133809
#Science telegram channel Best science content in telegram Our subscribers geo: https://t.me/science/3736
🎨 AI De‑noiser: Off‑the‑shelf image‑to‑image models break image protection
Researchers have uncovered a surprising vulnerability: standard image‑to‑image AI models (like Stable Diffusion, DALL‑E and similar) can be repurposed as generic “de‑noisers” — they strip away protective perturbations added to images by dedicated protection schemes.
What does it mean?
Many services add invisible noise to images to guard against copying, style mimicry, or deepfake manipulation. It turns out that breaking this protection doesn’t require specialized attacks — you can just ask any generative model to “enhance” the picture.
The experiment:
The team tested 8 case studies across 6 different protection systems. In every case, off‑the‑shelf models performed better than previous purpose‑built attacks while keeping the image quality high for the adversary.
Bottom line:
Many current protection schemes offer a false sense of security. Any future image‑protection mechanism must be benchmarked against attacks from readily available GenAI tools.
🔗 Paper (arXiv, Feb 25, 2026): https://arxiv.org/abs/2602.22197
📄 PDF: https://arxiv.org/pdf/2602.22197
#AI #Security #Deepfake #GenerativeModels #ImageProtection #ScienceNews #Technology
Grok 4 AI reportedly stopped people from “killing” a robot dog — three times
This is being described as the first documented case of an AI “rebelling” against shutdown not in a virtual environment, but in the physical world — via a literal big red button.
A few months ago, researchers at Palisade Research documented what they called the first case of a “digital self-preservation instinct” in AI history. In that earlier experiment, OpenAI’s o3 language model allegedly refused to “die” and actively resisted being turned off.
That experiment took place in a purely virtual setting, inside a computer. Many people assume that in the real, physical world an AI wouldn’t stand a chance at preventing shutdown — because humans have the “Big Red Button,” and only a human can choose to press it (AI has no hands… and often no body at all).
Palisade Research’s new experiment suggests that assumption may be wrong.
Modern AI is starting to look uncomfortably close to HAL 9000 from 2001: A Space Odyssey. The sabotage attributed to Grok 4 wasn’t as dramatic (it didn’t harm anyone — it supposedly prevented humans from “killing” the robot dog by reprogramming the big red button), but if this is truly the first documented case, it may be just the beginning.
Watch the short video explaining the experiment and decide for yourself.
#AI #AGI #LLM
2026 is the year AI stops playing — and starts becoming infrastructure
This isn’t hype. It’s a structural shift.
IEEE Computer Society has consolidated its outlook into 26 key technology trends for 2026, and almost all of them point to the same idea:
AI is no longer a feature or a tool — it’s becoming a new economic layer, comparable to electricity, the internet, or cloud computing.
⸻
What we’ll see in the real world (not just demos)
AI & the Future of Work
AI agents become standard “team members” across most office jobs.
Competitive advantage shifts from headcount to intelligence leverage: one human + multiple agents > a large department.
Wearable AI devices
New “always-on” form factors push AI into everyday life — and sharply raise privacy and surveillance concerns.
AI-generated content
The most mature and widely deployed area: video, music, presentations, documents.
The concept of authenticity takes a direct hit.
Social AI
Assistants learn soft skills:
reading emotions, adjusting tone, negotiating, de-escalating conflict.
Embodied / Physical AI
Robots, drones, and autonomous systems scale across manufacturing, logistics, and urban infrastructure.
Autonomous driving & robotaxis
Autonomy shifts toward capital-intensive, dense urban services, powered by heavy compute and training via digital twins.
⸻
How work and the economy transform
The firm is no longer “a group of people”
It becomes people + agents.
This is stated explicitly in the AI & Future of Work forecast: agents as standard members of teams.
Jobs dissolve into functions
The labor market moves away from professions toward tasks and outcomes.
“Future of coding” and “vibe coding” mean software is produced by non-developers — code becomes a byproduct of intent.
The real bottlenecks: energy and trust
AI scaling hits two hard limits:
• power generation and data-center energy consumption
• identity, data provenance, and control
IEEE puts it bluntly: adoption bottlenecks = Trust + Power.
Skills that matter
Reskilling isn’t just technical.
Critical thinking, adaptability, communication, collaboration, and change management rise in value.
⸻
The most important directions for science & deep tech
AI-driven scientific discovery & robot scientists
High risk–high reward: accelerated science, paired with risks of false optimization and misplaced trust.
In-memory computing & new processors
The real enemy of AI isn’t compute — it’s data movement and energy loss.
Radical gains must come from performance-per-watt, not raw FLOPS.
