XTHE

Meta has introduced a new AI model, Segment Anything Model 2 (SAM2), that can identify and track any object as it moves through a video. It builds on the first SAM, which only worked with images. This upgrade creates new possibilities for editing and analyzing videos.  

SAM2’s ability to segment objects in real-time is a major technical advance. It shows how AI can handle moving images and distinguish between different objects, even as they enter and exit the frame.  

To put SAM2’s achievement in context, segmentation means figuring out which parts of an image belong to which objects, and AI can do this, making it much easier to process or edit complex images. A capability that was the key feature of Meta’s first SAM. It has been used to segment coral reef images, analyze satellite imagery for disaster relief, and even analyze cellular images to detect skin cancer.  

Built on these image-based abilities, SAM2 expands segmentation to video, a shift not possible until now. To launch SAM2, Meta released a database of 50,000 videos used to train the model, along with 100,000 additional videos. Real-time video segmentation also requires significant computing power, so even though SAM2 is free now, that may change in the future.  

Segment Success 

With SAM2, video editors can select and modify objects in a scene more easily than with current editing tools, without having to adjust each frame by hand. Meta also sees SAM2 changing the interactive video landscape. People can select and move objects in live videos or virtual spaces using this AI model.  

Looking further ahead, Meta believes Sam2 could be important for developing and training computer vision systems, especially for self-driving cars. These systems need to track objects accurately and efficiently to understand and navigate their surroundings safely. The same could speed up labeling visual data, giving these AI systems better training material.  

While much of the AI video hype centers around generating videos from text prompts, the kind of editing capabilities provided by SAM2 might play an even bigger role in embedding AI into video creation. Models like OpenAI’s Sora, Runway, and Google Vue get a lot of attention for a reason, but SAM2 offers a different type of impact.  

While Meta may lead for now, other AI video developers are working on their own margins. For example, Google is testing video abstraction and object recognition features on YouTube. Adobe’s Firefly AI tools also focus on photo and video editing, with features like content overfill and auto reframe.

Source: Meta’s new AI model tags and tracks every object in your videos 

OpenAI has launched a new series of AI models designed to handle intricate reasoning in science, math, and coding. These models are built to spend more time thinking before answering, which helps them solve tougher problems with greater accuracy.  

The first model in this series, OpenAI 01 Preview, is now available through ChatGPT and the API. This release is part of an expanded strategy to roll out models better suited for complex logical tasks. It signifies a significant milestone in AI development, as the model presents remarkable progress in scientific, mathematical, and programming tasks.  

Advanced Reasoning Abilities 

The O1 Review Model is trained to spend more time thinking through problems than earlier versions. This helps it try different strategies, spot mistakes, and improve its answers as it works in tests. The text model in the series performed well on tough academic benchmarks, sometimes matching PhD-level skills, especially in physics, chemistry, and biology.  

The O1 Preview model also showed significant improvement in solving math problems. For example, in an international mathematics Olympiad qualifier, the earlier GPT-4-O solved 13% of the problems, while O1 Preview solved 83%. In core forces coding competitions, it scored in the 89th percentile, demonstrating a significant improvement over past models and reflecting the company’s commitment to responsible AI development. These new safety measures ensure the model adheres to the right guidelines more effectively through leveraging its reasoning abilities. For example, during tests intended to bypass the model’s safety rules, known as jailbreaking, the O1 Preview model attained a high compliance score, markedly outperforming previous models.  

OpenAI also improves its internal safety processes and works with AI safety institutes in the US and UK. These groups now have early access to research versions of the O1 model, allowing outside experts to test and review it. OpenAI aims to make future AI systems safer before releasing them to the public.  

Applications And Target Audience 

The O1 Preview model is especially well-suited for professionals working on complex science, coding, and math tasks. For instance, healthcare researchers can use it to annotate cell sequencing data. Physicists can rely on it to generate formulas concerning quantum optics, and developers can use it to build and execute integrated workflows. Its reasoning abilities make it a helpful tool for any field that requires in-depth problem-solving.  

To help developers, OpenAI has also released O1 Mini, a smaller and more affordable version of O1 Preview. The O1 Mini costs 80% less and is designed mainly for generating and debugging complex code. It does not have as much general knowledge as O1 Preview, but it is still a strong choice for tasks that need accurate reasoning at a lower price.  

Availability and Access 

Starting today, ChatGPT Plus and Team users can access both the O1 Preview and O1 Mini models through the model picker in ChatGPT.  
 
The current limits are:  

1. 50 queries per week for O1 Preview  

2. 50 queries per day for O1 Mini  

But OpenAI plans to raise these soon.  

ChatGPT Enterprise and Edu users will get access next week. However, the API for these models currently lacks features such as function calls and streaming, though these may be added in future updates.  

OpenAI has also expressed plans to make the O1 Mini available to ChatGPT-free users in the future. To broaden access, OpenAI plans to offer the O1 Mini model to ChatGPT-free users in the future, making its cutting-edge reasoning capabilities available to more people. OpenAI plans to launch Addison updates to the O1 Series, including features like browsing, file and image uploading, and more. These updates aim to make the models more versatile and more helpful to a broader audience.  

Along with the O1 Series, OpenAI will keep improving its GPT models so users can access the latest AI technology. By launching these RISBIN models, OpenAI shows its goal is to advance AI in terms of capability, safety, and real-world use.  

OpenAI Launches New Academy To Support AI Innovation In Low- And Middle-Income Countries 

OpenAI has launched the OpenAI Academy, a new program designed to support programmers and companies working with artificial intelligence. The Academy’s main goal is to help these groups address important challenges in their communities, especially in low- and middle-income countries. With this effort, OpenAI aims to ensure that the benefits of AI reach a wide range of groups worldwide, facilitating sustainable development and economic growth.  

