Toggle light / dark theme

Accurately forecasting weather remains a complex challenge due to the inherent uncertainty in atmospheric dynamics and the nonlinear nature of weather systems. As such, methodologies developed ought to reflect the most probable and potential outcomes, especially in high-stakes decision-making over disasters, energy management, and public safety. While numerical weather prediction (NWP) models offer probabilistic insights through ensemble forecasting, they are computationally expensive and prone to errors. Although ML models have been very promising in giving faster and more accurate predictions, they fail to represent forecast uncertainty, especially in extreme events. This makes ML-based models less useful in actual real-world applications.

The physics-based ensemble models, for example, the ENS from the European Centre for Medium-Range Weather Forecasts (ECMWF), rely on these simulations to produce probabilistic forecasts. These models properly represent the forecast distributions and joint spatiotemporal dependencies and require high computational resources and manual engineering. Conversely, the ML-based method, like GraphCast or FourCastNet, focuses only on deterministic forecasts and will minimize the errors in the mean outcome without considering any uncertainty. None of the attempts to generate probabilistic ensembles by MLWP produced realistic samples or competed with the accuracy of operational ensemble forecasts. Hybrid approaches like NeuralGCM embed ML-based parameterizations within traditional frameworks but have poor resolution and limited performance.

Researchers from Google DeepMind released GenCast, a probabilistic weather forecasting model that generates accurate and efficient ensemble forecasts. This machine learning model applies conditional diffusion models to produce stochastic trajectories of weather, such that the ensembles consist of the entire probability distribution of atmospheric conditions. In systematic ways, it creates forecast trajectories by using the prior states through autoregressive sampling and uses a denoising neural network, which is integrated with a graph-transformer processor on a refined icosahedral mesh. Utilizing 40 years of ERA5 reanalysis data, GenCast captures a rich set of weather patterns and provides high performance. This feature allows it to generate a 15-day global forecast at 0.25° resolution within 8 minutes, which is state-of-the-art ENS in terms of both skill and speed.

You might be keenly interested to know that this eagerness to produce responses is something tuned into AI. The AI maker has made various computational adjustments to get the AI to press itself to respond. Why so? Because people want answers. If they aren’t getting answers from the AI, they will go someplace else. That’s not good for the AI maker since they are courting views.

There is a ton of research taking place about AI hallucinations. It is one of the most pressing AI issues of our time.

AI hallucinations are considered a scourge on the future of generative AI and LLMs. Sadly, the state-of-the-art AI still has them, for example, see my analysis of OpenAI’s most advanced ChatGPT or new model o1 that still indeed emits AI hallucinations at the link here. They are like the energy bunny and seem to just keep running.

“This is a highly engineered design, but the fundamental concepts are fairly simple,” said Dr. Jie Yin. “And with only a single actuation input, our robot can navigate a complex vertical environment.”


What influence can marine life have on robotics? This is what a recent study published in Science Advances hopes to address as a team of researchers from the University of Virginia and North Carolina State University have developed the fastest swimming soft robot by taking cues from manta ray fins. This study holds the potential to help researchers, engineers, and scientists develop faster and more efficient swimming soft robots that can be used for a variety of purposes worldwide.

This study builds on a 2022 study conducted by this same team of researchers that explored swimming soft robots that exhibited butterfly strokes, achieving a then-record of 3.74 body lengths per second, along with demonstrating high power efficiency, low energy use, and high maneuverability. For this new study, the researchers developed fins used by manta rays with the goal of achieving greater results than before. The fins are flexible when not in use but become rigid when the researchers pumped air into the silicone body that encompasses the soft robot.

In the end, the researchers not only achieved low energy use and maneuverability, but also broke their own record of body lengths per second at 6.8. Additionally, the manta ray-inspired swimming soft robot was able to avoid obstacles, which was an improvement from their 2022 study.

When the brain is observed through imaging, there is a lot of “noise,” which is spontaneous electrical activity that comes from a resting brain. This appears to be different from brain activity that comes from sensory inputs, but just how similar—or different—the noise is from the signal has been a matter of debate.

New research led by a team at the University of Tokyo further untangles the relationship between internally generated noise and stimulus-related patterns in the brain, and finds that the patterns of spontaneous activity and stimulus-evoked response are similar in lower visual areas of the cerebral cortex, but gradually become independent, or “orthogonal,” as one moves from lower to higher visual areas.

The findings not only enhance our understanding of the mechanism that enables the brain to distinguish between signal and noise, but could also provide clues for developing noise-resistant incorporating a mechanism similar to that found in the biological brain. The study is published in the journal Nature Communications.

On December 5, 2024, the European Space Agency (ESA) achieved a milestone in space exploration with the successful launch of its Proba-3 mission, which aims to create artificial solar eclipses. This revolutionary mission could provide groundbreaking insights into the Sun’s mysterious atmosphere, the corona. By creating artificial eclipses, the two Proba-3 spacecraft will work together to block the Sun’s light, allowing scientists to observe its outer layers like never before. These solar eclipses will provide a close-up view of the corona for the first time, unlocking secrets that were previously beyond our reach.

The Proba-3 mission is built around a remarkable concept: two satellites, the Occulter and the Coronagraph, will fly in precise formation, separated by a distance of 500 feet. This configuration will allow the Occulter to block the Sun’s light and cast a shadow onto the Coronagraph, creating an artificial eclipse in orbit. By mimicking the conditions of a natural solar eclipse, scientists will be able to observe the Sun’s corona for extended periods, up to six hours at a time, far surpassing the fleeting moments provided by natural eclipses on Earth.

This level of precision, described by ESA as “down to the thickness of a fingernail,” is unprecedented in space exploration. The spacecraft rely on a suite of advanced technologies, including GPS, star trackers, lasers, and radio links, to maintain their exact positioning autonomously. This capability allows the spacecraft to operate as though they were a single, integrated observatory, delivering the optical performance required for such ambitious science objectives.

Chinese rivals, including Alibaba and DeepSeek, have released reasoning models like Marco-o1 and R1-Lite-Preview, are encroaching fast, challenging OpenAI’s dominance with open-source solutions and eclipsing o1-preview on certain third-party benchmarks.

These developments reflect the growing demand for large reasoning models (LRMs) capable of handling complex problem-solving tasks.

As OpenAI continues to refine its offerings, the rollout of o1 and ChatGPT Pro marks a milestone in its quest to provide accessible, high-performance AI tools. Whether these developments can maintain OpenAI’s leadership in an increasingly crowded market remains to be seen.

The BBC speaks to residents and travellers in some of the top-ranked countries on the 2024 Global Innovation Index to find out how cutting-edge technology benefits day-to-day life.

With the rise of AI, self-driving cars and wi-fi connected appliances, it can feel like innovation is everywhere these days. But certain countries are known for developing cutting-edge technologies that benefit residents and visitors alike.

To dive into those countries making the most impact in these areas, the World Intellectual Property Organisation recently released its 2024 Global Innovation Index, ranking 130 economies based on measures like their education system, technology infrastructure and knowledge creation (like patents filed or mobile apps created).