Lifeboat Foundation

CRISPR—Clustered Regularly Interspaced Short Palindromic Repeats—is the microbial world’s answer to adaptive immunity. Bacteria don’t generate antibodies when they are invaded by a pathogen and then hold those antibodies in abeyance in case they encounter that same pathogen again, the way we do. Instead, they incorporate some of the pathogen’s DNA into their own genome and link it to an enzyme that can use it to recognize that pathogenic DNA sequence and cut it to pieces if the pathogen ever turns up again.

The enzyme that does the cutting is called Cas, for CRISPR associated. Although the CRISPR-Cas system evolved as a bacterial defense mechanism, it has been harnessed and adapted by researchers as a powerful tool for genetic manipulation in laboratory studies. It also has demonstrated agricultural uses, and the first CRISPR-based therapy was just approved in the UK to treat sickle-cell disease and transfusion-dependent beta-thalassemia.

Now, researchers have developed a new way to search genomes for CRISPR-Cas-like systems. And they’ve found that we may have a lot of additional tools to work with.

Beyond glycemic benefits, users of the app also reported significantly less-related emotional distress than standard care (−1.9 vs 1.7 points in composite survey scores, P =0.03).

“We are currently working on making technology like this accessible to patients outside of research settings because we think it can really help patients in underserved areas who need high-touch care to get their under control,” Nayak said.

The researchers developed their custom voice-based AI app and had it powered by Alexa (Amazon wasn’t involved with the study). The software was equipped with titration algorithms by the American Association of Clinical Endocrinologists and the American College of Endocrinology and included emergency protocols to handle hypoglycemia and hyperglycemia.

In the era of big data and advanced artificial intelligence, traditional data storage methods are becoming inadequate. To address the need for high-capacity and energy-efficient storage solutions, the development of next-generation technologies is crucial.

Among these is resistive random-access memory (RRAM), which relies on altering resistance levels to store data. A recent study published in the journal Angewandte Chemie details the work of a research team who have pioneered a method for creating supramolecular memristors, one of the key components in the construction of nano-RRAM.

In the basement of Kirchhoff-Institut für Physik in Germany, researchers have been simulating the Universe as it might have existed shortly after the Big Bang. They have created a tabletop quantum field simulation that involves using magnets and lasers to control a sample of potassium-39 atoms that is held close to absolute zero. They then use equations to translate the results at this small scale to explore possible features of the early Universe.

The work done so far shows that it’s possible to simulate a Universe with a different curvature. In a positively curved universe, if you travel in any direction in a straight line, you will come back to where you started. In a negatively curved universe, space is bent in a saddle shape. The Universe is currently flat or nearly flat, according to Marius Sparn, a PhD student at Kirchhoff-Institut für Physik. But at the beginning of its existence, it might have been more positively or negatively curved.

A recent study published in IEEE Transactions on Control of Network Systems discusses how artificial intelligence (AI) can be used to control microgrids in the event of a long-term power outage caused by natural disasters or human error. This study was conducted by a team of researchers at UC Santa Cruz and holds the potential to improve power restoration techniques, which are traditionally controlled by local utility companies. One benefit of microgrids is they can function to power a small area, such as a town, until the primary utility source comes back online.

“Nowadays, microgrids are really the thing that both people in industry and in academia are focusing on for the future power distribution systems,” said Dr. Yu Zhang, who is an assistant professor of electrical and computer engineering at UC Santa Cruz and co-author on the study.

For the study, the researchers used an AI-based approach to develop a novel method where microgrids could draw power from renewable energy sources while being disconnected from the primary utility source, known as “islanding mode”, but can also function while being connected to the source, as well. This new method, which they refer to as constrained policy optimization (CPO), uses a machine learning algorithm that learns from outside input, such as real-time changes in environmental or power conditions, and makes the best-informed decisions on what to do next.

BMIs using intracortical electrodes, such as Utah arrays, are particularly adept at sensing fast changing (millisecond-scale) neural activity from spatially localized regions (1 cm) during behavior or stimulation that is correlated to activity in such spatially specific regions, for example, M1 for motor and V1 for vision. Intracortical electrodes, however, struggle to track individual neurons over longer periods of time, for example, between subsequent recording sessions^15,16. Consequently, decoders are typically retrained every day¹⁵. A similar neural population identification problem is also present with an ultrasound device, including from shifts in the field of view between experiment sessions. In the current study, we demonstrated an alignment method that stabilizes image-based BMIs across more than a month and decodes from the same neurovascular populations with minimal, if any, retraining. This is a critical development that enables easy alignment of a previous days’ models to a new day’s data and allows decoding to begin with minimal to no new training data. Much effort has focused on ways to recalibrate intracortical BMIs across days that do not require extensive new data^{18,19,20,21,22,23}. Most of these methods require identification of manifolds and/or latent dynamical parameters and collecting new neural and behavioral data to align to these manifolds/parameters. These techniques are, to date, tailored to each research group’s specific applications with varying requirements, such as hyperparameter tuning of the model²³ or a consistent temporal structure of data²². They are also susceptible to changes in function in addition to anatomy. For example, ‘out-of-manifold’ learning/plasticity alters the manifold²⁴ in ways that many alignment techniques struggle to address. Finally, some of the algorithms are computationally expensive and/or difficult to implement in online use²².

