Telecom companies last year spent nearly $295 billion in capital expenditures and over $1 trillion in operating expenditures. These large expenses are due in part to laborious manual processes that telcos face when operating networks that require continuous optimizations. For example, telcos must constantly tune network parameters for tasks — such as transferring calls from one network to another or distributing network traffic across multiple servers — based on the time of day, user behavior, mobility and traffic type. These factors directly affect network performance, user experience and energy consumption. To automate these optimization processes and save costs for telcos across the globe, NVIDIA today unveiled at GTC Paris its first AI Blueprint for telco network configuration. At the blueprint’s core are customized large language models trained specifically on telco network data — as well as the full technical and operational architecture for turning the LLMs into an autonomous, goal-driven AI agent for telcos. Automate Network Configuration With the AI Blueprint NVIDIA AI Blueprints — available on build.nvidia.com — are customizable AI workflow examples. They include reference code, documentation and deployment tools that show enterprise developers how to deliver business value with NVIDIA NIM microservices. The AI Blueprint for telco network configuration — built with BubbleRAN 5G solutions and datasets — enables developers, network engineers and telecom providers to automatically optimize the configuration of network parameters using agentic AI. This can streamline operations, reduce costs and significantly improve service quality by embedding continuous learning and adaptability directly into network infrastructures. Traditionally, network configurations required manual intervention or followed rigid rules to adapt to dynamic network conditions. These approaches limited adaptability and increased operational complexities, costs and inefficiencies. The new blueprint helps shift telco operations from relying on static, rules-based systems to operations based on dynamic, AI-driven automation. It enables developers to build advanced, telco-specific AI agents that make real-time, intelligent decisions and autonomously balance trade-offs — such as network speed versus interference, or energy savings versus utilization — without human input. Powered and Deployed by Industry Leaders Trained on 5G data generated by BubbleRAN, and deployed on the BubbleRAN 5G O-RAN platform, the blueprint provides telcos with insight on how to set various parameters to reach performance goals, like achieving a certain bitrate while choosing an acceptable signal-to-noise ratio — a measure that impacts voice quality and thus user experience. With the new AI Blueprint, network engineers can confidently set initial parameter values and update them as demanded by continuous network changes. Norway-based Telenor Group, which serves over 200 million customers globally, is the first telco to integrate the AI Blueprint for telco network configuration as part of its initiative to deploy intelligent, autonomous networks that meet the performance and agility demands of 5G and beyond. “The blueprint is helping us address configuration challenges and enhance quality of service during network installation,” said Knut Fjellheim, chief technology innovation officer at Telenor Maritime. “Implementing it is part of our push toward network automation and follows the successful deployment of agentic AI for real-time network slicing in a private 5G maritime use case.” Industry Partners Deploy Other NVIDIA-Powered Autonomous Network Technologies The AI Blueprint for telco network configuration is just one of many announcements at NVIDIA GTC Paris showcasing how the telecom industry is using agentic AI to make autonomous networks a reality. Beyond the blueprint, leading telecom companies and solutions providers are tapping into NVIDIA accelerated computing, software and microservices to provide breakthrough innovations poised to vastly improve networks and communications services — accelerating the progress to autonomous networks and improving customer experiences. NTT DATA is powering its agentic platform for telcos with NVIDIA accelerated compute and the NVIDIA AI Enterprise software platform. Its first agentic use case is focused on network alarms management, where NVIDIA NIM microservices help automate and power observability, troubleshooting, anomaly detection and resolution with closed loop ticketing. Tata Consultancy Services is delivering agentic AI solutions for telcos built on NVIDIA DGX Cloud and using NVIDIA AI Enterprise to develop, fine-tune and integrate large telco models into AI agent workflows. These range from billing and revenue assurance, autonomous network management to hybrid edge-cloud distributed inference. For example, the company’s anomaly management agentic AI model includes real-time detection and resolution of network anomalies and service performance optimization. This increases business agility and improves operational efficiencies by up to 40% by eliminating human intensive toils, overheads and cross-departmental silos. Prodapt has introduced an autonomous operations workflow for networks, powered by NVIDIA AI Enterprise, that offers agentic AI capabilities to support autonomous telecom networks. AI agents can autonomously monitor networks, detect anomalies in real time, initiate diagnostics, analyze root causes of issues using historical data and correlation techniques, automatically execute corrective actions, and generate, enrich and assign incident tickets through integrated ticketing systems. Accenture announced its new portfolio of agentic AI solutions for telecommunications through its AI Refinery platform, built on NVIDIA AI Enterprise software and accelerated computing. The first available solution, the NOC Agentic App, boosts network operations center tasks by using a generative AI-driven, nonlinear agentic framework to automate processes such as incident and fault management, root cause analysis and configuration planning. Using the Llama 3.1 70B NVIDIA NIM microservice and the AI Refinery Distiller Framework, the NOC Agentic App orchestrates networks of intelligent agents for faster, more efficient decision-making. Infosys is announcing its agentic autonomous operations platform, called Infosys Smart Network Assurance (ISNA), designed to accelerate telecom operators’ journeys toward fully autonomous network operations. ISNA helps address long-standing operational challenges for telcos — such as limited automation and high average time to repair — with an integrated, AI-driven platform that reduces operational costs by up to 40% and shortens fault resolution times by up to 30%. NVIDIA NIM and NeMo microservices enhance the platform’s reasoning and hallucination-detection capabilities, reduce latency and increase accuracy. Get started with the new blueprint today. Learn more about the latest AI advancements for telecom and other industries at NVIDIA GTC Paris, running through Thursday, June 12, at VivaTech, including a keynote from NVIDIA founder and CEO Jensen Huang and a special address from Ronnie Vasishta, senior vice president of telecom at NVIDIA. Plus, hear from industry leaders in a panel session with Orange, Swisscom, Telenor and NVIDIA.

