Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog – Page 2

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

Log in for authorized contributors

Godfather of AI: How To Make Safe Superintelligent AI

The co-inventor of modern AI and the most cited living scientist believes he’s figured out how to ensure AI is honest, incapable of deception, and never goes rogue. Yoshua Bengio – Turing Award Winner and founder of LawZero – is disturbed by the many unintended drives and goals present in today’s AIs, their ability to tell when they’re being tested, and demonstrated willingness to lie. AI companies are trying to stamp these out in a ‘cat-and-mouse game’ that Yoshua fears they’re losing.

But Yoshua is optimistic: he believes the companies can win this battle decisively with a single rearrangement to how AI models are trained, and has been developing mathematical proofs to back up the claim. The core idea is that instead of training AI to predict what a human would say, or to produce responses we’d rate highly, we should train it to model what’s actually true.

Learn more & full transcript: https://80k.info/bengio.

Yoshua argues this new architecture, which he calls “Scientist AI,” is a small enough change that we could keep almost all the techniques and data we use to train frontier AIs like Claude and ChatGPT. And that the new architecture need not cost more, could be built iteratively, and might be more capable as well as more honest.

Until recently, the biggest practical objection to Scientist AI was simple: the world wants agents, and Scientist AI isn’t one. But in new research, Yoshua has extended the design and believes the same honest predictor can be turned into a capable agent without losing its \.

Wealthy Individuals Funding Science is Good for Everyone

Excellent article on the importance of private funding for cutting-edge science.

“The skepticism toward private science funding is part of a broader anti-capitalist sentiment, likely fueled by real affordability problems in housing, healthcare, and education. These concerns are understandable. But directing private capital toward fundamental science benefits everyone, and treating this the same as other uses of wealth only ensures that money flows into megayachts rather than research.”

Private wealth funded most of history’s scientific breakthroughs. Stigmatizing it now is holding us all back.

Kanvas makes the microbiome druggable—and the implications are massive

Kanvas looks amazing! They’re systematically deciphering microbiomes and developing clinical-stage interventions to improve patient outcomes in oncology and beyond. Very impressive! I’m also especially interested in their approach to maternal envi­ron­mental enteric dysfunction (EED), which apparently affects 150M people!

Ever since the genomics revolution revealed how reliant the human organism is on its microscopic microbial cohabitants, the microbiome has been medicine’s most elusive frontier, promising better health if only we could untangle the trillions of interactions that influence nearly every facet of our physiology. But until now, effective medicines that harness the microbiome have been rare. Because of the diversity of microbial species and the complexity of host-to-microbe interactions, as well as the lack of a reliable, easily manufactured drug modality (the package that delivers a medicine’s therapeutic effect), the microbiome has been hard to treat, despite its importance to functions like immune response. Microbiome science has disappointed patients, doctors, founders, and investors.

That’s why DCVC is so excited about the cascade of recent developments at Kanvas Biosciences, which is moving the field beyond descriptive profiling of the microbiome to translating comprehensive biochemical insights into clinically useful products. In the past few weeks, the Princeton-based spatial biology company has kicked off a Phase 1 clinical trial for its first drug candidate, secured significant new backing from the Gates Foundation (closing a $48 million Series A financing, bringing Kanvas’s total funding to $78 million), and bolstered its scientific leadership by adding one of the most respected names in bioengineering to its board.

Clinical milestone

The most significant milestone in Kanvas’s evolution is the dosing of the first patients in a Phase I clinical trial for KAN-4. This live biotherapeutic product (LBP), resembling an ordinary pill, treats the colitis that many cancer patients develop after receiving immune checkpoint inhibitors (ICIs), allowing them to remain on the life-saving therapy longer.

One Type of Fat May Accelerate Brain Aging, But Scientists Say There’s a Fix

A higher level of the fat that gathers around organs has been linked to faster brain aging in a new study, with glucose and insulin the likely mediators.

The study, led by a team from Ben-Gurion University of the Negev (BGU) in Israel, suggests that reducing visceral fat can protect against brain atrophy.

Like other parts of the body, the brain doesn’t necessarily age at a consistent rate: wear and tear can increase or decrease, depending on numerous factors. Faster brain aging typically means a faster decline in mental performance, and a higher risk of brain diseases.

