Blogs

The hunt is on for anything that can surmount AI’s perennial memory wall–even quick models are bogged down by the time and energy needed to carry data between processor and memory. Resistive RAM (RRAM)could circumvent the wall by allowing computation to happen in the memory itself. Unfortunately, most types of this nonvolatile memory are too unstable and unwieldy for that purpose.Fortunately, a potential solution may be at hand. At December’s IEEE International Electron Device Meeting (IEDM), researchers from the University of California, San Diego showed they could run a learning algorithm on an entirely new type of RRAM.“We actually redesigned RRAM, completely rethinking the way it switches,” says Duygu Kuzum, an electrical engineer at the University of California, San Diego, who led the work.RRAM stores data as a level of resistance to the flow of current. The key digital operation in a neural network—multiplying arrays of numbers and then summing the results—can be done in analog simply by running current through an array of RRAM cells, connecting their outputs, and measuring the resulting current.Traditionally, RRAM stores data by creating low-resistance filaments in the higher-resistance surrounds of a dielectric material. Forming these filaments often needs voltages too high for standard CMOS, hindering its integration inside processors. Worse, forming the filaments is a noisy and random process, not ideal for storing data. (Imagine a neural network’s weights randomly drifting. Answers to the same question would change from one day to the next.) Moreover, most filament-based RRAM cells’ noisy nature means they must be isolated from their surrounding circuits, usually with a selector transistor, which makes 3D stacking difficult.Limitations like these mean that traditional RRAM isn’t great for computing. In particular, Kuzum says, it’s difficult to use filamentary RRAM for the sort of parallel matrix operations that are crucial for today’s neural networks.So, the San Diego researchers decided to dispense with the filaments entirely. Instead they developed devices that switch an entire layer from high to low resistance and back […]

Most 3D design software requires visual dragging and rotating—posing a challenge for blind and low-vision users. As a result, a range of hardware design, robotics, coding, and engineering work is inaccessible to interested programmers. A visually-impaired programmer might write great code. But because of the lack of accessible modeling software, the coder can’t model, design, and verify physical and virtual components of their system. However, new 3D modeling tools are beginning to change this equation. A new prototype program called A11yShape aims to close the gap. There are already code-based tools that let users describe 3D models in text, such as the popular OpenSCAD software. Other recent large-language-model tools generate 3D code from natural-language prompts. But even with these, blind and low-vision programmers still depend on sighted feedback to bridge the gap between their code and its visual output. Blind and low-vision programmers previously had to rely on a sighted person to visually check every update of a model to describe what changed. But with A11yShape, blind and low-vision programmers can independently create, inspect, and refine 3D models without relying on sighted peers.A11yShape does this by generating accessible model descriptions, organizing the model into a semantic hierarchy, and ensuring every step works with screen readers. The project began when Liang He, assistant professor of computer science at the University of Texas at Dallas, spoke with his low-vision classmate who was studying 3D modeling. He saw an opportunity to turn his classmate’s coding strategies, learned in a 3D modeling for blind programmers course at the University of Washington, into a streamlined tool. “I want to design something useful and practical for the group,” he says. “Not just something I created from my imagination and applied to the group.” Re-imagining Assistive 3D Design With OpenSCADA11yShape assumes the user is running OpenSCAD, the script-based 3D modeling editor. The program adds OpenSCAD features to connect each component of modeling across three application UI panels. OpenSCAD allows users to create models entirely through typing, […]

Boardroom priorities are shifting from financial metrics toward technical oversight. Although market share and operational efficiency remain business bedrocks, executives also must now manage the complexities of machine learning, the integrity of their data systems, and the risks of algorithmic bias.The change represents more than just a tech update; it marks a fundamental redefinition of the skills required for business leadership.Research from the McKinsey Global Institute on the economic impact of artificial intelligence shows that companies integrating it effectively have boosted profit margins by up to 15 percent. Yet the same study revealed a sobering reality: 87 percent of organizations acknowledge significant AI skill gaps in their leadership ranks.That disconnect between AI’s business potential and executive readiness has created a need for a new type of professional education.The leadership skills gap in the AI eraTraditional business education, with its focus on finance, marketing, and operations, wasn’t designed for an AI-driven economy. Today’s leaders need to understand not just what AI can do but also how to evaluate investments in the technology, manage algorithmic risks, and lead teams through digital transformations.The challenges extend beyond the executive suite. Middle managers, project leaders, and department heads across industries are discovering that AI fluency has become essential for career advancement. In 2020 the World Economic Forum predicted that 50 percent of all employees would need reskilling by 2025, with AI-related competencies topping the list of required skills.IEEE | Rutgers Online Mini-MBA: Artificial IntelligenceRecognizing the skills gap, IEEE partnered with the Rutgers Business School to offer a comprehensive business education program designed for the new era of AI. The IEEE | Rutgers Online Mini-MBA: Artificial Intelligence program combines rigorous business strategy with deep AI literacy.Rather than treating AI as a separate technical subject, the program incorporates it into each aspect of business strategy. Students learn to evaluate AI opportunities through financial modeling, assess algorithmic risks […]

Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.ICRA 2026: 1–5 June 2026, VIENNAEnjoy today’s videos! To train the next generation of autonomous robots, scientists at Toyota Research Institute are working with Toyota Manufacturing to deploy them on the factory floor.[ Toyota Research Institute ]Thanks, Erin!This is just one story (of many) about how we tried, failed, and learned how to improve ‪drone delivery system.Okay but like you didn’t show the really cool bit...?[ Zipline ]We’re introducing KinetIQ, an AI framework developed by Humanoid, for end-to-end orchestration of humanoid robot fleets. KinetIQ coordinates wheeled and bipedal robots within a single system, managing both fleet-level operations and individual robot behaviour across multiple environments. The framework operates across four cognitive layers, from task allocation and workflow optimization to VLA-based task execution and reinforcement-learning-trained whole-body control, and is shown here running across our wheeled industrial robots and bipedal R&D platform.[ Humanoid ]What if a robot gets damaged during operation? Can it still perform its mission without immediate repair? Inspired by self-embodied resilience strategies of stick insects, we developed a decentralized adaptive resilient neural control system (DARCON). This system allows legged robots to autonomously adapt to limb loss, ensuring mission success despite mechanical failure. This innovative approach leads to a future of truly resilient, self-recovering robotics.[ VISTEC ]Thanks, Poramate!This animation shows Perseverance’s point of view during drive of 807 feet (246 meters) along the rim of Jezero Crater on Dec. 10, 2025, the 1,709th Martian day, or sol, of the mission. Captured over two hours and 35 minutes, 53 Navigation Camera (Navcam) image pairs were combined with rover data on orientation, wheel speed, and steering angle, as well as data from Perseverance’s Inertial Measurement Unit, and placed into a 3D virtual […]

While quantum computers continue to slowly grind toward usefulness, some are pursuing a different approach—analog quantum simulation. This path doesn’t offer complete control of single bits of quantum information, known as qubits—it is not a universal quantum computer. Instead, quantum simulators directly mimic complex, difficult-to-access things, like individual molecules, chemical reactions, or novel materials. What analog quantum simulation lacks in flexibility, it makes up for in feasibility: quantum simulators are ready now.“Instead of using qubits, as you would typically in a quantum computer, we just directly encode the problem into the geometry and structure of the array itself,” says Sam Gorman, quantum systems engineering lead at Sydney-based startup Silicon Quantum Computing.Yesterday, Silicon Quantum Computing unveiled its Quantum Twins product, a silicon quantum simulator, which is now available to customers through direct contract. Simultaneously, the team demonstrated that their device, made up of 15,000 quantum dots, can simulate an often-studied transition of a material from an insulator to a metal, and all the states between. They published their work this week in the journal Nature.“We can do things now that we think nobody else in the world can do,” Gorman says. The Powerful ProcessThough the product announcement came yesterday, the team at Silicon Quantum Computing established its Precision Atom Qubit Manufacturing process following the startup’s establishment in 2017, building on the academic work that the company’s founder, Michelle Simmons, led for over 25 years. The underlying technology is a manufacturing process for placing single phosphorus atoms in silicon with subnanometer precision.“We have a 38-stage process,” Simmons says, for patterning phosphorus atoms into silicon. The process starts with a silicon substrate, which gets coated with a layer of hydrogen. Then, by means of a scanning-tunneling microscope, individual hydrogen atoms are knocked off the surface, exposing the silicon underneath. The surface is then dosed with phosphine gas, which adsorbs to the surface only in places where the silicon is […]