Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Deep learning has become an essential part of computer vision, with deep neural networks (DNNs) excelling in predictive performance. However, they often fall short in other critical quality dimensions, such as robustness, calibration, or fairness. While existing studies have focused on a subset of these quality dimensions, none have explored a more general form of “well-behavedness” of DNNs. With this work, we address this gap by simultaneously studying nine different quality dimensions for image classification. Through a large-scale study, we provide a bird’s-eye view by analyzing 326 backbone models and how different training paradigms and model architectures affect the quality dimensions. We reveal various new insights such that (i) vision-language models exhibit high fairness on ImageNet-1k classification and strong robustness against domain changes; (ii) self-supervised learning is an effective training paradigm to improve almost all considered quality dimensions; and (iii) the training dataset size is a major driver for most of the quality dimensions. We conclude our study by introducing the QUBA score (Quality Understanding Beyond Accuracy), a novel metric that ranks models across multiple dimensions of quality, enabling tailored recommendations based on specific user needs.

Read more | >

From JAMA Cardiol ogy: A centralized, population health coordinator-led notification and clinical support pathway improved the initiation of antihypertensive therapy in patients with left ventricular hypertrophy.

Despite the recognition that poorly controlled hypertension leads to adverse cardiovascular events, there are often barriers in care systems that contribute to substandard recognition and treatment.19 Notably, prior work employing trained nonphysicians focused on closing gaps in cardiovascular disease management has yielded significant improvements in disease-specific metrics using remote, centralized interventions.20-25 Similarly, there is a growing body of evidence demonstrating the effectiveness of clinician-directed support systems—often in the form of “nudges”—that have made meaningful advances in a variety of clinical outcomes.26,27 Whether a methodologic approach combining clinician nudges with the support of trained nonphysicians can be applied to LVH-associated diseases—including hypertension—is unknown.

Accordingly, the NOTIFY-LVH pragmatic randomized clinical trial28 sought to determine whether potentially underutilized echocardiogram data could be leveraged to improve patient care by augmenting the traditional ambulatory care framework. Specifically, this study tested whether a centralized clinical support pathway targeting clinicians of patients with LVH on their prior echocardiograms would increase the rate of treatment for hypertension and the earlier diagnosis of LVH-associated diseases.

Read more | >

Although lifespan has long been the focus of ageing research, the need to enhance healthspan — the fraction of life spent in good health — is a more pressing societal need. Caloric restriction improves healthspan across eukaryotes but is unrealistic as a societal intervention. Here, we describe the rewiring of a highly conserved nutrient sensing system to prevent senescence onset and declining fitness in budding yeast even when aged on an unrestricted high glucose diet. We show that AMPK activation can prevent the onset of senescence by activating two pathways that remove excess acetyl coenzyme A from the cytoplasm into the mitochondria — the glyoxylate cycle and the carnitine shuttle. However, AMPK represses fatty acid synthesis from acetyl coenzyme A, which is critical for normal cellular function and growth. AMPK activation therefore has positive and negative effects during ageing. Combining AMPK activation with a point mutation in fatty acid synthesis enzyme Acc1 that prevents inhibition by AMPK (the A2A mutant) allows cells to maintain fitness late in life without reducing the mortality associated with advanced age. Our research shows that ageing in yeast is not intrinsically associated with loss of fitness, and that metabolic re-engineering allows high fitness to be preserved to the end of life.

The authors have declared no competing interest.

Read more | >

Adeno-associated virus (AAV) is a prominent method for delivering genes in vivo. Therapeutic delivery to target cells is achieved through full capsids containing the gene cargo. However, the presence of empty capsids in the AAV drug product can reduce therapeutic effectiveness, necessitating their detection at various stages of the AAV production process. Traditional methods for assessing the AAV empty/full (E/F) ratio are often slow, labor-intensive, and require significant optimization.

Consider a novel, rapid, and high-throughput approach for determining the AAV E/F ratio using Octet® Biolayer Interferometry (BLI) alongside Octet® AAVX Biosensors. This cutting-edge technique evaluates intact viral capsids and is perfect for screening both crude and purified samples, offering a quicker and more efficient workflow with results available in as little as 30 minutes.

Discover the advantages of this innovative method and enhance your AAV workflow by downloading the technical note.

Read more | >

Imagine walking into your kitchen and instantly knowing if the fish you bought yesterday is still fresh—or entering an industrial site with sensors that immediately alert you to hazardous gas leaks. This isn’t science fiction—it’s the promise behind our newly developed nanomechanical sensor array, a powerful tool we’ve created to detect and analyze complex gases in real-time.

In our recent study published in Microsystems & Nanoengineering, we introduce a miniaturized array of silicon and polymer-based capable of detecting various gases quickly and accurately.

This array utilizes a simple yet ingenious principle: when gas molecules enter the sensor, they diffuse into specific polymers, causing them to swell slightly. This swelling generates detected by tiny piezoresistive sensors embedded in silicon. It’s like watching a sponge expand as it absorbs water—but at a microscopic scale, with the expansion measured electrically to detect and identify gases.

Read more | >

Astronomers have identified an exoplanet named Enaiposha, also known as GJ 1214 b, located 47 light-years from Earth. Initially classified as a mini-Neptune, further observations suggest it may belong to a different planetary category.

Read more | >

We now know it isn’t just neutron stars that emit such pulses. A white dwarf and a red dwarf star have been discovered closely orbiting each other emitting radio pulses every two hours. Their findings means we know it isn’t just neutron stars that emit such pulses, but these are spaced unusually far apart.

An international team of astronomers led by Dr Iris de Ruiter, now at the University of Sydney, has shown that a white dwarf and a red dwarf star orbiting each other every two hours are emitting radio pulses.

Thanks to follow-up observations using optical and x-ray telescopes, the researchers were able to determine the origin of these pulses with certainty. The findings explain the source of such radio emissions found across the Milky Way galaxy for the first time.

Read more | >