Enabling Local Innovators With AI 

Many countries with growing technology sectors have skilled programmers and companies working to solve community problems. But these regions often lack access to advanced AI training, resources, and infrastructure. OpenAI wants to provide these tools and support so local talent can make a difference in areas like healthcare, agriculture, education, and finance.  

The OpenAI Academy will provide programmers and companies with the resources they need to deploy AI to concrete solutions. Participants will get access to advanced AI tools, training, and support from OpenAI experts. This will help them build their skills, improve their AI projects, and solve problems in their communities.  

Academy Offerings: Training API Credits And Community Assistance 

OpenAI is committing significant resources to the Academy to help these innovators succeed. Key components of the program include:  

1. Training and Technical Guidance: Participants will receive expert support from OpenAI to help them use AI effectively to solve complex problems.  

2. API Credits: OpenAI is offering $1M in API credits to help developers and companies use its models in their projects. This will make it easier for more people to start working with AI, join forces on projects, and drive collective innovation. This will create a supportive environment where participants can learn from one another and develop new AI applications.  

3. Contests and incubators column: The academic plans to partner with philanthropists to create contests and incubators that will further invest in organizations on the front lines of their communities. These contests will reward organizations for developing AI-based solutions that tackle urgent local challenges.  

By building a platform for partnership and development, OpenAI wants to increase the global impact of AI. The Academy’s goal is to ensure AI benefits reach everywhere, especially where they are needed most.  

Success Stories and Global Impact 

OpenAI has already helped AI developers and groups working for social good. For example, KOBI, which won the OpenAI prize at the Tools competition, created an AI tool to help students with dyslexia learn to read. iStem, a winner of the Turn.io chat for impact context contest, uses AI to make content more accessible to people with vision impairments in India. OpenAI supported these groups with API credits and technical support, helping them grow their projects and keep making a difference.  

These stories show how AI can help solve global problems, from better education to making things more accessible for marginalized groups. The OpenAI Academy aims to build on these successes by supporting even more programmers and institutions worldwide.  

Expanding Language Access For AI Development 

To make AI more accessible, OpenAI has made sure that AI resources are available in many languages. The company recently funded and released a professional translation of the massive metadata Language Understanding (MM value) benchmark into 14 languages, including Arabic, Bengali, Hindi, Swahili, and Yoruba. This helps more people use AI in their own languages.  

A Step Towards Global AI Inclusivity 

OpenAI’s new initiative is an important step toward making AI development increasingly inclusive. It makes sure that those who understand the unique cultures and challenges of their regions can tailor AI applications to meet local needs by supporting programmers and institutions across diverse regions. OpenAI plans to bring the benefits of AI to a larger audience, enabling communities worldwide to use technology for economic growth and to tackle challenges.  

The OpenAI Academy is ready to help innovators worldwide use AI to solve big challenges and create new chances for responsible growth and technology progress.

Source: OpenAI Unveils New Reasoning Model Series: OpenAI o1-Preview 

Project Kuiper is Amazon’s low-Earth-orbit satellite network. Its goal is to close the digital gap by bringing fast, affordable internet to communities without reliable access through traditional means.  

To get online with Kuiper, customers install a rugged outdoor antenna called a customer terminal that connects directly with satellites racing overhead. Past versions of this equipment were often too bulky, complex, or expensive, preventing LEO satellite networks from fully realizing their vision. Now, Kuiper aims to change the game.  

Project Kuiper targets tens of millions of customers, setting an early goal to build a customer terminal for under $500. Engineers met this in 2020 by inventing a compact, lightweight antenna. They’ve continued refining designs to make terminals even smaller, cheaper, and more powerful.  

Building on these advancements, Amazon recently shared the results of its ongoing work. Read on to discover more about the small, powerful antennas and the technology that makes them possible.  

Continuing this momentum, at a satellite industry conference in Washington, D.C., Amazon provided a first look at three engineering models that will anchor its customer terminal portfolio.  

The standard customer terminal is for homes and small businesses. It is compact, less than 11 inches square, 1 inch thick, under 5 lb, and it will deliver up to 400 Mbps. Amazon expects to produce it for under $400.  

An Ultra-Compact Design To Help Connect Even More Customers 

The 11-inch-square design will be Project Kuiper’s smallest and most affordable customer terminal, weighing just 1 pound and offering speeds up to 100 Mbps. It offers portability and affordability, creating opportunities to serve even more customers worldwide. This design will connect residential customers who need an even lower-cost model, as well as government and enterprise customers pursuing applications such as ground mobility and the Internet of Things (IoT).  

Project Kuiper’s largest terminal meets demanding needs for businesses, government, and telecom users. It measures 19 by 30 inches and delivers speeds up to 1 Gbps.  

“Our mission with Project Kuiper goes beyond simply connecting the unserved and underserved it’s about delivering outstanding quality, reliability, and value,” said Rajiv Badyal, Amazon’s Vice President of Technology for Project Kuiper. “Every choice has pushed us to raise the bar for experience, and our diverse terminals prove our commitment.”  

Powered by Amazon Design Custom Chips 

Project Kuiper customer terminals are powered by an Amazon-designed baseband chip, code-named Prometheus. This chip combines:  

• The processing power of a 5G modem found in modern smartphones  

• The ability of a base station to handle thousands of users simultaneously  

• The strength of a microwave backhaul antenna for robust point-to-point connections  

All of these features are combined into one custom chip.  

