Lifeboat Foundation

Machine learning influences numerous aspects of modern society, empowers new technologies, from Alphago to ChatGPT, and increasingly materializes in consumer products such as smartphones and self-driving cars. Despite the vital role and broad applications of artificial neural networks, we lack systematic approaches, such as network science, to understand their underlying mechanism. The difficulty is rooted in many possible model configurations, each with different hyper-parameters and weighted architectures determined by noisy data. We bridge the gap by developing a mathematical framework that maps the neural network’s performance to the network characters of the line graph governed by the edge dynamics of stochastic gradient descent differential equations. This framework enables us to derive a neural capacitance metric to universally capture a model’s generalization capability on a downstream task and predict model performance using only early training results. The numerical results on 17 pre-trained ImageNet models across five benchmark datasets and one NAS benchmark indicate that our neural capacitance metric is a powerful indicator for model selection based only on early training results and is more efficient than state-of-the-art methods.

New data confirms that variation in our Universe’s expansion rate (or the Hubble ‘tension’) is not an error in measurement.

Summary: A new experimental cancer drug could ease cognitive difficulties for those with Rett syndrome, a rare autism-linked disorder, by enhancing brain cell functions. The drug, ADH-503, improves the activity of microglia, which are crucial for maintaining neural networks.

Researchers found that healthy microglia restored synapse function in brain organoids mimicking Rett syndrome. This breakthrough suggests potential therapies for Rett syndrome and other neurological conditions.

Using AI-generated datasets to train future generations of machine learning models may pollute their output, a concept known as model collapse, according to a new paper published in Nature. The research shows that within a few generations, original content is replaced by unrelated nonsense, demonstrating the importance of using reliable data to train AI models.

Generative AI tools such as large language models (LLMs) have grown in popularity and have been primarily trained using human-generated inputs. However, as these AI models continue to proliferate across the Internet, computer-generated content may be used to train other AI models—or themselves—in a recursive loop.

Ilia Shumailov and colleagues present mathematical models to illustrate how AI models may experience model collapse. The authors demonstrate that an AI may overlook certain outputs (for example, less common lines of text) in training data, causing it to train itself on only a portion of the dataset.

Astronomers have discovered a black hole with a mass about 33 times greater than that of our sun, the biggest one known in the Milky Way aside from the supermassive black hole lurking at the center of our galaxy.

The newly identified black hole is located about 2,000 light-years from Earth — relatively close in cosmic terms — in the constellation Aquila, and has a companion star orbiting it, researchers said on Tuesday. A light year is the distance light travels in a year, 5.9 trillion miles.

Black holes are extraordinarily dense objects with gravity so strong that not even light can escape, making it difficult to spot them. This one was identified through observations made in the European Space Agency’s Gaia mission, which is creating a huge stellar census, because it caused a wobbling motion in its companion star. Data from the European Southern Observatory’s Chile-based Very Large Telescope and other ground-based observatories were used to verify the black hole’s mass.

Sexual size dimorphism in mammals, often linked to sexual selection, can impacts genome evolution. This study finds sexual dimorphism in body size is associated with expanded gene families for olfactory functions and contracted gene families for brain development.

When seawater gets cold, it gets viscous. This fact could explain how single-celled ocean creatures became multicellular when the planet was frozen during “Snowball Earth,” according to experiments.

The new spacesuit urine filtration system by Cornell researchers recycles astronauts’ urine into drinkable water, aiming to improve comfort and efficiency on future Moon and Mars missions.

Currently, astronauts must relieve themselves inside their spacesuits during spacewalks. This is not only uncomfortable and unhygienic, but also wasteful. Unlike the wastewater management system on the International Space Station (ISS), the water from urine during spacewalks is not recycled.

Researchers at Cornell are addressing this issue with a novel urine collection and filtration system inspired by the ‘stillsuits’ from the sci-fi franchise Dune. Like these ‘stillsuits’, their prototype absorbs and purifies urine, and recycles it into drinking water.

For generations, researchers have been pondering the question of how and where consciousness is formed in the brain.

Professor Ekrem Dere from Ruhr University Bochum, Germany, proposes a new approach to researching conscious cognitive information processing. He advocates defining phases of conscious cognitive processes on the basis of behavioral observations and learning curves.