Waymo is recalling thousands of its self-driving cars in the US over a software issue that could allow vehicles to drive into flooded roads. According to a letter posted on the National Highway Traffic Safety Administration (NHTSA) website on Tuesday, the voluntary recall affects nearly 3,800 robotaxis that use the company's fifth and sixth-generation automated driving systems. It follows an incident on 20 April in San Antonio, Texas, where an empty Waymo vehicle entered a flooded road and was swept into a creek. The company, which hopes to be operating a robotaxi service in London by September, said it was working on additional software safeguards, according to CNBC. The BBC has contacted Waymo, which is owned by Google's parent company Alphabet, for comment.

John Pendry is known for creating an invisibility cloak. John Pendry's kitchen is dominated by a huge photograph of what looks like the view through a kaleidoscope: dizzying shards of purple, green, yellow and white. Given that Pendry is famous above all else for inventing an invisibility cloak - a device that can bend light around objects - I wonder if I am looking at something related to that. But no, he tells me, the image simply shows crystals of vitamin C magnified many times. All that invisibility-cloak stuff is in the past, he says, and he has moved on to "more exciting things".

Large language models (LLMs) have shown remarkable capabilities in language understanding and generation.

The Department of Homeland Security said agents arrested an illegal alien after he “weaponized" his car during a chaotic attempt to evade arrest in a San Antonio parking lot.

Animal traffickers face up to 20 years in prison and a $250,000 fine. Breakthroughs, discoveries, and DIY tips sent every weekday. It should go without saying, but please don't smuggle spider monkeys. While responding to a human trafficking case earlier this year, United States Border Patrol agents in Laredo, Texas, found two of these tiny primates . The driver failed to yield and fled the scene, leading officers to respond.

Earth's turbulent atmosphere makes it hard to detect new planets from the ground. Astronomer Rebecca Jensen-Clem is working out how to find them anyway. The pendant on Rebecca Jensen-Clem's necklace is only about an inch wide, composed of 36 silver hexagons entwined in a honeycomb mosaic. At the Keck Observatory, in Hawaii, just as many segments make up a mirror that spans 33 feet, reflecting images of uncharted worlds for her to study. Jensen-Clem, an astronomer at the University of California, Santa Cruz, works with the Keck Observatory to figure out how to detect new planets without leaving our own. Typically, this pursuit faces an array of obstacles: Wind, fluctuations in atmospheric density and temperature, or even a misaligned telescope mirror can create a glare from a star's light that obscures the view of what's around it, rendering any planets orbiting the star effectively invisible.

Future deep learning models will be distinguished by systems that perpetually learn through interaction, imagination, and cooperation, blurring the line between training and inference. This makes continual learning a critical challenge, as methods that efficiently maximize bidirectional transfer across learning trajectories will be essential. Replay is on track to play a foundational role in continual learning, allowing models to directly reconcile new information with past knowledge. In practice, however, replay is quite unscalable, doubling the cost of continual learning when applied naively. Moreover, the continual learning literature has not fully synchronized with the multi-task fine-tuning literature, having not fully integrated highly scalable techniques like model merging and low rank adaptation into a replay-enabled toolset that can produce a unified model in the face of many sequential tasks. In this paper, we begin by applying and analyzing low rank adaptation in a continual learning setting. Next, we introduce consolidation, a phasic approach to replay which leads to up to 55\% less replay samples being needed for a given performance target. Then, we propose sequential merging, an offshoot of task arithmetic which is tailored to the continual learning setting and is shown to work well in combination with replay. Finally, we demonstrate that the developed strategies can operate synergistically, resulting in a highly scalable toolset that outperforms standalone variants.