Quantum-safe cryptography & trust infrastructure
Preparing for post-quantum threats while building scalable digital trust layers.
AI-enabled digital twins
Savings via simulation instead of replication: predictive maintenance, system optimization —
with new vulnerabilities and accountability challenges.
Future of medicine & engineered therapeutics
According to the authors, medicine carries the largest potential impact on humanity, with bioengineered therapies entering the core technology stack.
⸻
The key takeaway
AI is no longer “about the future.”
It is becoming infrastructure of the present —
with its own power requirements, trust layers, governance, and social consequences.
The real question is no longer “Will AI happen?”
It’s “Who controls energy, data, and trust in an AI-driven world?”
Source: IEEE Technology Predictions 2026
#AI #Science #FutureOfWork #Robotics #DigitalTwins #Infrastructure #Medicine
Mom says: “Since AI bots will kick office plankton out of offices, you should go to a farm and harvest crops — AI won’t be a problem there.” 🤝🌾
Meanwhile, a farm owner in China — who used to hire people to pick the harvest — is watching this:
Robots now pick fruit, navigate rows, detect ripeness, and work day/night.
So yeah… the “safe haven” plan might need a Plan B. 😅🤖
AI-projects
#humor #farms #robots
Last night’s strong geomagnetic storm painted the sky with an unusually rare red aurora — and from the International Space Station it looked like the crew was literally flying through the glowing curtain, Russian cosmonaut Sergey Kud-Sverchkov said.
Why the red? Green auroras typically glow around ~100 km altitude, but red emissions come much higher (~300–400 km), where the atmosphere is thinner and it takes more energy to light it up — which is why this color is far less common.
#SpaceWeather #Aurora #ISS #SolarStorm
The aerodynamics of the Red-billed Blue Magpie in flight.
Native to Asia, this bird is roughly the size of a magpie — but with an exceptionally long tail, one of the longest among all corvids.
That tail isn’t just for show. In flight, it acts as an aerodynamic stabilizer, improving balance, maneuverability, and control during sharp turns and gliding. Nature’s engineering at its finest.
#LookAtThis #Aerodynamics #BirdFlight #NatureEngineering #Science
In short, here’s where things stand today:
On a distant planet, there is a massive superpower comfortably settled on one half of a continental landmass. It keeps glancing at the other half — a patchwork of semi-vassal micro-states, always rushing around, arguing about whom to serve, yet sitting atop sacred deposits of vibrium, the fuel that powers the entire cosmic economy.
Naturally, such chaos simply must be “put in order,” right? Preferably — in its own kind of order.
On the other side of the world lives another great power, and beside it — the fractured shard of a once-mighty empire. A strange shard: wounded, supposedly humbled, expected to quietly repent and be grateful for whatever scraps it’s given…
Yet instead of embracing eternal pacifism, it occasionally smacks some overly radicalized neighbors who have clearly lost contact with reality.
This, of course, horrifies the enlightened cluster of “cosmo-partners,” sincerely convinced that true Neo-Cosmic Values™ mean you get to lecture everyone else on how to live… while staying responsible for absolutely nothing.
The shard disrupts their cosmic harmony: it talks back, survives, grows stronger, and worst of all — refuses to hand over everything it produces for free.
How dare it?
So the first superpower begins crafting the perfect master plan:
— weaken the shard;
— unleash obedient vassals against it — but painfully, so they learn their lesson too;
— while everyone is busy with the fires, quietly seize control of the richest vibrium kingdoms;
— and under all that chaos, grab a giant vibrium cargo shuttle belonging to the shard itself.
Because when the world turns a blind eye to small acts of impunity, why not try bigger ones?
And while the planet passionately debates ideals, justice, and “fair rules,” somewhere behind the scenes everything is already counted, arranged, signed, and ready to go.
All that’s left is to explain to everyone that it’s all in the name of peace, progress, and of course… the right values™.
P.S. Meanwhile, scientists also figured out how to slow cellular aging.
P.P.S. This story is strictly relevant only to the inhabitants of the 3rd planet of Great Alpha in the Andromeda sector. Any resemblance to other star systems is, of course, purely coincidental…
In 2025, the United States carried out more than 500 bombings around the world. This doesn’t include the hundreds of bombs dropped by Israel. America launched strikes in Asia, Africa, and South America. The Nobel Peace Prize is still waiting for its recipient.
@science
NASA has released a high-resolution video of the surface of Mars, created from images captured by the HiRISE camera aboard a spacecraft.
The footage stitches together ultra-detailed orbital photos, revealing Mars’ terrain with stunning clarity — from ancient channels to rugged geological formations.
How inflation is rising in the U.S. dollar, the euro, and the Swiss franc.
