Lifeboat Foundation

View insights.

Often when Dr. Thomas Valley sees a new patient in the intensive care unit at Michigan Medicine in Ann Arbor, he clamps a pulse oximeter on their finger – one of the many devices he uses to gauge their health and what course of care they might require, whether they are a child having seizures, a teenage car accident victim or an older person with Covid-19.

But recently, Valley, an assistant professor in the University of Michigan’s Division of Pulmonary and Critical Care, realized first-hand that the small device may yield less accurate oxygen readings in patients with dark skin.

Continue reading “Scientists are searching for solutions after studies show pulse oximeters don’t work as well for people of color” »

In 2019, astronomers observed the nearest example to date of a star that was shredded, or “spaghettified,” after approaching too close to a massive black hole.

That tidal disruption of a sun-like star by a black hole 1 million times more massive than itself took place 215 million light years from Earth. Luckily, this was the first such event bright enough that astronomers from the University of California, Berkeley, could study the optical light from the stellar death, specifically the light’s polarization, to learn more about what happened after the star was torn apart.

Their observations on Oct. 8, 2019, suggest that a lot of the star’s material was blown away at high speed—up to 10,000 kilometers per second—and formed a spherical cloud of gas that blocked most of the high-energy emissions produced as the black hole gobbled up the remainder of the star.

Researchers from North Carolina State University have developed a new approach to federated learning that allows them to develop accurate artificial intelligence (AI) models more quickly and accurately. The work focuses on a longstanding problem in federated learning that occurs when there is significant heterogeneity in the various datasets being used to train the AI.

Federated learning is an AI training technique that allows AI systems to improve their performance by drawing on multiple sets of data without compromising the privacy of that data. For example, federated learning could be used to draw on privileged patient data from multiple hospitals in order to improve diagnostic AI tools, without the hospitals having access to data on each other’s patients.

Federated learning is a form of machine learning involving multiple devices, called clients. The clients and a centralized server all start with a basic model designed to solve a specific problem. From that starting point, each of the clients then trains its local model using its own data, modifying the model to improve its performance. The clients then send these “updates” to the centralized server. The centralized server draws on these updates to create a hybrid model, with the goal of having the hybrid model perform better than any of the clients on their own. The central server then sends this hybrid model back to each of the clients. This process is repeated until the system’s performance has been optimized or reaches an agreed-upon level of accuracy.

The Future of Earth: 1,000 Years From Now.

In the last 250 years, humans have drastically and irreversibly transformed the Earth. Greenhouse gases emitted by human industries have changed the planet’s climate, presenting the single greatest threat humanity has ever faced. If humans can cause such incredible damage to the Earth in 250 years, what will our planet look like in 1,000 years time?

Continue reading “The Future of Earth: 1000 Years From Now” »

In recent years, roboticists have developed increasingly advanced robotic systems, many of which have artificial hands or robot hands with multiple fingers. To complete everyday tasks in both homes and public settings, robots should be able to use their “hands” to efficiently grasp and manipulate objects.

Enabling dexterous manipulation involving multiple fingers in robots, however, has so far proved challenging. This is primarily because it is an advanced skill that entails an adaptation to the shape, weight, and configuration of objects.

Researchers at Universität Hamburg have recently introduced a new approach to teach robots to grasp and manipulate objects using a multi-fingered robotic hand. This approach, introduced in IEEE Transactions on Neural Networks and Learning Systems, allows a robotic hand to learn from humans through teleoperation and adapt its manipulation strategies based on human hand postures and the data gathered when interacting with the environment.

A new King’s-led study, published in the Proceedings of the National Academy of Sciences, has found that a single factor (a protein coding gene known as Sox8) can make non-ear cells adopt ear character during embryo development. The findings not only demonstrate how cell fate decisions are regulated in the embryo but may also inform reprogramming and regenerative strategies for the ear developmental malformations.

Responsible for the sense of hearing and balance, the inner ear is critically important for communication with the environment. In humans, developmental malformations of the ear have life-long consequences, while age-related hearing defects affect a large proportion of the population. Currently, there are no therapies that involve biological approaches—only hearing aids or cochlear implants, as how the ear normally develops is not fully understood and many of the controlling factors are poorly characterized.

Researchers from the Faculty of Dentistry, Oral and Craniofacial Sciences at King’s, in collaboration with colleagues from the Francis Crick Institute, explored the earliest steps in ear development to determine what causes cells to become ear cells, and what makes them different from cells which form other sense organs.

The gravitational constant G determines the strength of gravity—the force that makes apples fall to the ground or pulls the Earth in its orbit around the sun. It is part of Isaac Newton’s law of universal gravitation, which he first formulated more than 300 years ago. The constant cannot be derived mathematically; it has to be determined through experiment.

Over the centuries, scientists have conducted numerous experiments to determine the value of G, but the scientific community isn’t satisfied with the current figure. It is still less precise than the values of all the other fundamental natural constants—for example, the speed of light in a vacuum.

One reason gravity is extremely difficult to quantify is that it is a very weak force and cannot be isolated: when you measure the gravity between two bodies, you also measure the effect of all other bodies in the world.

Waste heat is a promising source of energy conservation and reuse, by means of converting this heat into electricity—a process called thermoelectric conversion. Commercially available thermoelectric conversion devices are synthesized using rare metals. While these are quite efficient, they are expensive, and in the majority of cases, utilize toxic materials. Both these factors have led to these converters being of limited use. One of the alternatives is oxide-based thermoelectric materials, but the primary drawback these suffer from is a lack of evidence of their stability at high temperatures.

A team led by Professor Hiromichi Ohta at the Research Institute for Electronic Science at Hokkaido University has synthesized a barium cobalt oxide thermoelectric converter that is reproducibly stable and efficient at temperatures as high as 600°C. The team’s findings have been published in the journal ACS Applied Materials & Interfaces.

Thermoelectric conversion is driven by the Seebeck effect: When there is a temperature difference across a conducting material, an electric current is generated. However, efficiency of thermoelectric conversion is dependent on a figure called the thermoelectric figure of merit ZT. Historically, oxide-based converters had a low ZT, but recent research has revealed many candidates that have high ZT, but their stability at high temperatures was not well documented.

Moore’s Law needs a hug. The days of stuffing transistors on little silicon computer chips are numbered, and their life rafts—hardware accelerators—come with a price.

When programming an accelerator—a process where applications offload certain tasks to system hardware especially to accelerate that task—you have to build a whole new software support. Hardware accelerators can run certain tasks orders of magnitude faster than CPUs, but they cannot be used out of the box. Software needs to efficiently use accelerators’ instructions to make it compatible with the entire application system. This translates to a lot of engineering work that then would have to be maintained for a new chip that you’re compiling code to, with any programming language.

Now, scientists from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) created a new programming language called “Exo” for writing high-performance code on hardware accelerators. Exo helps low-level performance engineers transform very simple programs that specify what they want to compute, into very complex programs that do the same thing as the specification, but much, much faster by using these special accelerator chips. Engineers, for example, can use Exo to turn a simple matrix multiplication into a more complex program, which runs orders of magnitude faster by using these special accelerators.