Format as: Powers Project Kuiper customer terminals, satellites, and ground gateway antennas enabling each satellite to process up to 1 terabit per second (Tbps) of traffic.  

Getting Ready To Offer Commercial Service 

Drawing from its history of mass device production, Amazon is leveraging this expertise as Project Kuiper designs and produces its customer terminals. To meet the anticipated demand, the team is expanding infrastructure to prepare for producing tens of millions of units.  

Following these preparations, Project Kuiper is getting ready to launch its first two prototype satellites on the first flight of the United Launch Alliance (ULA) Vulcan Center rocket. This mission will provide engineers with real-world data on how the systems operate in space and will allow them to test the full communications network simultaneously.  

Project Kuiper is expanding its operations to prepare for commercial services. For instance, the team has started building a dedicated satellite production facility in Kirkland, Washington, and plans to begin mass-producing satellites by the end of 2023. Project Kuiper aims to launch the first production satellites in the first half of 2024 and plans to offer this service to its earliest customers later that year. 

Source: Here’s your first look at Project Kuiper’s low-cost customer terminals 

After visiting 24 of the 63 US National Parks, I have learned how rewarding it is to explore the outdoors with family and friends and encountered challenges like finding the best trailheads or bathrooms.  

Google Maps added four updates to simplify planning your next park trip and help you find key information.  

Quickly See What A Park’s Must-Do Attractions Are 

Popular spots, such as attractions and trailheads, are now easier to find thanks to community input.  

Search for a park and view photo highlights. Tap for details and reviews to plan your trip quickly.  

See The Popular Trails From Beginning To End In Maps 

When you search for a trail, its route is highlighted, and you get community info on type and difficulty.  

Find Your Way Around With More Detailed Directions 

After selecting places and trails to visit, the maps will provide accurate directions. This month, park entrances will be highlighted, and maps will guide you directly to trailheads with hiking and biking directions.  

Take Google Maps Offline With You, No Matter Which Park You Visit 

Cell service can be unreliable in parks. Now, download offline maps to navigate without an internet connection.  

Planning your next adventure? Here are the top-rated national parks in the US based on Google Maps data:  

• Great Smoky Mountains National Park  

• Bryce Canyon National Park  

• Glacier Bay National Park and Reserve  

• Canyons of the Colorado National Park  

• Rocky Mountain National Park  

• Acadia National Park  

• Badlands National Park  

• Mount Rainier National Park  

• Carlsbad Caverns National Park  

• Grand Teton National Park  

These updates launch in April for all US national parks and will expand globally. Spend less time planning and more time exploring.

Source:  4 Maps updates to help you explore U.S. national parks

Microsoft has expanded its Azure Space program by leveraging SpaceX’s Starlink and Azure Orbital Cloud Access to deliver fast, direct-to-cell connections to U.S. first responders and public sector groups. This update delivers satellite, cellular, and fiber networks together so people working in remote areas can get secure, reliable, high-speed communications. Tests with the National Interagency Fire Center have shown how well this works.  

Primary Updates and Capabilities 

• Azure Orbital Cloud Access is currently being offered for user testing (“in preview”) and combines Starlink satellites with software-defined terrestrial networks. This setup helps deliver connections with minimal time delays (low latency).  

• 1st Responder Support: This service helps emergency teams stay connected, for example, by linking firefighters in remote locations to FireNet, a cloud-based app used to manage wildfires.  

• Direct-to-Cell & IoT: The system enables users and IoT devices to communicate securely and directly, even when regular networks are down.  

• Satellite partnerships: Along with SpaceX, Microsoft partners with ACS, Intel, and K, set to provide secure global satellite services across multiple orbits.  

• Security focus: This platform is built to meet strict US government security standards, leveraging AI and threat intelligence to guard against cyber threats.  

This expansion is part of Microsoft’s broader effort to make Azure the top choice for space-based data-intensive workloads and cloud-powered connectivity.  

Microsoft has provided an update on its Azure Space Platform, extending cloud services to remote areas and soon launching private previews of Azure Orbital Cloud Access for fast global cloud connectivity.  

Azure Orbital Cloud Access uses Jupiter Networks’ software-defined wide-area network (WAN) technology, which manages large-scale computer networks, to connect satellites to both wireless and fiber-optic networks.  

Microsoft began its Azure Space Initiative two years ago to position Azure as a key contender in the space and satellite cloud market.  

Last year, Microsoft released Azure Orbital for user testing (called ‘Preview’), letting customers control and communicate with Microsoft and partner satellites from ground facilities worldwide, without incurring additional costs to send data to Azure’s server network. The original service is now named Azure Orbital Ground Station, which allows Microsoft to offer various services, including the new Azure Orbital Cloud Access.  

Microsoft has also announced that, as your orbital ground station, its first ground station-as-a-service is generally available in partnership with Microsoft. The service enables satellite operators to focus on their satellites and use the cloud more reliably, reducing costs and latency, accelerating time-to-market, and enhancing security with Azure.  

Microsoft Azure Announces Several New Space Partnerships 

Azure Orbital products with partners, including:  

• Airbus  

• American Vol  

• Aero Space  

• Black Shark AI  

• ESRI  

• Unit Packet  

• Enterprise  

• iDirect  

• Intelsat  

• Kratos  

• KSAT  

• Loft Orbital  

• Lupia  

• Omni Space  

• Orbital Insights  

• SES  

• Sky Watch  

• SpaceX  

• Thales  

• Alenia Space  

• USA Electro Dynamics  

• Viasat  

• Explore  

By teaming up with these partners, Microsoft aims to bring satellite-based communications to enterprise cloud operations. Azure’s integration of 5G and satellite technology is designed to help satellite vendors transition from analog to digital systems.  

