Robot Arms in Orbit, a Google Game Builder and AI Skin Checks Top This Week's Data News

Space, voting and personal assistants

Seattle-based Orbital Robotics has built autonomous robotic arms that NASA will test on the International Space Station with its free-flying Astrobee robot. According to the roundup, the system lets Astrobee approach, capture and manipulate satellite-like targets, showing how autonomous servicing could work in microgravity. The test aims to show how free-flying robots could eventually repair, upgrade or reposition satellites without astronauts.

On the consumer side, Colorado-based The KyleGroup has created a prompt to help voters interpret complex ballot measures. Users copy official ballot language into an AI chatbot and ask for a plain-English explanation along with arguments for and against.

New York-based startup Persona has launched a personal AI assistant built around a wearable band. The device activates when a user presses a button or flicks their wrist, then sends encrypted data to Persona's cloud model. The model can order food, book travel, take notes and answer messages, and aims to become more personalized by learning user preferences over time.

Google has launched Playground, an AI-powered platform for building browser-based games through text prompts rather than coding. Creators can choose a genre, 2D or 3D visuals and game mechanics, and can upload images that the platform converts into game assets.

Health and biology take a large share

The Department of Health and Human Services has announced a program and three projects to accelerate U.S. clinical trials with AI tools. SURPASS would simulate trials, analyze incoming data and automate startup tasks. STACK aims to expand trial sites, COMMONS would build national research-data infrastructure, and CINCH would help patients identify relevant cancer trials.

Researchers at New York University trained AI models to detect signs of heart-transplant rejection by analyzing electrocardiograms alongside blood-test biomarkers. The system learned from 5,300 EKG readings of 2,357 transplant recipients and compared them with biopsy results. In a test of 38 patients, the combined model correctly identified 94 percent of patients who were not experiencing rejection, which the roundup says suggests it could reduce unnecessary invasive biopsies.

The U.S. government has also partnered with Google, Meta and the nonprofit Biohub to build large open biological datasets for training AI models. The initiative will collect standardized measurements of how cells respond to different conditions, giving researchers data to build predictive models of cellular behavior.

At St Luke's Hospital in Bradford, England, clinicians use an AI system to speed up assessments of suspicious skin lesions. The tool analyzes three photographs in minutes and helps identify cases that can be discharged or referred for specialist review, while clinicians double-check every result. The hospital says patients seen per session rose from 24 to 32 and unnecessary biopsies fell by 10 percent.

Ecosystems and workplace training

Researchers at the University of Wollongong in Australia developed an AI mapping system to monitor Antarctica's scattered ecosystems. It combines field observations, drone imagery and satellite data, allowing researchers to map vegetation from small patches of moss to broader landscape patterns. The approach aims to support long-term biodiversity monitoring across areas that are difficult and expensive to visit.

Seattle-based startup Yoodli has expanded its AI platform from speech and interview practice into broader workplace training. Employees can rehearse conversations with AI personas, generate training courses from a prompt and receive coaching inside tools such as Zoom and Teams. Managers get reports on whether employees' skills are improving.

What it means

The items are mostly early-stage projects, pilots or launches, and the roundup describes many of their benefits as aims or possibilities rather than proven results. The clearest reported figures come from healthcare: the NYU model's 94 percent correct identification of non-rejection cases in a 38-patient test, and the Bradford hospital's reported gains in patients seen and fewer unnecessary biopsies. Across the list, the stated common goal is to cut routine work, whether servicing satellites, reading ballots, running trials or screening skin lesions.