AI model extrapolates beyond training data to predict diverse antimicrobial structures.
Category: robotics/AI – Page 171
A framework for soft mechanism driven robots
Soft robots excel in safety and adaptability, yet their lack of structural integrity and dependency on open-curve movement paths restrict their dexterity. Conventional robots, albeit faster due to sturdy locomotion mechanisms, are typically less robust to physical impact. We introduce a multi-material design and printing framework that extends classical mechanism design to soft robotics, synergizing the strengths of soft and rigid materials while mitigating their respective limitations. Using a tool-changer equipped with multiple extruders, we blend thermoplastics of varying Shore hardness into monolithic systems. Our strategy emulates joint-like structures through biomimicry to achieve terrestrial trajectory control while inheriting the resilience of soft robots. We demonstrate the framework by 3D printing a legged soft robotic system, comparing different mechanism syntheses and material combinations, along with their resulting movement patterns and speeds. The integration of electronics and encoders provides reliable closed-loop control for the robot, enabling its operation across various terrains including sand, soil, and rock environments. This cost-effective framework offers an approach for creating 3D-printed soft robots employable in real-world environments.
Soft mechanism driven robots, made via multi-material 3D printing, combine soft and rigid components for robust, adaptable locomotion. This framework balances flexibility and strength, enabling effective operation across varied terrains.
Sam Altman Announces GPT-5 Timeline Update, Cancels o3 As Standalone Model
In today’s AI news, OpenAI will ship GPT-5 in a matter of months and streamline its AI models into more unified products, said CEO Sam Altman in an update. Specifically, Altman says the company plans to launch GPT-4.5 as its last non-chain-of-thought model and integrate its latest o3 reasoning model into GPT-5.
In other advancements, Harvey, a San Francisco AI startup focused on the legal industry, has raised $300 million in a funding round led by Sequoia that values the startup at $3 billion — double the amount investors valued it at in July. The Series D funding round builds on the momentum and reflects investors’ enthusiasm for AI tools …
Meanwhile, Meta is in talks to acquire South Korean AI chip startup FuriosaAI, according to people familiar with the matter, a deal that could boost the social media giant’s custom chip efforts amid a shortage of Nvidia chips and a growing demand for alternatives. The deal could be completed as early as this month.
Then, AI took another step into Hollywood today with the launch of a new filmmaking tool from showbiz startup Flawless. The product — named DeepEditor — promises cinematic wizardry for the digital age. For movie makers, the tool offers photorealistic edits without a costly return to set.
In videos, join IBM’s Boris Sobolev as he explains how model customization can enhance reliability and decision-making of agentic systems. Discover practical tips for data collection, tool use, and pushing the boundaries of what your AI can achieve. Supercharge your AI agents for peak performance!
CEO and cofounder Andrew Feldman about his startup Then, Moderator Marc Pollefeys (Professor of Computer Science at ETH Zurich and Director of the Microsoft Mixed Reality and AI Lab in Zurich) leads an expert panel. The discussion will focus on advancements in robotics and the impact of embodied AI in complex, real-world scenarios. Speakers include; Marco Hutter (Director of the Robotic Systems Lab at ETH and Senior Director of Research at the AI Institute) Péter Fankhauser (Co-Founder & CEO at ANYbotics) Raquel Urtasun (Founder & CEO at Waabi and Professor of Computer Science at the University of Toronto).
We close out with, Databricks CEO Ali Ghodsi speaks exclusively with Worldwide Exchange Anchor Frank Holland about the new partnership announced on Thursday between the data analytics startup and the European tech giant.
Utility Engineering
As AIs rapidly advance and become more agentic, the risk they pose is governed not only by their capabilities but increasingly by their propensities, including goals and values. Tracking the emergence of goals and values has proven a longstanding problem, and despite much interest over the years it remains unclear whether current AIs have meaningful values. We propose a solution to this problem, leveraging the framework of utility functions to study the internal coherence of AI preferences. Surprisingly, we find that independently-sampled preferences in current LLMs exhibit high degrees of structural coherence, and moreover that this emerges with scale. These findings suggest that value systems emerge in LLMs in a meaningful sense, a finding with broad implications. To study these emergent value systems, we propose utility engineering as a research agenda, comprising both the analysis and control of AI utilities. We uncover problematic and often shocking values in LLM assistants despite existing control measures. These include cases where AIs value themselves over humans and are anti-aligned with specific individuals. To constrain these emergent value systems, we propose methods of utility control. As a case study, we show how aligning utilities with a citizen assembly reduces political biases and generalizes to new scenarios. Whether we like it or not, value systems have already emerged in AIs, and much work remains to fully understand and control these emergent representations.