Microsoft has expanded its partnership with SES to launch the Satellite Communication Virtualization Program. This initiative will create a fully virtualized satellite ground network that supports a range of network functions and edge applications.  

The program will help industry players adopt standards more quickly and enable remote updates. It also simplifies introducing services like network slicing, new virtual functions, and edge cloud applications.  

Microsoft’s collaboration with KSAT and PIXEL centers on analyzing satellite data in Microsoft’s cloud, with a focus on services that derive insights from space-based observations.  

Microsoft provided an update on its partnership with Loft Orbital. They are developing an on-orbit computing system to build, test, and validate space software in Azure. They deploy applications to satellites using Loft’s infrastructure.  

The first Loft Satellite with Azure capabilities will launch next year, enabling governments and companies to deploy software applications directly to space hardware via Azure.  

Microsoft’s sustainability product team is working with Muon Space to develop enterprise ESG analytics products based on Muon’s Earth Systems data.  

Microsoft Starlink Combines Cloud Computing And Satellite Connectivity 

Microsoft is expanding space by partnering with SpaceX to merge Azure with Starlink’s satellite internet. The new Azure Orbital Cloud Access Preview for government customers will prioritize network traffic using SpaceX’s satellites and Azure Edge devices.  

Amazon Web Services, Microsoft’s main cloud competitor, launched its Aerospace and Satellite Solutions Unit and created AWS Ground Station and Project Kuiper, direct competitors to SpaceX’s Satellites in cloud and satellite connectivity.  

Google, owned by Alphabet, also partnered with SpaceX to connect Starlink satellites to Google Cloud, further advancing cloud-satellite integration.  

Alibaba is expanding its cloud reach in the Asia Pacific, launching new data centers to strengthen its presence in a competitive global market.  

Free Report: Learn How to Benefit From the Growing Electric Vehicle Industry 

Globally, electric car sales continue their remarkable growth, even after breaking records in 2021. High gas prices have increased this time, but so have the growing comfort features and technology of EVs, so the passion for EVs will be around long after gas prices normalize. Not only are manufacturers seeing record-high profits, but producers of EV-related technology are raking in the dough as well.  

Do you know how to cash in? If not, we have the perfect report for you, and its free today. Don’t miss your chance to download Zack’s top five stocks for the electric vehicle revolution at no cost and with no obligation. 

Source:  Microsoft (MSFT) Expands Azure Space Connectivity Offerings

Apple’s private cloud compute is a new, highly secure system for running complex AI tasks that go beyond what your device can handle. It is built to ensure user data is never kept or made inaccessible by Apple.  

With verifiable transparency, Apple enables external security experts to audit the exact code executing on silicon-based servers. This mechanism transitions privacy from a stated assurance to a technically verifiable property.  

How Apple’s Private Cloud Compute Ensures Deletion 

PCC operates as a stateless architecture, processing data only for the duration of each request and ensuring no persistence post-completion.  

• Cryptographic erasure: Data is destroyed immediately after fulfillment of the request. Through cryptographic erasure, recovery is invisible even if server hardware is compromised.  

• No Administrative Access: The design restricts even Apple administrators from accessing user data during processing.  

• Secure Enclave and Verified Boot: PCC integrates security features from iPhone and Mac, such as Secure Enclave and Secure Boot, to ensure only authenticated code executes on servers.  

How To Verify Data Deletion (Transparency Logs) 

Apple gives researchers and, in some cases, users ways to check these privacy claims:  

1. Audit logs column: Apple publishes immutable logs detailing all software deployed on PCC, ensuring tamper-evident record keeping.  

2. Virtual research environment (VRE): security researchers leverage a VRE, functioning as a private cloud compute node on an Apple silicon Mac, to verify software integrity and ensure no data is stored.  

3. Device-level verification: Devices verify, via cryptographic attestation, that a server is running code that matches transparency log entries before initiating communication.  

How Users Can View Their Own Data Usage 

While experts check the code, regular users can keep track of their own data used through the Settings app:  

• To view your data usage, open the Settings app. Tap Privacy and Security, then tap Apple Intelligence Report.  

• Functionality: This feature lets users create a report of requests sent to the private cloud compute, showing data for the last 15 minutes or the last 7 days.  

• Export Capability Column: Users can export the report as a .json file for closer inspection.  

Note: The Apple Intelligence report might be empty if no requests were sent to the cloud, since many tasks are handled locally.  

Apple Intelligence is our personal intelligence system that brings generative models to iPhone, iPad, and Mac for features that require handling complex data with larger models. We developed Private Cloud Compute (PCC), a new cloud intelligence system designed for private AI processing. PCC stands out by combining custom Apple silicon and a secure operating system to deliver end-to-end security, ensuring that personal user data sent to PCC is only accessible to the user, not even to Apple. This extends Apple’s device-level. We have developed privacy standards for the cloud, setting PCC apart from standard cloud AI approaches. We believe it is the most advanced security architecture ever created for cloud AI computing at scale.  

Apple focused on device processing to keep user data secure and private. When user data is managed in the cloud, we use security measures such as end-to-end encryption or temporary processing with random identifiers to protect user privacy.  

Making secure, private AI processing in the cloud is a challenge. Data centers use powerful AI hardware for complex machine learning models, but this requires unencrypted access to user requests, making end-to-end encryption unworkable. As a result, Cloud AI must rely on traditional security methods, which present key challenges:  

• Verifying and enforcing privacy in Cloud AI is difficult. If a service claims not to log data, this is hard to confirm. Software changes can introduce logging without detection, and load balancers might log many user requests during troubleshooting.  