Stellar Blade is a mashup of classic arcade action and modern Soulslike about a defense force officer with big tits and an even bigger ass saving the post-apocalyptic ruins of Earth from the aliens who destroyed it. Paradise lost? Not for the gooners

Malika Favre and George Wu bring the fun back to our feeds.

AI, Meta, Llama, open source, cybersecurity, protection tools, information security, threat protection, artificial intelligence ## Introduction In a world where artificial intelligence is becoming increasingly prevalent, the importance of cybersecurity cannot be overstated. Meta, the parent company of Facebook, is making strides in this domain. Recently, the company announced the release of a...

Navigating the dense urban canyons of cities like San Francisco or New York can be a nightmare for GPS systems. The towering skyscrapers block and reflect satellite signals, leading to location errors of tens of meters. For you and me, that might mean a missed turn. But for an autonomous vehicle or a delivery robot, that level of imprecision is the difference between a successful mission and a costly failure. These machines require pinpoint accuracy to operate safely and efficiently. Addressing this critical challenge, researchers from the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland have introduced a groundbreaking new method for visual localization during CVPR 2025 Their new paper, “FG2: Fine-Grained Cross-View Localization by Fine-Grained Feature Matching,” presents a novel AI model that significantly enhances the ability of a ground-level system, like an autonomous car, to determine its exact position and orientation using only a camera and a corresponding aerial (or satellite) image. The new approach has demonstrated a remarkable 28% reduction in mean localization error compared to the previous state-of-the-art on a challenging public dataset. Key Takeaways: Superior Accuracy: The FG2 model reduces the average localization error by a significant 28% on the VIGOR cross-area test set, a challenging benchmark for this task. Human-like Intuition: Instead of relying on abstract descriptors, the model mimics human reasoning by matching fine-grained, semantically consistent features—like curbs, crosswalks, and buildings—between a ground-level photo and an aerial map. Enhanced Interpretability: The method allows researchers to “see” what the AI is “thinking” by visualizing exactly which features in the ground and aerial images are being matched, a major step forward from previous “black box” models. Weakly Supervised Learning: Remarkably, the model learns these complex and consistent feature matches without any direct labels for correspondences. It achieves this using only the final camera pose as a supervisory signal. Challenge: Seeing the World from Two Different Angles The core problem of cross-view localization is the dramatic difference in perspective between a street-level camera and an overhead satellite view. A building facade seen from the ground looks completely different from its rooftop signature in an aerial image. Existing methods have struggled with this. Some create a general “descriptor” for the entire scene, but this is an abstract approach that doesn’t mirror how humans naturally localize themselves by spotting specific landmarks. Other methods transform the ground image into a Bird’s-Eye-View (BEV) but are often limited to the ground plane, ignoring crucial vertical structures like buildings. FG2: Matching Fine-Grained Features The EPFL team’s FG2 method introduces a more intuitive and effective process. It aligns two sets of points: one generated from the ground-level image and another sampled from the aerial map. Here’s a breakdown of their innovative pipeline: Mapping to 3D: The process begins by taking the features from the ground-level image and lifting them into a 3D point cloud centered around the camera. This creates a 3D representation of the immediate environment. Smart Pooling to BEV: This is where the magic happens. Instead of simply flattening the 3D data, the model learns to intelligently select the most important features along the vertical (height) dimension for each point. It essentially asks, “For this spot on the map, is the ground-level road marking more important, or is the edge of that building’s roof the better landmark?” This selection process is crucial, as it allows the model to correctly associate features like building facades with their corresponding rooftops in the aerial view. Feature Matching and Pose Estimation: Once both the ground and aerial views are represented as 2D point planes with