Contrasting these manifold-based methods, our decoder alignment algorithm leverages the intrinsic spatial resolution and field of view provided by fUS neuroimaging to perform decoder stabilization in a way that is intuitive, repeatable and performant. We used a single fUS frame (∼ 500 ms) to generate an image of the current session’s anatomy and aligned a previous session’s field of view to this single image. Notably, this did not require any additional behavior for the alignment. Because we only relied upon the anatomy, our decoder alignment is robust, can use any off-the-shelf alignment tool and is a valid technique so long as the anatomy and mesoscopic encoding of relevant variables do not change drastically between sessions.

It remains an open question as to how much the precise positioning of the ultrasound transducer during each session matters for decoder performance, especially out-of-plane shifts or rotations. In these current experiments, we used linear decoders that assumed a given image pixel is the same brain voxel across all aligned data sessions. To minimize disruptions to this pixel–voxel relationship, we performed image alignment within the 2D plane. As we could only image a 2D recording plane, we did not correct for any out-of-plane brain shifts between sessions that would have disrupted the pixel–voxel mapping assumption. Future fUS-BMI decoders may benefit from three-dimensional (3D) models of the neurovasculature, such as registering the 2D field of view to a 3D volume^25,26,27 to better maintain a consistent pixel–voxel mapping.

“We want the robot to ask for enough help such that we reach the level of success that the user wants. But meanwhile, we want to minimize the overall amount of help that the robot needs,” said Allen Ren.

A recent study presented at the 7th Annual Conference on Robotic Learning examines a new method for teaching robots how to ask for further instructions when carrying out tasks with the goal of improving robotic safety and efficiency. This study was conducted by a team of engineers from Google and Princeton University and holds the potential to design and build better-functioning robots that mirror human traits, such as humility. Engineers have recently begun using large language models, or LLMs—which is responsible for designing ChatGPT—to make robots more human-like, but this can also come with drawbacks, as well.

“Blindly following plans generated by an LLM could cause robots to act in an unsafe or untrustworthy manner, and so we need our LLM-based robots to know when they don’t know,” said Dr. Anirudha Majumdar, who is an assistant professor of mechanical and aerospace engineering at Princeton University and a co-author on the study.

Continue reading “Teaching Robots to Ask for Help: A Breakthrough in Enhancing Safety and Efficiency” »

The explosive growth of generative AI over the last year has been truly phenomenal. Kick-started by the public release of ChatGPT (was it really only a year ago?), it’s now everywhere. Keen to ride the wave, every app from Office to eBay has been adding generative capabilities, and growing numbers of us are finding uses for it in our everyday and professional lives.

Given its nature, it’s not surprising that content creators, in particular, have found it a powerful addition to their toolset. Marketing agencies, advertising creatives, news organizations and social media influencers have been among the most enthusiastic early adopters.

While it brings great opportunities for improving efficiency and automating manual, repetitive elements of creative work, it also throws up significant challenges. Issues around copyright, spam content, hallucination, the formulaic nature of algorithmic creation and bias all need to be considered by professionals planning on adopting it into their workflow.

When galaxies collide, their supermassive black holes enter into a gravitational dance, gradually orbiting each other ever closer until eventually merging. We know they merge because we see the gravitational beasts that result, and we have detected the gravitational waves they emit as they inspiral. But the details of their final consummation remain a mystery. Now a new paper published on the pre-print server arXiv suggests part of that mystery can be solved with a bit of dark matter.

Just as the famous three-body problem has no general analytical solution for Newtonian gravity, the two-body problem has no general solution in general relativity. So, we have to resort to computer simulations to model how black holes orbit each other and eventually merge.

For binary black holes that are relatively widely separated, our simulations work really well, but when black holes are close to each other things get complicated. Einstein’s equations are very nonlinear, and modeling the dynamics of strongly interacting black holes is difficult.