• Delivering runtime transparency for AI in the cloud is difficult. Cloud AI services are often unclear about the software they use, and these details are usually kept private. Even if a service uses only open-source software that researchers can inspect, there is no common way for a user, device, or browser to confirm that it is connecting to an unmodified version of the software or to notice if the software has changed.  

• Limiting privileged access in cloud AI environments is difficult. Operations require ongoing monitoring, and during incidents, administrators use tools like SSH. Even with restrictions, enforcing access limits is hard. For example, administrators may inadvertently copy confidential data and steal credentials. Risk of user data theft. Apple devices like the iPhone and Mac can handle computation locally. The security and privacy benefits are obvious. Users control their own devices. Researchers can inspect both hardware and software, and secure boot ensures runtime transparency. Apple does not have privileged access. For example, the data protection file encryption system prevents Apple from disabling or guessing an iPhone passcode.  

However, to handle advanced requests, Apple Intelligence sometimes needs to use larger models in the cloud to meet our users’ security and privacy standards. We are extending our device security approach to the cloud.  

When PCC is available in beta, we will share more details and address researchers’ questions in our next post.

Source: Private Cloud Compute: A new frontier for AI privacy in the cloud 

News Summary 

• The Newton Physics Engine, now in NVIDIA ISAAC Lab, helps researchers and developers build more capable and adaptable robots.  

• The new ISAAC GR00T Open Foundation model enables robots to reason like humans. It breaks complex tasks into simpler steps and uses prior knowledge and common sense to finish them.  

• NVIDIA COSMOS world-based models let developers generate diverse data, enabling faster, scale-out training of physical AI models.  

• Researchers at top universities such as Stanford, ETH Zurich, and the National University of Singapore are using NVIDIA’s accelerated computing and software to advance robotics research.  

• Top robot developers such as Agility Robotics, Boston Dynamics, Disney Research, Figure AI, Franka Robotics, Hexagon, Skild AI, Solomon, and Techman Robot are now using NVIDIA ISAAC and Omniverse technologies.  

NVIDIA announced today that the open-source Newton Physics Engine is now available in NVIDIA ISAAC Lab, along with the open NVIDIA ISAAC GR00T and 1.6 reasoning, vision, language, and action model for robot skills and new AI infrastructure. These technologies give developers and researchers an open first robotics platform that speeds up iteration, standardizes testing, unifies training with on-robot inference, and helps robots safely and reliably transfer skills from simulation to the real world.  

Humanoids are the next frontier of physical AI, requiring the ability to reason, adapt, and act safely in a volatile world, said Rev. Lebaredian, vice president of Omniverse and Simulation Technology at NVIDIA. With these latest updates, developers now have the three components to bring robots from research into everyday life:  

• ISAAC GR00T as the robot’s brain  

• Newton as their body  

• NVIDIA Omniverse as their training ground  

Newton Opens New Standard For Physical Simulation In Robotics 

Robotics learn faster and more safely in simulation, but humanoid robots with their complex joints, balance, and movements challenge today’s physics engines. More than a quarter of a million robotics developers worldwide need accurate physics so that the skills they teach robots in simulation can be used safely and reliably in the real world.  

NVIDIA announced today the beta release of Newton, an open-source GPU-accelerated physics engine managed by the Linux Foundation. Newton is built on the NVIDIA Warp and OpenUSD frameworks and was developed in collaboration with Google DeepMind, Disney Research, and NVIDIA. It is available now.  

Thanks to Newton’s flexible design and support for various physics solvers, developers can now simulate very complex robotic actions, such as walking through snow or gravel and handling cups and fruits, and then successfully deploying them in the real world.  

Recent adopters of Newton include respected research labs and universities like ETH Zurich Robotic Systems Lab, Technical University of Munich, and Peking University, as well as robotics company Lightwheel and simulation engine company Style3D.  

Cosmos Reason Enhances Robot Thinking in the New Open Isaac Gr00t N1.6 Model 

Robots need to interpret uncertain instructions and adapt to unfamiliar situations to handle real-world tasks.  

The newest ISAAC GR00T N1.6 robot foundation model, coming soon to Hugging Face, will include NVIDIA Cosmos Reason, an open, customizable reasoning vision-language model designed for physical AI. Cosmos Reason works like the robot’s brain, turning implicit instructions into step-by-step plans. It uses prior knowledge, common sense, and physics to help robots handle new situations and perform a wide range of tasks.  

Cosmos Reason has been downloaded over 1 million times. It currently ranks highest on the physical reasoning leaderboard on Hugging Face. It can also organize and label large sets of both real and synthetic data for model training. Cosmos Reason 1 is now available as a simple NVIDIA NIM microservice for deploying AI models.  

ISAAC GR00T N1.6 now allows humanoid robots to move and handle objects simultaneously. This gives them more freedom in their torso and arms, making it easier to do challenging tasks such as opening heavy doors.  

Developers can further train GR00T models with the open-source NVIDIA physical AI dataset on Hugging Face. Users have downloaded this dataset over 4.8 million times, and it now offers thousands of artificial and real-world examples.  

Top robot makers like AeiRobot, Franka Robotics, LG Electronics, Lightwheel, Mentee Electronics, Neura Robotics, Solomon Techman Robot, and UCR are testing ISAC GR00T-N models to build general-purpose robots.  