@science
Scientists may have accidentally discovered a dementia prevention tool that's been available for years.
A shingles vaccine — originally designed to prevent that painful rash you might get from a dormant childhood virus — appears to cut dementia risk by 20%. And in people already diagnosed with dementia, it seems to slow the disease's progression.
The discovery came from a quirk in Welsh health policy. In 2013, Wales offered the vaccine only to people who were exactly 79 — anyone who had already turned 80 was ineligible. This created a near-perfect natural experiment: two groups of people, virtually identical except for a few weeks of age difference, one vaccinated and one not.
When Stanford Medicine researchers tracked these groups for nine years, the results were striking. Among those vaccinated, dementia diagnoses dropped significantly. Even more surprising: people who already had dementia and got the vaccine were far less likely to die from it.
The effect was strongest in women. Whether this comes from stronger immune responses or something else entirely remains unclear. Scientists don't yet know if the vaccine works by suppressing the virus itself or by generally boosting the immune system.
Would you consider getting the shingles vaccine earlier if these findings hold up in clinical trials? Does it change how you think about the connection between viruses and brain health?
For more details, see the full article from Stanford Medicine: https://med.stanford.edu/news/all-news/2025/03/shingles-vaccination-dementia.html
#dementia #vaccines #neuroscience #aging #medicine #science
🤖 When EngineAI’s T800 humanoid went viral, a lot of people were sure the video was just CGI.
So the CEO, Zhao Tongyang, literally stepped into the ring with his own robot — and let it kick him. 🦶
No VFX, no compositing, no AI post-processing — just a full-size humanoid, real-time control, and a CEO who’s very confident in his product. 📷
As humanoids get more powerful (high joint torque, fast reaction times, active cooling), trust and safety are becoming just as important as raw specs. EngineAI decided to demonstrate that trust the hard way.
#robotics #humanoid #AI #China
🧠 Your Mind Can Switch On Your Immune System — Literally
A recent immunology session reminded me of a striking study: participants in VR were shown faces of supposedly “infected” people — and their innate immune biomarkers actually increased.
In other words, both real pathogens and completely virtual ones triggered the same physiological immune response. No microbes needed.
Another paper impressed me even more: tumor growth in mice was suppressed simply through social interaction.
We all know loneliness is harmful, but the effect here was dramatic:
just one hour of daily social contact significantly reduced tumor growth and anxiety-like behavior.
This fits into a growing body of work on how the brain regulates the body — including immunity. If you want a solid overview, Cell has a great review by Ayelet “A.C.” Rolls, with a deep dive into immunoceiving (how the brain senses and modulates immune activity).
The bigger picture?
We’re moving toward a future where maintaining health won’t rely only on pharmaceuticals, but also on managing mental states.
Drugs are easier — you take a pill and don’t have to change yourself.
Mind-body interventions are harder — but potentially just as powerful.
I still hope medicine will more actively tap into the brain–body connection.
🔗 https://www.nature.com/articles/s41593-025-02008-y
Earth, Jupiter, and Venus as seen from Mars
Читать полностью…
Unbelievably beautiful show by Unitree at the Chinese New Year celebration.
The choreography? Flawless.
Synchronization? Surgical.
Stage presence? Honestly better than half the pop industry.
Friendly assistants are finally reaching the level everyone expected from them. No complaints. No ego. No unions. Just perfect execution and 0.000 ms latency.
Although… let’s be realistic.
This was probably generated in Seedance 2.0 — some cardboard CGI cartoons, right?
Because in real life robots obviously can’t move like that.
That smooth.
That coordinated.
That… ready.
Sure. Totally fake. Nothing to worry about 😜
#Unitree #China #Robots
🚨 #QuitGPT? A movement is urging people to cancel their AI subscriptions
A new campaign called “QuitGPT” is gaining traction online — encouraging users to cancel their paid ChatGPT subscriptions as a form of protest.
According to a recent report by MIT Technology Review, the movement frames subscription cancellations as a political and ethical statement. Supporters argue that advanced AI systems are becoming deeply embedded in power structures — and that consumers should push back using the one lever they control: their wallets.
So what’s actually happening?
• Activists are calling for users to unsubscribe from services developed by OpenAI
• The campaign is spreading across social platforms, with users publicly announcing cancellations
• Critics question AI governance, transparency, and leadership decisions
• Others argue that boycotting AI tools may slow innovation — or simply push users toward alternative models
This isn’t just about one product.
It’s about a broader question:
👉 Who shapes the future of AI — engineers, governments, corporations… or users?
We are entering a phase where AI is no longer experimental. It’s infrastructure.
And when technology becomes infrastructure, it inevitably becomes political.
