Physical AI

Harvard particle physicist Matthew Schwartz has released 36 research papers written in collaboration with the AI model Claude. The move is drawing attention in the physics community, where people are debating what mass AI-assisted publishing means for scientific authorship, peer review and the future role of human researchers in theoretical physics.

A Wall Street Journal opinion piece argues that regulation of the artificial intelligence industry should be grounded in physical reality rather than abstract policy frameworks. The author's central claim is that the physical constraints of computing, energy and infrastructure could serve as a more practical basis for governing AI development than rules written around speculative capabilities. The piece has sparked debate about how best to oversee a fast-moving technology.

Nvidia is pushing heavily into physical AI, technology that lets machines understand and act in the real world. The company says the approach could make robotaxis safer and speed up development of humanoid robots, positioning its chips and software platforms at the centre of the next wave of robotics beyond generative AI.

Amazon Web Services has released an open-source Physical AI Toolchain aimed at robotics developers. The toolchain is intended to help teams build and deploy AI-powered physical systems, and its open-source release makes the tooling freely available to the robotics community. Details on its full feature set have not yet been widely reported.

Nvidia is making a major push into what it calls physical AI, technology designed to let robots and autonomous vehicles operate safely in the real world. The company says the approach could make robotaxis more reliable and accelerate the development of humanoid robots. The announcement is drawing attention as a signal of Nvidia's ambitions beyond chips for data centers, extending into robotics and self-driving systems.

A robotics writer continues a 13-part series on social physical AI, arguing that AI systems' understanding of people should be treated as a continuous loop rather than one-shot predictions. The installment focuses on designing for repair: systems that notice misunderstandings and correct them. Readers in AI, design and robotics circles are engaging with the inclusive, human-centred framing.

The ninth instalment of a thirteen-part series on social physical AI asks a pointed question: if a medical AI delivers factually correct information, is that enough? The piece argues that even accurate information can be unsafe when delivered in the wrong context, and calls for social controls around AI systems alongside technical accuracy, touching on ethics, privacy and inclusive design.

Researchers have introduced a machine learning framework built on physics-based principles that predicts the lifespan of electric vehicle drive systems. By combining physical models of component degradation with data-driven learning, the approach aims to give manufacturers and owners more reliable estimates of how long drivetrains will last, supporting maintenance planning and design improvements in the fast-growing EV sector.

Basketball Hall of Famer Shaquille O'Neal says he would be willing to play a game of basketball against an AI-powered robot. The remarks come amid growing public discussion about whether advanced robots and AI systems could ever match human athletes in physical competition.

News outlets are explaining how AI-based hurricane forecast models differ from the traditional 'spaghetti' models long used to track storms. Traditional models come from physics-based computer simulations whose varied paths appear as spaghetti-like lines, while AI models learn patterns from historical weather data and can run faster and sometimes more accurately. The comparison matters as forecasters weigh how much to trust AI guidance during hurricane season.

Amazon Web Services has released an open-source toolchain aimed at physical AI, giving robotics developers tools to build, train and deploy AI systems that interact with the real world. The move positions AWS against rivals courting the robotics sector and gives developers a vendor-backed, freely available alternative for embodied AI development workflows.

A new report from Bioengineer.org claims that an AI tutoring system scored highly on a physics exam yet performed poorly in real classroom settings. The item highlights a gap between AI systems' ability to answer exam questions and their effectiveness at actually teaching students. Details about the system, the exam, or the classroom trials are limited to the headline.

The University of Delaware has launched a new institute focused on 'human-centered' artificial intelligence, with an initial aim of helping physical therapists treat patients. The initiative signals a growing push to apply AI in healthcare while keeping clinicians and patients at the center of decision-making. Details on funding and specific projects have not yet been widely reported.

Meta has acquired a robotics AI company in a move aimed at accelerating its development of humanoid robots, according to a report. The purchase signals Meta's intention to push further into physical AI systems, building on its artificial intelligence research to enter the robotics hardware space. Details on the acquired company and deal value have not been disclosed.

Fast Company explores how technologies born in video games, such as simulations and game engines, are paving the way for physical AI, where artificial intelligence is embodied in robots that interact with the real world. The piece examines how gaming tools can train and test embodied systems, a topic drawing attention as companies invest heavily in robotics and AI beyond screens.

TwelveLabs has released Pegasus 1.6, a video understanding model aimed at physical AI applications such as robotics. The company says the model can analyze video input to help machines and robots interpret real-world visual environments, extending multimodal AI beyond screen-based tasks into embodied systems operating in physical spaces.

Odyssey-3, a new world model, is being presented as a single AI base capable of controlling robots, autonomous cars and drones. The claim is that one shared model of the physical world can replace separate systems for each type of machine. Details on performance, backers and real-world deployments have not yet been widely reported.

TechCrunch examines Safeworld, a company working on generative AI-powered robots, and the public trust problem it faces. The central question is whether the company can persuade people that robots driven by generative AI will not hurt them, amid ongoing public anxiety about AI safety and physical robotics.

Commentary in the robotics press examines how nations pursuing 'sovereign AI' — building their own independent artificial intelligence capabilities — still end up reliant on the same underlying infrastructure providers. The piece links this tension to the emerging race in physical AI, where countries compete to develop robots and embodied systems, arguing that national ambitions are constrained by a shared set of landlords controlling compute, data centres and key technology layers.

Fast Company examines how decades of game development — simulated physics, virtual agents, and interactive 3D environments — could inform the next generation of embodied AI systems such as robots. The piece argues that the game industry's experience building believable worlds and training agents inside them offers a useful blueprint for teaching machines to act in the physical world.

Startup Hark has introduced an AI agent ahead of a planned hardware launch in 2027, according to Bloomberg. The company is positioning the software as a preview of its technology strategy, with the physical product still two years away. Details about the agent's capabilities, pricing, and the hardware itself remain limited in early coverage.

Researchers have introduced an AI agent designed to assist with physics research, with potential applications in quantum computing and dark matter detection. According to the announcement, the system could accelerate complex calculations and data analysis in these fields, where experiments generate enormous datasets and require highly precise modelling. Coverage so far is limited to the initial release, and details about the team, funding and validation remain sparse.

A new report examines the fast-growing industry built around safeguarding data centers, the physical facilities powering cloud computing and artificial intelligence. As operators pour billions into new sites, spending on security, fire protection, cooling and resilience has become a major business of its own, drawing attention from investors tracking the AI infrastructure boom.