Launching New Cosmos World Core Models for Physical AI Development 

NVIDIA has announced new updates to its Open Cosmos WFM (World Foundation Models), which have been downloaded over 3 million times. These updates let developers generate a wide range of data to speed up the training of physical AI models at scale using text, image, and video prompts.  

• Cosmos Predict 2.5, coming soon, combines the strengths of three Cosmos WFMs into a single model. This reduces complexity, saves time, and improves efficiency. It can generate longer videos up to 30 seconds and supports multi-view camera outputs for world simulations.  

• Cosmos Transfer 2.5, also coming soon, is 3.5 times smaller than the previous and delivers faster, higher-quality results. It can create photo-realistic synthetic data, meaning lifelike images and scenes generated by computers from 3D simulations and spatial control inputs such as depth segmentation, edges, and high-definition maps.  

A New Workflow For Teaching Robots To Grasp Objects 

Teaching a robot to pick up an object is one of robotics’ hardest challenges. It goes beyond moving an arm; it turns concepts into precise actions, a skill robots learn through trial and error.  

The new Dexterous Grasping workflow in the developer preview of iSight Lab 2.3, built on the NVIDIA Omniverse platform, trains robots with multi-fingered hands and articulated arms in a simulated computer-generated environment. Using an automated series of progressively more challenging training tasks, the workflow adjusts technical factors such as gravity (which affects how objects fall), friction (which determines how surfaces slide against each other), and object weight, helping robots learn skills in unpredictable settings as well.  

Boston Dynamics Atlas robots used this workflow to learn grasping, which greatly improved their ability to manipulate objects. Top robotics developers such as Agility Robotics, Boston Dynamics, Figure AI, Hexagon, Skild, AI Solomon, and Techman Robot are now using NVIDIA ISAAC and Omniverse technologies.  

Testing Learned Robotic Robot Skills in Simulation 

Teaching a robot a new skill, such as picking up a cup or walking across a room, is very challenging. Testing these skills on a real robot takes a lot of time and can be costly.  

The solution lies in simulation, which enables testing a robot’s learned skills across countless scenarios, tasks, and settings. Even in simulation, developers tend to build fragmented, simplified tests that do not reflect the real world. A robot that learns to steer a perfect, simple simulation will fail the moment it faces real-world complexity.  

NVIDIA and Lightwheel are joining forces to help developers run complex, large-scale simulation tests without having to start from scratch. ISAAC Lab Arena, their freely available framework for scalable experiments and standardized testing, will be available soon.  

New NVIDIA AI Infrastructure Enables Robotics Workloads Anywhere 

NVIDIA announced an AI infrastructure designed for the most challenging workloads, enabling developers to leverage advanced technologies and software libraries.  

NVIDIA GB200/NVL72 is a rack-scale system with 36 Grace CPUs and 72 Blackfire GPUs. Major cloud providers are using it to speed up AI training and inference. Tasks include intricate reasoning and physical AI tasks.  

• NVIDIA RTX Pro Servers, which offer a single hardware and software architecture for every robot development workload, including: Model training, Teaching robots to perform tasks using data, Synthetic data generation (creating simulated data to use in development and testing), Robot teach learning (enabling robots to improve at tasks through algorithms), Simulation (virtually testing robots before real-world deployment). RTX PRO servers have been adopted by the RAI Institute.  

• NVIDIA’s Jetson Thor, powered by a Blackwell GPU, lets robots run multiple AI workflows, such as intelligent interactions, and enables real-time on-robot inference. This is a major step forward for high-performance physical AI tasks and applications, such as humanoid robotics. Partners such as Figure AI, Galbot, Google DeepMind, Mentee Robotics, Meta, Skild AI, and Unity have adopted Jetson Thor.  

NVIDIA Advances Progress in Robotics Research 

NVIDIA technologies, including GPUs, simulation frameworks, and CUDA-accelerated libraries, were referenced in nearly half of CoRL’s accepted papers, with use across leading research labs and institutions such as Carnegie Mellon, University of Washington, ETH Zurich, and National University of Singapore.  

CoRL also featured BEHAVIOR, a Robotic Learning Benchmark project from the Standard Vision and Learning Lab, and Taccel, a high-performance simulation platform for advancing vision-based tactile robotics developed by Peking University.  

Find out more about NVIDIA’s robotics research at CORL, taking place from September 27 to October 2 in Seoul.

Source: NVIDIA Accelerates Robotics Research and Development With New Open Models and Simulation Libraries 

Story Highlights 

• Dell is reviving its XPS laptops with the new XPS 14 and XPS 16, and will bring back the XPS 13 later this year at its most affordable price yet.  

• Alienware will double its laptop lineup to reach more gamers, introducing anti-glare OLED displays and new Intel Core Ultra 200HX processors, as well as AMD 2nd Gen Ryzen 9000X3D processors for desktops.  

• Dell Alienware is launching two new monitors, including the world’s first 52-inch 6K display.  

Dell Technologies is growing its consumer and gaming lineup by reviving the XPS laptop line with new designs and pricing, expanding Alienware’s range, and debuting two new ultra-sharp monitors.  

Perspectives 

Jeff Clark, Vice Chairman and Chief Operating Officer at Dell Technologies, said, “We’re getting back to our roots with a fresh focus on consumer and gaming. XPS is back, better than ever, with a complete redesign that delivers outstanding craftsmanship in our thinnest, lightest form factors yet. We’re also bringing back the XPS 13 as our most accessible XPS ever. In gaming, we’re building on recent momentum and effectively doubling Alienware’s notebook lineup. These moves are about broadening our portfolio and expanding our coverage so we can reach more customers with the best products at every price point.”  

