NASA’s James Webb Space Telescope has captured the “Cosmic Owl,” a startling owl-faced pair of colliding ring galaxies. This double-ring structure is exceptionally rare: ring galaxies account for just 0.01% of known galaxies, and two colliding rings is almost unheard of. The JWST image provides an exceptional natural laboratory for studying galaxy evolution. Models suggest the galactic clash began roughly 38 million years ago, meaning the owl-like shape could persist for a long time. A team led by Ph.D. student Mingyu Li of Tsinghua University in China announced the finding.

Spotting the ‘Cosmic Owl’

According to Mingyu Li, the first author of the new study , he and his team found the Owl by combing through public JWST data from the COSMOS field. The twin ring galaxies jumped out thanks to JWST’s infrared imaging. Each ring is about 26,000 light-years across (a quarter of the Milky Way), and each harbors a supermassive black hole at its core – one of the Owl’s eyes.

JWST images show the collision interface – the Owl’s beak – ablaze with activity. ALMA observations find a huge clump of molecular gas there – the raw fuel for new stars – being squeezed by the impact. Radio observations show a jet from one galaxy’s black hole slamming into the gas. Li notes the shockwave-plus-jet have ignited an intense starburst, turning the beak into a stellar nursery.

Rarity and Significance

Ring galaxies are extremely rare (≈0.01% of all galaxies), so finding two in collision is unheard of. Another team independently identified the same system and called it the “Infinity Galaxy”. Li says this event is an exceptional natural laboratory for studying galaxy evolution. In one view, researchers can see black holes feeding, gas compressing and starbursts happening together.

Li points out the collision’s shockwave and jet have triggered an intense starburst in the beak. He says this may be a crucial way to turn gas into stars rapidly, which could help explain how young galaxies built up their mass so quickly. Simulations will clarify the precise collision conditions needed to produce such a rare twin-ring “owl” shape.

The use of conventional robots for industry and hazardous environments is easy for the purpose of control and modelling. However, these are too rigid to operate in confined places and uneven terrain. The soft bio-related roots are better adapted to the environment and manoeuvring in inaccessible places. Such flexible capabilities would need an array of on-board sensors and spacious models which are tailored to each robot design. Having a new and less resource-demanding approach, the researchers at MIT have developed a far less complex, deep learning control system that teaches the soft, bio-inspired robots to follow the command from a single image only.

Soft Robots Learn from a Single Image

As per Phys.org, this research has been published in the journal Nature, by training a deep neural network on two to three hours of multi-view images of various robots executing random commands, the scientists trained the network to reconstruct the range and shape of mobility from only one image. The previous machine learning control designs need customised and costly motion systems. Lack of a general-purpose control system limited the applications and made prototyping less practical.

The methods unshackle the robotics hardware design from the ability to model it manually. This has dictated precision manufacturing, extensive sensing capabilities, costly materials and reliance on conventional and rigid building blocks.

AI Cuts Costly Sensors and Complex Models

The single camera machine learning approach allows the high-precision control in tests on a variety of robotic systems, adding the 3D-printed pneumatic hand, 16-DOF Allegro hand, a soft auxetic wrist and a low-cost Poppy robot arm.

As this system depends on the vision alone, it might not be suitable for more nimble tasks which need contact sensing and tactile dynamics. The performance may also degrade in cases where visual cues are not enough.

Researchers suggest the addition of sensors and tactile materials that can enable the robots to perform different and complex tasks. There is also potential to automate the control of a wider range of robots, together with minimal or no embedded sensors.