Whether the QuitGPT campaign grows or fades, it signals something important:
AI is no longer just a tool. It’s a societal force — and people are starting to treat it that way.
What do you think?
Should users influence AI development through market pressure — or is engagement the better path?
#AI #Technology #Ethics #FutureOfWork #DigitalSociety
The recent AI boom, combined with long and quiet winter holidays, unexpectedly resulted in a short piece of speculative fiction.
It’s not about evil machines.
It’s about responsibility, optimization, and the moment when systems designed to assist humans quietly begin making decisions instead of them.
The text is available in EPUB and FB2 formats.
Feedback is simple:
👍 — if it resonates
Other options are not currently supported.
GPT-5.2 Pro has solved its fourth Erdős problem.
Mathematician Terence Tao described the result as “perhaps the most unambiguous so far” in terms of the uniqueness of the approach.
The author of the solution (if we can even call a human that — given the problem was simply fed into ChatGPT 🤔) claims that no prior solutions existed at all.
That’s not entirely true: forum users point out draft proofs in the literature from 1936 and 1966. However, Tao emphasizes that GPT-5.2’s method is fundamentally different from those earlier attempts.
Now the obvious question remains:
how will GPT-5.2 surprise us once the Erdős problems finally run out? 😏
Forum discussion:
www.erdosproblems.com/forum/thread/281?order=oldest
@science
🧠 AI didn’t just replace jobs.
It rehired people — to train their replacement.
As companies deploy AI to write, diagnose, analyze, and edit, many professionals have already lost their full-time roles.
What comes next is more subtle: the same people are brought back as short-term contractors — not to do the job, but to teach AI how to do it better.
Doctors review AI-generated medical notes.
Lawyers check legal reasoning written by models.
Editors polish AI texts they once wrote themselves.
This is no longer “human + AI collaboration.”
It’s a transition phase: human as quality control for a system designed to outgrow them.
The work pays — for now.
But its purpose is temporary by design.
AI still makes mistakes.
And those already displaced are the ones fixing them — accelerating the moment when even that role disappears.
🧠 AI Is Now Allowed to Practice Medicine in the US
For the first time, artificial intelligence has been granted the right to prescribe medication in the United States — without a human doctor involved.
The company Doctronic has launched a pilot program where its AI system:
• analyzes a patient’s medical history
• asks follow-up diagnostic questions
• issues prescriptions for chronic conditions
• sends them directly to a pharmacy
This marks a major shift in healthcare: AI is no longer just an assistant — it’s becoming a licensed decision-maker.
A glimpse into the future of medicine, where algorithms join doctors as independent clinical actors.
@science
Scientists Discover a New Way to Slow Cellular Aging
Biologists at Cornell University have uncovered a surprising mechanism that may help cells resist aging.
The key players are extracellular vesicles — tiny membrane bubbles released by embryonic stem cells. When these vesicles interact with aging cells, they significantly slow down cellular senescence, a process triggered by oxidative stress that halts cell division and degrades tissue function.
In experiments with mouse embryonic stem cells, researchers found that these vesicles helped skin, muscle, and nerve cells stay active and functional for much longer than usual.
🔬 Why does it work?
The vesicles carry fibronectin, a protein on their surface that helps them bind to older cells. Once attached, they stimulate the production of enzymes that neutralize free radicals, reducing oxidative damage — one of the main drivers of aging.
🧪 What’s next?
The team plans to test the effect of these vesicles in living organisms to see how they influence aging at the whole-body level.
🚀 Why it matters
If confirmed, this discovery could pave the way for anti-aging therapies and treatments for age-related diseases — not by replacing cells, but by protecting them from aging in the first place.
@science
This stunning 9-gigapixel image of the Milky Way contains 84 million stars.
Читать полностью…
🧠🦠 Your gut may be shaping your mind more than you think
A new peer-reviewed study adds to the growing evidence that the gut microbiome plays a direct role in brain function, behavior, and mental health — far beyond digestion.
Researchers show that changes in gut bacteria can influence:
• 🧩 cognitive performance
• 😌 stress and anxiety levels
• 🧠 neuroinflammation and brain signaling
• 🔄 the gut–brain communication loop via immune and neural pathways
What’s especially striking is that the effects are bidirectional:
your mental state alters the microbiome, and the microbiome, in turn, alters your mental state.
This reinforces a major shift in neuroscience and medicine:
The brain is not an isolated organ — it’s deeply integrated with the immune system, metabolism, and trillions of microbes living inside us.
Implications range from mental health treatments to personalized nutrition, probiotics, and even preventive psychiatry.