XPS Returns With Complete Redesign 

Dell re-introduces the XPS line with a streamlined design for the first time. The XPS logo appears on the cover. The XPS 14 and XPS 16 feature CNC-machined aluminum, improved interfaces, sharp displays, and strong performance. They lead the industry in battery life, up to 27 hours of Netflix or over 40 hours of local video playback.  

These are Dell’s thinnest laptops, measuring just 14.6 mm. The XPS-14 weighs roughly 3 lb, more than half a pound lighter than the previous generation, and the XPS-16 weighs almost 3.6 lb, lighter than its predecessor. Available with Tandem OLED (organic light-emitting diode) display options and Intel Core Ultra Series 3 processors. The built-in Intel Arc graphics feature 12 XE cores, specialized processing units that enhance graphics performance. The new XPS line delivers impressive visuals with a balanced focus on portability and performance.  

Later this year, Dell will add more XPS models in new styles and prices, including the XPS 13, the thinnest, lightest, and most affordable XPS yet.  

Alienware Expands to Reach More Gamers 

Alienware expands its laptop lineup with an ultra-slim 17mm gaming laptop combining high performance and portability, and a new entry-level laptop for greater affordability. These updates aim to broaden Alienware’s reach across gamer types.  

With these additions, Alienware offers choices for everyone, from dedicated and casual gamers to those with hybrid interests.  

Alienware debuts antiglare OLED displays on the 16 Area 51 and 16X Aurora laptops, delivering stunning OLED visuals with breakthrough anti-glare technology. Responding to one of the community’s top requests, these displays let gamers experience the deep blacks, vibrant shades, and exceptional contrast of OLED in any lighting condition. These laptops, along with the Alienware 18 Area 51, feature the powerful new Intel Core Ultra 200HX processors. Additionally, the Alienware Area 51 desktop is equipped with AMD’s new Ryzen 9850X 3D processor and 3D V-Cache technology, pushing the ceiling of gaming performance.  

Industry-Leading Dell UltraSharp Monitors 

Dell is introducing two new monitors.  

The Dell Atreshop 52 Thunderbolt Hub monitor is the world’s first 52-inch 6K monitor and the first monitor to achieve the highest tier 6 of TUV Rhineland Low Blue Light certification. It is designed for financial traders, data scientists, engineers, and executives who need maximum screen real estate without a multi-monitor setup.  

The Dell UltraSharp 32 4K QD OLED Monitor, ACS 2026 Innovation Award honoree, is the world’s first commercial display HDR True Black 500 QD OLED monitor with anti-glare, low-reflectance (AGLR) technology. Two years ago, Alienware brought QD-OLED to gamers. Now, with enhanced panel efficiency, improved HDR performance, and anti-glare technology, Dell is bringing QD-OLED to creative professionals who demand perfect color precision in any lighting condition.  

For details, including pricing and availability, visit www.dell.com/ces2026.

Source: DELL TECHNOLOGIES AT CES 2026: XPS Returns, Alienware Expands and UltraSharp Monitors Set New Standards 

News Summary 

• AMD has launched the Ryzen AI 400 and Pro 400 series of processors. The Ryzen AI 400 series targets consumer and gaming systems, offering up to 60 NPU TOPS for Co-Pilot + PCs and advanced AI features, while the Pro 400 series is tailored for business applications.  

• The Ryzen AI Max models extend these capabilities, bringing powerful AI and graphics to slim laptops, workstations, and compact devices for creators, gamers, and developers.  

• AMD introduced the Ryzen AI Halo, a mini PC, distinct from the processors above. Halo runs on Ryzen AI Max + performance and targets AI developers who need an out-of-the-box mini PC platform for rapid edge AI development.  

• In addition to its AI lineup, AMD announced the Ryzen 7 9850X 3D, the company’s fastest gaming processor, built on Zen 5 and AMD 3D V-Cache technologies. Unlike AI-oriented chips, this processor is purpose-built for peak gaming performance and is designed exclusively for gamers.  

• Alongside these product launches, AMD announces robust yearly growth in OEM adoption of Ryzen AI processors, with more consumer, business, and gaming systems set to launch through 2026.  

• The new ROCm 7.2 software for Windows and Linux now supports Ryzen AI for 400 series of processors for use with Comfy UI.  

Highlighting its ongoing efforts, AMD introduced its latest mobile and desktop processors at CES 2026, emphasizing expanded AI capabilities, gaming performance, and business features across a broader range of systems.  

Continuing the business focus, AMD also introduced the Ryzen AI Pro series, specifically for business laptops. This series features AI acceleration, robust security, and management capabilities distinct from consumer-focused models.  

With AI now central to PCs, AMD is expanding its hardware and software lineup to accelerate AI use. For example, the new Ryzen AI Halo platform’s ROCm7.2 support for Ryzen AI 400 series and an AI bundle for Adrenalin Edition make AI development and use easier and faster.  

For gamers, AMD is introducing a new top gaming CPU, the Ryzen 7 9850X3D. It improves on the Ryzen 7 9800X3D by 400 MHz, offering even better gaming performance. Ready on users will also get FSR Redstone, which adds machine running frame generation and upscaling to the latest AAA games, moving AMD closer to a complete AI computing platform.  

AI is changing what PCs can’t do, and AMD is at the forefront, said Jack Huynh, Senior Vice President and General Manager of AMD’s Computing and Graphics Group. We are making smarter, faster, and more compelling experiences possible for everyone.  

