insulae during solutions that were solved without than with...
Post a Comment
Read more

Vitruvius, DQI, Medel, and Catherine

Vitruvius "He states that all buildings should have three attributes: firmitas, utilitas, and venustas (strength, utility, and beauty), principles reflected in much Ancient Roman architecture.  "His discussion of perfect proportion in architecture and the human body led to the famous Renaissance drawing of the Vitruvian Man by Leonardo da Vinci. "In the Roman conception, architecture needed to take into account everything touching on the physical and intellectual life of man and his surroundings. DQI "There are [nowadays] three models of design quality indicator : DQI which is applicable to all building types DQI for schools which is applicable to school buildings. This model of DQI is being used on all current school projects in the UK and forms part of the Department for Children, Schools and Families 'Minimum Design Standard' for new school buildings. DQI for health buildings which was released in beta format in June 2012 on the DQI website. Elena Med...
Post a Comment
Read more

Uncovering compute

Charting America’s datacenters with open source intelligence isn’t a perfect practice. "'In the discovery phase, some data centers will be so obscure that we won’t find news, rumors, or existing databases mentioning them. While larger data centers are more likely to be reported due to their significance and physical footprint, there are many smaller data centers (<100 MW) that could add up to significant levels of AI compute,' Epoch AI said. "But Epoch AI continues to expand its toolset and look through more satellite imagery with the goal of mapping Big Tech’s newest project.  "The goal is to cast light into the darkness.  "'Even if we have a perfect analysis of a data center, we may still be in the dark about who uses it, and how much they use,' Epoch AI’s website said. 'AI companies like OpenAI and Anthropic make deals with hyperscalers such as Oracle and Amazon to rent compute, but the arrangement for any given data center is sometimes sec...
Post a Comment
Read more

Vernon

" Vernon is becoming an unlikely hub for Southern California’s AI infrastructure. "'The race toward superintelligence is reshaping infrastructure requirements across every industry,' said Michael Wall, executive vice president at Prime Data Centers, which built the Vernon data center. 'We’re working to give businesses the foundation they need to build and deploy the next generation of AI models —faster, more efficiently and at massive scale.' "Prime and other developers, including Goodman Group, CoreSite, and Digital Realty, are planning hundreds of megawatts of new data center capacity in Vernon, said Darren Eades, managing director of JLL, who specializes in data centers."
Post a Comment
Read more

Synthetic everything

"Instagram's top exec Adam Mosseri has made it clear that he expects AI content to overtake non-AI imagery and the significant implications that shift has for its creators and photographers. "Mosseri shared the thoughts in a lengthy post about the broader trends he expects to shape Instagram in 2026.  "And he offered a notably candid assessment on how AI is upending the platform. 'Everything that made creators matter —the ability to be real, to connect, to have a voice that couldn’t be faked —is now suddenly accessible to anyone with the right tools,' he wrote. 'The feeds are starting to fill up with synthetic everything.' "But Mosseri doesn't seem particularly concerned by this shift.  "He says that there is a lot of amazing AI content  and that the platform may need to rethink its approach to labeling such imagery by fingerprinting real media, not just chasing fake ."
Post a Comment
Read more

Foreign tech workers avoiding US

"Major tech firms, including Amazon, Microsoft, and Google, have reportedly urged overseas staff to return to the US quickly while simultaneously warning them to limit dependents’ travel. "At the same time, all these top tech companies, and many more, have been laying people off.  "Even if you can trace your ancestry back to the Mayflower, we’re living in a time of tech job insecurity.  "While the US raises barriers, rival tech hubs are pitching themselves as open and predictable.  "Canada, Europe, and parts of Asia are marketing fast‑track visas and remote‑work‑friendly policies aimed squarely at the engineers, founders, and researchers who now view the US as too much trouble for too little certainty."
Post a Comment
Read more

An 11-qubit atom processor in silicon

"An important requirement for scaling up is the ability to extend high-fidelity entanglement non-locally across several spin registers. "Here we address this challenge with an 11-qubit atom processor composed of two multi-nuclear spin registers that are linked by means of electron exchange interaction.  "Through the advancement of calibration and control protocols, we achieve single-qubit and multi-qubit gates with all fidelities ranging from 99.10% to 99.99%.  "By entangling all combinations of local and non-local nuclear-spin pairs, we map out the performance of the processor and achieve state-of-the-art Bell-state fidelities of up to 99.5%.  "We then generate Greenberger–Horne–Zeilinger (GHZ) states with an increasing number of qubits and show entanglement of up to eight nuclear spins. "By establishing high-fidelity operation across interconnected nuclear spin registers, we realize a key milestone towards fault-tolerant quantum computation with atom pro...
Post a Comment
Read more