Specialized RNA molecules could counter ALS neurodegeneration

Misshapen proteins cause a mess of trouble—particularly in neurodegenerative diseases. But a new study suggests it’s possible that giving them a little bit of extra support could keep them working correctly, and even reverse the damage they have caused.

The new research focuses on one such aberrant protein, TDP-43, which binds to RNA in the cell’s nucleus and is responsible for regulating thousands of human genes. If TDP-43 turns from a healthy, liquid-like phase into diseased, fibrous solid-like aggregates, its presence can be fatal.

This protein is one of the key drivers of the diseases amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD)—a discovery first made by pioneering Penn Medicine scientists Virginia M.-Y. Lee, Ph.D., MBA, and the late John Trojanowski, MD, Ph.D.

Dynamically Tunable Singular States Through Air-Slit Control in Asymmetric Resonant Metamaterials

This study presents a novel method for dynamically tuning singular states in one-dimensional (1D) photonic lattices (PLs) using air-slit-based structural modifications. Singular states, arising from symmetry-breaking-induced resonance radiation, generate diverse spectral features through interactions between resonance modes and background radiation. By strategically incorporating air slits to break symmetry in 1D PLs, we demonstrated effective control of resonance positions, enabling dual functionalities including narrowband band pass and notch filtering. These singular states originate from asymmetric guided-mode resonances (aGMRs), which can be interpreted by analytical modeling of the equivalent slab waveguide. Moreover, the introduction of multiple air slits significantly enhances spectral tunability by inducing multiple folding behaviors in the resonance bands.

Blood as the mirror and modulator of aging: mechanistic insights and rejuvenation strategies

Aging is a complex process influenced by changes in our blood that affect how quickly we age. Scientists have shown that blood contains important molecules and cellular components — including proteins, metabolites, and immune cells — that can either accelerate or slow aging. Tools such as the ‘proteomic aging clock’ predict age and disease risk based on blood protein profiles, whereas emerging multi-omics approaches integrate metabolomic and immunomic data. Large-scale analyses of circulating factors reveal how these components change with age and identify markers of organ-specific aging. Certain blood-borne molecules can predict diseases such as heart disease and Alzheimer disease. These findings demonstrate that aging does not occur uniformly across tissues. Overall, studying diverse blood components provides valuable insight into aging biology and offers opportunities to develop strategies that promote healthier aging and improve long-term health.

This summary was initially drafted using artificial intelligence, then revised and fact-checked by the author.

Void-Filled Material Stops Intense Electron Beam

An intense electron beam is stopped more efficiently by a highly porous material than by a less porous material, suggesting new strategies for controlling beams.

New experiments show that porous materials consisting mostly of empty space can absorb the energy carried by an ultraintense electron beam more effectively than porous media with higher mass densities. The finding contradicts the prevailing notion that denser and thicker obstacles always provide more stopping power and suggests that the microstructure of a material fundamentally changes its electron-stopping ability. Simulations by the experimental team revealed the physical mechanisms behind this “anomalous-stopping” effect, which the researchers believe provides a new way to control the propagation of electron beams in extreme environments [1].

The study focuses on relativistic electron beams (REBs), which travel at close to the speed of light. REBs that carry currents in the mega-ampere regime can deliver petawatts (1015 watts) of power to a small target in a pulse lasting for a few picoseconds. This high intensity makes them ideal for creating and probing extreme states of matter that exist in stars, planetary cores, or nuclear events. The short bursts of intense energy provided by REBs are also used in inertial-confinement fusion—a scheme in which high-power lasers heat a fuel pellet and trigger nuclear fusion.

The first direct observation of laser-created isolated hopfions

Over the past few decades, some physicists worldwide have been investigating unusual particle-like magnetic structures known as topological solitons. These structures could potentially be leveraged to develop new cutting-edge technologies, such as new magnetic memory devices and computing systems.

A type of topological solitons that has proven to be difficult to realize experimentally is the hopfion. This is a three-dimensional (3D) structure comprised of closed loops of continuously swirling spin textures, which can resemble linked or knotted vortex strings.

Researchers at South China University of Technology, Nankai University, Forschungszentrum Jülich, South China Normal University, University of Luxembourg, and Uppsala University recently reported the first direct observation of isolated hopfions in a magnetic material, which were created using laser pulses.

/* */