Introducing the AMD Ryzen AI 400 series and AMD Ryzen AI Pro 400 series 

AMD launches Ryzen AI 400 and Pro 400 series for consumer and business Co-Pilot + PCs. Built on XEN5 and XDNA2 NPUs, they deliver up to 60 TOPs of NPU AI compute and meet Co-Pilot + PC requirements. With up to 12 CPU cores, built-in Radeon 800M graphics, and faster memory, these chips offer strong performance, long battery life, and smart features across many devices.   

The Ryzen AI Pro 400 series is designed for businesses and modern IT needs. It offers strong performance, advanced security with AMD Pro technologies, easy management, and sustained reliability. These benefits translate to enhanced data protection, simplified service deployment, and reduced downtime for IT teams. IT teams can confidently upgrade their systems, knowing these processors offer reliable enterprise-level features and the same AI capabilities as the rest of the Ryzen AI Pro 400 series. This gives business users a consistent, high-quality experience and IT decision makers great performance and value.  

With the latest Ryzen AI processors, AMD is helping move AI PCs from early adopters to everyday tools by providing fast AI-powered services such as real-time translation, intelligent suggestions, and local data processing. For privacy, these new chips offer greater computing power, broader device compatibility, and an improved on-device user experience, helping users work and create more efficiently. Alongside our partners, the Ryzen AI 400 series is driving the next generation of fast, smart computing that delivers phenomenal AI benefits for everyday tasks.  

Pricing and Availability 

Systems with AMD Ryzen AI 400 series and Ryzen AI Pro 400 series will be available from top brands like Acer, Asus, Dell, HP, Gigabyte, and Lenovo starting in Q1 2026. Desktops with the Ryzen AI 400 series will launch later in Q2 2026.

Source: AMD Expands AI Leadership Across Client, Graphics, and Software with New Ryzen, Ryzen AI, and AMD ROCm Announcements at CES 2026 

On March 2 at Mobile World Congress Barcelona 2026, Intel and Ericsson announced an expanded partnership to accelerate the telecom industry’s adoption of AI-powered sixth-generation wireless networks.  

The partnership extends their decades-long relationship and focuses on developing compute infrastructure, connectivity systems, and cloud platforms for commercial 6G deployments.  

Ericsson, valued at about $38 billion, ranks among the top three telecom equipment vendors worldwide alongside Nokia and Huawei.  

The company primarily serves wireless carriers but is expanding into enterprise markets with 5G and cloud-based communication platforms.  

The partnership targets technical areas like:  

• Radio Access Networks  

• Packet Core Systems  

• Edge Computing  

• Security Systems  

They will collaborate on standards and ecosystem readiness to enable operators to launch AI-native networks rather than just add AI to existing systems. At the event, both showcased joint technology projects in their booths.  

Moving From Research To Practical World Use 

Borje Ekholm, Ericsson’s President and CEO, called the initiative a move from minor network upgrades to full infrastructure redesign.  

6G is not simply an iteration of mobile technology. Ekholm said it is the infrastructure that will distribute AI across devices, the edge, and the cloud. Ericsson’s long history of network innovation and large-scale operator deployments enables us to lead practical integration across the value chain and move 6G from research into commercial reality.  

Global standards groups are still developing sixth-generation wireless standards, and the industry expects commercial rollouts in the early 2030s. 6G is designed to offer speedier data speeds, lower latency, and support for more devices than 5G.  

Industry groups say that 6G’s main feature is to embed machine learning and AI into the network, from the start, rather than adding them later to systems not designed for them.  

Lip-Bu Tan, Intel’s CEO, described the company’s technical priorities within the partnership. Intel’s ambition is to be the undisputed technology leader in combining RAN, core, and edge AI to enable an effortless transition to AI-native 6G environments. Tan said: “Together with Ericsson, we will continue to demonstrate that the future of network connectivity is open, power-efficient, secure, and grounded in intelligent AI inference.”  

Open Connectivity refers to network designs with standardized interfaces that enable operators to use multi-vendor equipment.  

This method, called Open RAN, helps operators avoid reliance on a single supplier, lowers overall deployment costs, and gives smaller equipment makers a chance to compete with larger companies.  

The partnership aims to accelerate and optimize AI workloads across the network by using AI-optimized processors at cell sites and edge locations and by building systems for both network processing and machine learning on unified hardware.  

Intel’s Xeon processors will be used in cloud RAN systems, while future Ericsson chips will use Intel’s manufacturing technology.  

AI For Networks And Networks For AI 

The companies are forging ahead to achieve two ambitious goals: harnessing AI to revolutionize network operations by managing radio resources for casting traffic and seamlessly automating infrastructure management.  

The second goal involves building networks specifically designed to support AI applications running on devices and in cloud environments. In this vision, compute and connectivity are treated as integrated rather than separate layers, directly linking to the evolving operational needs enabled by AI.  

This integrated design contrasts with today’s mobile networks, where most computing happens in data centers and on users’ devices. Looking ahead, processing will be spread out across the network using radio equipment and edge servers for distributed computing, reflecting the demands of emerging AI-powered applications.  

Further, supporting this evolution, the plan also includes adding sensing abilities to the network. With these enhancements, the infrastructure can notice and react to what’s happening around it, not just move data from one place to another.  

Ericsson and Intel’s long-standing collaboration on cloud-based radio access and 5G core systems sets the stage for their latest venture. A 6G partnership focused on maximizing energy efficiency, championing open interfaces, and embedding cutting-edge AI.  

Industry groups like 3GPP are gearing up to launch official work on 6G standards, with the world eagerly awaiting the commercial debut of 6G technology projected for 2030-2032.

Source: Intel and Ericsson Partner Up to Accelerate AI-Native 6G Networks