Rust never sleeps

"The programming world is at a transition point. For system programming, Rust and AI are replacing C and manual coding. "C is prone to memory errors, accounting for about 70% of all operating system security holes. Memory-safe Rust prevents these problems. As Microsoft stated in 2019, 'What separates Rust from C and C++ is its strong safety guarantees.' That's not to say you still can't screw up with Rust. You can. The first Common Vulnerabilities and Exposures (CVE) security bug has also been identified: the Rust Android Binder driver bug, CVE-2025-68260.  "Meanwhile, over in Windows land, in early 2025, Check Point Research found a bug in the Rust-based Graphics Device Interface (GDI) component in Windows 11 24H2. "Linux developers appreciate that AI can make their work easier. However, as Torvalds said, 90% of today's AI industry is hype. He's also warned that using AI to generate serious, long-lived production code can be a horrible idea ...
Post a Comment
Read more

Combustion

"Around a million years ago, at the very beginning of the biggest disruption to life’s energetic capabilities since the emergence of cyanobacteria, humans learned to harness a new kind of energy source: combustion .  While other animals are limited by the metabolic energy they can harvest biologically, fire —an explosion of energy released as fuel reacts with oxygen —gave our ancestors an external energy source that was orders of magnitude more powerful. "Our ancestors cooked food, using fire as a form of digestion to assist the work of teeth, jaws and guts on tough, raw foods.  "Fire kept our ancestors warm without fur, scared off predators and cleared landscapes for hunting.  "One evolutionary response to this extracorporeal bounty of energy was a growth in brain size and specialization, including the development of a bigger prefrontal cortex.  "As a consequence, humans are smart, cooperative and exceptionally social. Those traits made us capable of a remarka...
Post a Comment
Read more

Hopfield networks to create

"Krotov and his colleagues recently developed a new deep learning architecture called the energy transformer. "Typical AI architectures are usually found by trial and error. But Krotov thinks energy transformers could be designed more intentionally with a specific energy landscape in mind, like a more complex take on a Hopfield network. "Though Hopfield networks were originally designed to remember, researchers are now exploring how they can be used to create.  "Image generators such as Midjourney are powered by diffusion models , which are themselves inspired by the physics of diffusion. To train them, researchers add noise to the training data —say, pictures of cats —and then teach the model to remove the noise. That’s a lot like what a Hopfield network does, except instead of always landing on the same cat picture, a diffusion model removes non-cat  noise from a noisy, random starting state to produce a new cat. "It turns out that diffusion models can be und...
Post a Comment
Read more

Nadella hands-on 🙆‍♂️

"Microsoft CEO Satya Nadella has sharply criticized the company's own AI assistant in internal communications, stating that integrations connecting Copilot with Gmail and Outlook 'don't really work' for the most part and are 'not smart,' according to a December 28, 2025 report from The Information . "The frank assessment reveals growing concerns within Microsoft about Copilot's technical execution and market performance, despite the company's public positioning of AI as central to its growth strategy.  "Nadella has assumed an unusually hands-on role in fixing the struggling product, essentially becoming what sources describe as the company's top product manager."
Post a Comment
Read more

Flocking patterns

" Murmuration  by Le Patin Libre is the latest contemporary dance show on ice (and perhaps the only one) that’s been years in the making. "Skaters who grow up in the world of organized sport have been conditioned to show off jazz hands for such-and-such gala or throw in a triple axel for so-and-so points. "The biggest challenge for them was to decondition all these conditions and find skaters who were willing to be open-minded to exploring choreography that wasn't so set. "Of course, there were still some recognizable jumps and spins in there —you can't get rid of the basics —but what the cast did come up with was an impressive, iridescent echo of birds in flight. "The animal kingdom presents some very unique flocking patterns.  "In geese migrations to warmer weather, the wind currents of the leader’s flapping provide support to the geese following behind, and their signature V-formation is the most aerodynamic shape to make use of this tactic.  ...
Post a Comment
Read more

Agents need total access 🫥

"For individuals, Farmer from the Ada Lovelace Institute says, many people have already built up intense relationships with existing chatbots and may have shared huge volumes of sensitive data with them during the process, making them different from other systems that have come before. "'Be very careful about the quid pro quo when it comes to your personal data with these sorts of systems,' Farmer says. 'The business model these systems are operating on currently may well not be the business model that they adopt in the future.' "'Even if, let's say, you genuinely consent and you genuinely are informed about how your data is used, the people with whom you interact might not be consenting,' Véliz, the Oxford associate professor, says. 'If the system has access to all of your contacts and your emails and your calendar and you’re calling me and you have my contact, they're accessing my data too, and I don't want them to'."
Post a Comment
Read more

Developers

"They are tired of hearing about artificial intelligence as a panacea without pragmatism. They want to hear about how they can pragmatically and easily harness it —now —for real-life use cases. "Indeed, we’ve spent the last few years bombarded by hyperbolic talk about AI (Robotaxis anyone?). How it’s going to transform life as we know it. How it’s going to take our jobs. When it will become sentient… "Meanwhile, AI has kind of quietly become part of the fabric of our lives —not by changing our lives or taking our jobs or becoming sentient, but by making our lives and our jobs easier.  "For example, when I Googled When will AI become sentient?  (and When did Skynet become self-aware,  for comparison purposes), I didn’t have to comb through results one at a time but instead read the AI-generated summary of the most relevant content at the top, with sources."
Post a Comment
Read more