One explanation for how people can plan efficiently despite limited cognitive resources is that we possess a set of adaptive planning strategies and know when and how to use them. But how are these strategies acquired? While previous research has studied how individuals learn to choose among existing strategies, little is known about the process of forming new planning strategies. In this work, we propose that new planning strategies are discovered through metacognitive reinforcement learning. To test this, we designed a novel experiment to investigate the discovery of new planning strategies. We then present metacognitive reinforcement learning models and demonstrate their capability for strategy discovery as well as show that they provide a better explanation of human strategy discovery than alternative learning mechanisms. However, when fitted to human data, these models exhibit a slower discovery rate than humans, leaving room for improvement.

This paper describes a novel approach to adaptive manufacturing in the context of small batch production and customization. It focuses on integrating task-level planning and reasoning with reinforcement learning (RL) in the SkiROS2 skill-based robot control platform. This integration enhances the efficiency and adaptability of robotic systems in manufacturing, enabling them to adjust to task variations and learn from interaction data. The paper highlights the architecture of SkiROS2, particularly its world model, skill libraries, and task management. It demonstrates how combining RL with robotic manipulators can learn and improve the execution of industrial tasks. It advocates a multi-objective learning model that eases the learning problem design. The approach can incorporate user priors or previous experiences to accelerate learning and increase safety. Spotlight video: https://youtu.be/H5PmZl2rRbs?si=8wmZ-gbwuSJRxe3S&t=1422 SkiROS2 code: https://github.com/RVMI/skiros2 SkiROS2 talk at ROSCon: https://vimeo.com/879001825/2a0e9d5412 SkiREIL code: https://github.com/matthias-mayr/SkiREIL

Scientists and tech industry leaders, including high-level executives at Microsoft and Google, issued a new warning Tuesday about the perils that artificial intelligence poses to humankind. "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war," the statement said. Sam Altman, CEO of ChatGPT maker OpenAI, and Geoffrey Hinton, a computer scientist known as the godfather of artificial intelligence, were among the hundreds of leading figures who signed the statement, which was posted on the Center for AI Safety's website. Worries about artificial intelligence systems outsmarting humans and running wild have intensified with the rise of a new generation of highly capable AI chatbots such as ChatGPT. It has sent countries around the world scrambling to come up with regulations for the developing technology, with the European Union blazing the trail with its AI Act expected to be approved later this year.

One of today's goals for industrial robot systems is to allow fast and easy provisioning for new tasks. Skill-based systems that use planning and knowledge representation have long been one possible answer to this. However, especially with contact-rich robot tasks that need careful parameter settings, such reasoning techniques can fall short if the required knowledge not adequately modeled. We show an approach that provides a combination of task-level planning and reasoning with targeted learning of skill parameters for a task at hand. Starting from a task goal formulated in PDDL, the learnable parameters in the plan are identified and an operator can choose reward functions and parameters for the learning process. A tight integration with a knowledge framework allows to form a prior for learning and the usage of multi-objective Bayesian optimization eases to balance aspects such as safety and task performance that can often affect each other. We demonstrate the efficacy and versatility of our approach by learning skill parameters for two different contact-rich tasks and show their successful execution on a real 7-DOF KUKA-iiwa.

We present a general framework for training safe agents whose naive incentives are unsafe. As an example, manipulative or deceptive behaviour can improve rewards but should be avoided. Most approaches fail here: agents maximize expected return by any means necessary. We formally describe settings with 'delicate' parts of the state which should not be used as a means to an end. We then train agents to maximize the causal effect of actions on the expected return which is not mediated by the delicate parts of state, using Causal Influence Diagram analysis. The resulting agents have no incentive to control the delicate state. We further show how our framework unifies and generalizes existing proposals.

ESA's SpaceBok robot is designed to walk, hop, and run in low-gravity environments. From free-flying droids to humanoids, from crawlers to inflatable torsos, space robots of myriad types are now being considered for missions in low Earth orbit, on interplanetary spacecraft, and on other worlds. It might sound like a prop list from a Star Wars movie, but space agencies and their contractors are developing a panoply of robotic assistants with a serious aim in mind: to boost the productivity and safety of astronauts. The idea behind robot assistants is multifaceted: one aim is to offload time-consuming repetitive tasks like space station cleaning and inventory making from crew members to free-flying or humanoid robots. Ground robots controlled from, say, spacecraft orbiting the Moon or Mars could construct human habitats ahead of a landing, or perform reconnaissance ahead of human exploration missions.

Within buildings 10 and 30 of the Siemens complex on the outskirts of Munich, the next generation of German workers are toiling over a range of test projects. The assignments are carefully chosen to impart the skills needed to continue the German miracle in automated manufacturing. In one room, a group of young men train to be automotive mechatronic engineers. They've just spent the past week feverishly programming a diminutive working model of an automated production line--complete with sensors, conveyor belts, and tools that work without human input. They're able to discuss their work in surprisingly good English, but what sets them apart from their peers in the US is that none of them attend a university.

Washington, Dec 26: Researchers have developed an automated technology that combines imaging with digital analysis and machine learning to help physicians detect melanoma at its early stages. People with melanoma often have mole-looking growths on their skin that tend to be irregular in shape and colour and can be hard to tell apart from benign ones, making the disease difficult to diagnose. "There is a real need for standardisation across the field of dermatology in how melanomas are evaluated," said James Krueger, Professor at Rockefeller University in the US. "Detection through screening saves lives but is very challenging visually, and even when a suspicious lesion is extracted and biopsied, it is confirmed to be melanoma in only about 10 per cent of cases," said Krueger. In the new approach, images of lesions are processed by a series of computer programmes that extract information about the number of colours present in a growth and other quantitative data.

Housemarque, the Finnish developer behind Resogun and Dead Nation, hasn't had the best year. I visited its Helsinki headquarters back in September to see how the studio was following up Resogun, the surprise hit of the PlayStation 4 launch. What I found was a unique company struggling to hold on to the identity it believes in. Housemarque made its name with Stardust. Originally released for the Amiga in the early '90s, the series rose to prominence with the digital release of Super Stardust HD on the PlayStation 3. The studio has since become a specialist in digital-only games, almost all of which can trace their lineage back to the arcade. The isometric shooter Dead Nation was the studio's next big hit, going on to become one of the bestselling digital-only titles for PlayStation 3, while the Ikaruga-meets-Metroid platformer Outland was critically acclaimed. But it was during the launch of the PlayStation 4 that Housemarque would make the biggest impact. Resogun took the basic premise behind the arcade classic Defender and turned it into a modern shooter. With cylindrical stages and a custom voxel-based engine, the game was by far the strongest PlayStation 4 exclusive of its time, and one of scant few highlights of the console's November 2013 launch. Sony clearly knew as much: It made Resogun free to all members of its PlayStation Plus subscription service, and as a result the game was downloaded by millions of PlayStation 4 owners.

EdTech runs four or five years behind TechCrunch headlines. Accordingly, hot topics in tech this year–AI, robotics, VR, wearables–will become widespread in education in a few years. Makerspaces and online learning are both expected to be widely adopted by schools in one year's time or less to encourage students to take ownership of their education by creating and provide them with ubiquitous access to digital tools, discussion forums, rich media and more. The time to adopt for robotics and virtual reality are estimated within two to three years, while artificial intelligence and wearable technology are expected to be mainstream in schools within four to five years. Horizon Report is published by CoSN, the association of district EdTech professionals, and the New Media Consortium.