Robot Finishes Marathon

RAIBO 2, developed by a research team from the Korea Advanced Institute of Science & Technology (KAIST), has become the first four-legged robot in the world to finish a full marathon, showcasing its walking control technology and durability. It participated in the 22nd Sangju Marathon, completing the 42.195-kilometer course in four hours, 19 minutes and 52 seconds.

The course includes two 50-meter-high hills at the 14-km and 28-km marks, making it a challenging route for runners and for quadruped robots due to unexpected obstacles. The team, used an advanced learning algorithm for RAIBO 2, allowing it to train its walking mechanism in an environmental simulation to replicate diverse terrains such as slopes, stairs and icy paths.

More information:

https://www.koreatimes.co.kr/www/tech/2024/11/129_386562.html

Digital Holographic Imaging BCI

Researchers at the Johns Hopkins Applied Physics Laboratory (APL) in Laurel, Maryland, and the Johns Hopkins School of Medicine in Baltimore have achieved a breakthrough in noninvasive, high-resolution recording of neural brain activity. They demonstrated that neural tissue deformations may provide a novel signal for brain activity that has the potential to be leveraged for future BCI devices. The team developed a digital holographic imaging (DHI) system to identify and validate the signal as tissue deformation that occurs during neural activity. Tissue deformation is only tens of nanometers in height, so the DHI system was developed with sensitivities at the nanometer scale.

The DHI system operates by actively illuminating the tissue with a laser and recording the light scattered from the neural tissue on a special camera. This information is processed to form a complex image of the tissue from which magnitude and phase information can be precisely recorded to spatially resolve changes in brain tissue velocity. Numerous fundamental tests were conducted over several years to ensure the signal the team identified was in fact correlated to when neurons fired. The neural signal was challenging to identify because of competing noise from physiological clutter such as blood flow, heart rate and respiratory rate. They also discovered that the clutter could also provide insight into the health of an individual.

More information:

https://www.jhuapl.edu/news/news-releases/241114-noninvasive-brain-computer-interface

Jailbreak LLM-Driven

University of Pennsylvania researchers developed an algorithm that can jailbreak robots controlled by a large language model (LLM). The RoboPAIR algorithm uses an attacker LLM to provide prompts to a target LLM, adjusting the commands until they bypass the safety filters. It also employs a judge LLM to ensure the attacker LLM produces prompts that consider the target LLM's physical limitations, such as certain obstacles in the environment.

One finding the scientists found concerning was how jailbroken LLMs often went beyond complying with malicious prompts by actively offering suggestions. The researchers stressed that prior to the public release of their work, they shared their findings with the manufacturers of the robots they studied, as well as leading AI companies. They also noted they are not suggesting that researchers stop using LLMs for robotics.

More information:

https://spectrum.ieee.org/jailbreak-llm

Robot Cleans Bathroom Sink

A robotic arm has mastered the surprisingly complex task of sink washing, showing off its ability to learn. With the help of an AI model trained in surgery videos, a robot system has successfully carried out difficult surgical tasks as skillfully as a human. Researchers decided to let their robotic arms learn to do the task by observing someone else doing it.

Researchers developed a cleaning sponge equipped with force and position sensors and had a person use it to clean just the front edge of a sink that had been sprayed with dyed gel imitating dirt. They then used the data to train a neural network and then fed those patterns to the robot and let them inform its movements as it sets out about the task.

More information:

https://newatlas.com/robotics/robot-cleans-sink/