AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
In the absence of numerous robotic demonstrations, Being-M0.7 first learns the dynamics of motion from more than 10,000 hours of human data, then translates it to the control of a humanoid's entire body.
A digital twin produces the demonstrations needed for adaptation. In the trial on a Stretch 3 robot, the personalized policy succeeds in 63.9% of the shots.
Sensors and inference don't always follow a fixed rhythm. Chronos encodes physical time to better withstand delays and asynchronous data flows.
Derived from Cosmos Policy, SmolWAM reduces the computation and memory required. It achieves 95.5% success over 400 LIBERO episodes in the published protocol.
The system combines reusable locomotion with gestures from tennis videos. Real trials report 90% of successful forehands and exchanges with humans.
A code agent reviews the controller based on real tests under human authorization. The insertion success rate goes from 2 out of 21 to 17 out of 21 between generations.
Trained on 15,000 hours of manipulation videos from twelve different robots, the RoboJEPA family shows that planning progresses predictably with calculation.
By cross-referencing crop type, the product used and weather data, the system adjusts the application rate and travel speed, detects rows with 30 per cent greater accuracy and minimises drift.
The start-up pays volunteers fitted with sensors to record everyday tasks – a rare resource for robot training.
Its autonomous electric railcars, which run in trains without couplings, aim to bring rail transport back into the picture for journeys of less than 800 km, which are currently dominated by lorries.
The manufacturer of Atlas is bringing on board an AI specialist rather than a robotics engineer, at a time when Hyundai is expecting it to deliver 30,000 humanoid robots by 2028.
Spun out of Carnegie Mellon’s Safe AI Lab, the start-up puts robots through their paces in thousands of scenarios involving virtual humans who stumble, crouch or carry loads.
Following a fight between the influencer Frankie LaPenna and a remotely operated EngineAI T800 humanoid, the California Athletic Commission demanded that these events be halted on the grounds that they lacked authorisation.
Rather than carrying a load, the robot tows it: a single control system manages starting, turning and stopping, whether with human passengers or not, and without the need for any specific adjustments.
A team of researchers has proposed two adaptors that enable general-purpose robots to operate da Vinci instruments without modifying them, with a view to making robotic surgery more accessible.
The Munich-based manufacturer of industrial robotic arms has completed a $40 million funding round – mainly comprising shares sold by employees – led by Sequoia and Lightspeed, to accelerate its growth in the United States.
Researchers have proposed a flexible capacitive skin covering the entire hand, with the same arrangement of 202 sensors on both humans and robots, to train robots using human demonstrations.
The new dataset and its test set of 1,000 questions reveal that the best video models still struggle to contextualise the use of a tool within what they see.
A test suite comprising 18 tasks and 3,100 demonstrations on a Unitree G1 reveals a gap between selecting a tool and the complete execution of the task.
Tested on ex vivo pig colons, the system successfully completes 85.4 per cent of straight-line journeys and 70 per cent of 90-degree turns thanks to a policy that predicts sequences of actions.