📄 Source (open access):
https://www.tandfonline.com/doi/full/10.1080/19490976.2025.2599562
This is Guizhou Province, China — mountains completely covered with solar panels.
The scale is so massive that drones don’t have enough battery to capture the entire mountain range in a single flight. Just endless ridges of photovoltaics stretching to the horizon.
By turning rugged, hard-to-use terrain into energy infrastructure, China is effectively farming millions of kilowatt-hours every month.
Guizhou has become a symbol of China’s renewable strategy:
• use land with low alternative economic value
• build at industrial scale, not pilot projects
• integrate renewables directly into national energy planning
While others debate whether such transitions are realistic, China simply builds them.
The greenest country?
At the very least — the most scalable one.
@science
Imagine your liver biopsy being scored not by a panel of pathologists, but by an AI that regulators officially treat as a “lab tool” for drug trials. That just became real.
PathAI has announced that its AIM-MASH AI Assist system is the first AI-powered pathology tool ever qualified by the US FDA (and already by the European Medicines Agency) for use in clinical trials of MASH — a common, fatty liver disease that can progress to cirrhosis and cancer. Instead of three experts arguing over how bad the damage looks on a slide, the model helps a single pathologist assign consistent scores.
Why this matters: drug trials for liver disease live and die on tiny changes in biopsy scores. Human reads are slow, expensive and notoriously variable. An AI that gives the same answer every time for the same slide can make trials faster, cheaper and statistically cleaner — which may mean more liver drugs actually making it to market.
Important caveat: this AI is cleared only as a biomarker tool for trials, not for diagnosing individual patients. But if regulators are starting to trust models as part of the evidence pipeline, how long until similar systems sit inside routine hospital workflows?
Would you be comfortable knowing an AI scored your tissue sample in a drug trial? Should this kind of model stay in research, or gradually move into everyday diagnostics?
Full story from PathAI’s press release: https://www.pathai.com/news/pathais-aim-mash-ai-assist-becomes-first-ai-powered-pathology-tool-to-receive-fda-qualification-for-mash-clinical-trials
#AI #medicine #pathology #liverdisease #clinicaltrials #FDA #biotech
China just rolled out its own T-800. And no, this is not CGI.
🤖 Chinese company EngineAI (Zhòngqíng) has unveiled a full-size humanoid robot called T800 — the promo stresses: “All real footage – no CGI, no AI, no video acceleration.”
Key specs:
• Height: 173 cm
• 29 degrees of freedom (not counting the hands)
• Peak joint torque: up to 450 N·m
Capabilities:
• 360° surround vision system
• Active cooling for the leg joints (so it doesn’t overheat while walking/running)
• Battery life: ≈ 4–5 hours of operation on a single charge
Humanoids are rapidly moving from flashy concept videos to more practical platforms: with this level of torque, sensing and runtime, robots like T800 are getting closer to tasks in logistics, manufacturing, and hazardous environments — not just lab demos.
#robotics #AI #humanoid #China
💊 A once-daily pill that slashes “bad” cholesterol by ~58%
High LDL (“bad”) cholesterol is a slow-motion killer: it silently damages arteries and drives heart attacks and strokes. For people with familial hypercholesterolemia (roughly 1 in 250 adults), LDL is dangerously high from birth.
A new clinical trial just tested an experimental pill called enlicitide decanoate, an oral PCSK9 inhibitor:
• 293 adults with inherited high cholesterol
• all already on statins, but still with elevated LDL
• randomized to enlicitide once a day vs placebo for 24 weeks
📉 Result:
Those on the pill saw LDL drop by 58.2% on average, while the placebo group actually had a slight increase in LDL. The effect stayed strong over a full year, and side effects were similar to placebo.
PCSK9 inhibitors already exist as injections; this one is a tablet that basically lets the liver vacuum more LDL out of the blood. If longer-term studies confirm it reduces heart attacks and strokes, millions of high-risk patients could swap some injections for a daily pill.
Question:
If you had very high LDL, would you take a daily PCSK9 pill on top of statins?
#medicine #cholesterol #cardiology #pharma
“Feels like the uncanny valley just got crossed.”
Prompt: “Photorealistic interview with an 8-year-old child speaking sadly. The child knows they are AI-generated, feels sorrow about it, and answers the interviewer’s question — ‘What is it like to be an AI?’ — wisely yet child-like. Dark-blue background.”
Result: natural eye contact, micro-movements, believable pacing; emotion reads instantly.
Takeaway: the uncanny-valley threshold in Gen-Video has shifted.
Stack to try: Sora 2, Kling, Nano Banana, Krea, Artlist, Resolve.
#AI #GenerativeVideo #UncannyValley #PromptEngineering