Scientists develop novel self-healing electronic skin for health monitoring
Researchers have achieved a breakthrough in wearable health technology by developing a novel self-healing electronic skin (E-Skin) that repairs itself in seconds after damage. This could potentially transform the landscape of personal health monitoring.
In a study published in Science Advances, scientists demonstrate an unprecedented advancement in E-Skin technology that recovers over 80% of its functionality within 10 seconds of being damaged—a dramatic improvement over existing technologies that can take minutes or hours to heal.
The technology seamlessly combines ultra-rapid self-healing capabilities, reliable performance in extreme conditions, advanced artificial intelligence integration, and highly accurate health monitoring systems. This integration enables real-time fatigue detection and muscle strength assessment with remarkable precision.
New device helps people with spinal cord injuries grasp large, heavy objects
More than 15 million people worldwide are living with spinal cord injury (SCI), which can affect their sensory and motor functions below the injury level. For individuals with SCI between C5 and C7 cervical levels, this can mean paralysis affecting their limbs and limited voluntary finger and wrist flexion, making it difficult to grasp large, heavy objects.
Now, a team of UC Berkeley engineers from the Embodied Dexterity Group has developed a wearable device to enhance grasping functionality in this population. Dubbed the Dorsal Grasper, this assistive device leverages voluntary wrist extension and uses supernumerary robotic fingers on the back of the hand to facilitate human-robot collaborative grasping.
In a study recently featured in IEEE Transactions on Neural Systems and Rehabilitation Engineering, the researchers demonstrated for the first time how the Dorsal Grasper can expand users’ graspable workspace. Test subjects found that they could easily grasp objects anywhere they could reach their arm, without having to rotate their bodies, which can cause wheelchair users to lose their balance.
AI program plays the long game to solve decades-old math problems
A game of chess requires its players to think several moves ahead, a skill that computer programs have mastered over the years. Back in 1996, an IBM supercomputer famously beat the then world chess champion Garry Kasparov. Later, in 2017, an artificial intelligence (AI) program developed by Google DeepMind, called AlphaZero, triumphed over the best computerized chess engines of the time after training itself to play the game in a matter of hours.
More recently, some mathematicians have begun to actively pursue the question of whether AI programs can also help in cracking some of the world’s toughest math problems. But, whereas an average game of chess lasts about 30 to 40 moves, these research-level math problems require solutions that take a million or more steps, or moves.
In a paper appearing on the arXiv preprint server, a team led by Caltech’s Sergei Gukov, the John D. MacArthur Professor of Theoretical Physics and Mathematics, describes developing a new type of machine-learning algorithm that can solve math problems requiring extremely long sequences of steps. The team used their new algorithm to solve families of problems related to an overarching decades-old math problem called the Andrews–Curtis conjecture. In essence, the algorithm can think farther ahead than even advanced programs like AlphaZero.
Dual-domain architecture shows almost 40 times higher energy efficiency for running neural networks
Combining concepts from statistical physics with machine learning, researchers at the University of Bayreuth have shown that highly accurate and efficient predictions can now be made as to whether a substance will be liquid or gaseous under given conditions. They have published their findings in Physical Review X.
Observation of a glass of water reveals that the water exists in two distinct phases: liquid and gas. Even at room temperature, water molecules are constantly evaporating from the surface of the liquid water and passing into the gas phase. At the same time, some of the water molecules from the gas condense back into the liquid.
The transition from one phase to the other depends on temperature and pressure. Above a critical temperature, the simultaneous coexistence of gas and liquid disappears. The resulting supercritical fluid no longer forms an interface. This is important for industrial processes such as separation, cleaning and production.