rich feature descriptors, the model computes the similarity between them. It then samples a sparse set of the most confident matches and uses a classic geometric algorithm called Procrustes alignment to calculate the precise 3-DoF (x, y, and yaw) pose. Unprecedented Performance and Interpretability The results speak for themselves. On the challenging VIGOR dataset, which includes images from different cities in its cross-area test, FG2 reduced the mean localization error by 28% compared to the previous best method. It also demonstrated superior generalization capabilities on the KITTI dataset, a staple in autonomous driving research. Perhaps more importantly, the FG2 model offers a new level of transparency. By visualizing the matched points, the researchers showed that the model learns semantically consistent correspondences without being explicitly told to. For example, the system correctly matches zebra crossings, road markings, and even building facades in the ground view to their corresponding locations on the aerial map. This interpretability is extremenly valuable for building trust in safety-critical autonomous systems. “A Clearer Path” for Autonomous Navigation The FG2 method represents a significant leap forward in fine-grained visual localization. By developing a model that intelligently selects and matches features in a way that mirrors human intuition, the EPFL researchers have not only shattered previous accuracy records but also made the decision-making process of the AI more interpretable. This work paves the way for more robust and reliable navigation systems for autonomous vehicles, drones, and robots, bringing us one step closer to a future where machines can confidently navigate our world, even when GPS fails them. Check out the Paper. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Jean-marc MommessinJean-marc is a successful AI business executive .He leads and accelerates growth for AI powered solutions and started a computer vision company in 2006. He is a recognized speaker at AI conferences and has an MBA from Stanford.Jean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/AI-Generated Ad Created with Google’s Veo3 Airs During NBA Finals, Slashing Production Costs by 95%Jean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Highlighted at CVPR 2025: Google DeepMind’s ‘Motion Prompting’ Paper Unlocks Granular Video ControlJean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Snowflake Charts New AI Territory: Cortex AISQL & Snowflake Intelligence Poised to Reshape Data AnalyticsJean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Exclusive Talk: Joey Conway of NVIDIA on Llama Nemotron Ultra and Open Source Models

This article was originally published on GameDevDigest.comEnjoy!What was Radiant AI, anyway? - A ridiculously deep dive into Oblivion's controversial AI system and its legacyblog.paavo.meConsider The Horse Game - No I don’t think every dev should make a horse game (unlike horror, which I still think everyone should at least one). But I do think every developer should at least look at them, maybe even play one because, it is very important that you understand the importance of genre, fandom, and how visibility works. Even if you are not making a horse game, the lessons you can learn by looking at this sub genre are very similar to other genres, just not as blatantly clear as they are with horse games.howtomarketagame.comMaking a killing: The playful 2D terror of Psycasso® - I sat down with lead developer Benjamin Lavender and Omni, designer and producer, to talk about this playfully gory game that gives a classic retro style and a fresh (if gruesome) twist.UnityIntroduction to Asset Manager transfer methods in Unity - Unity's Asset Manager is a user-friendly digital asset management platform supporting over 70 file formats to help teams centralize, organize, discover, and use assets seamlessly across projects. It reduces redundant work by design, making cross-team collaboration smoother and accelerating production workflows.UnityVideosRules of the Game: Five Tricks of Highly Effective Designers - Every working designer has them: unique techniques or "tricks" that they use when crafting gameplay. Sure, there's the general game design wisdom that everyone agrees on and can be found in many a game design book, but experienced game designers often have very specific rules that are personal to them, techniques that not everyone knows about or even agrees with. In this GDC 2015 session, five experienced game designers join the stage for 10 minutes each to share one game design "trick" that they use.Game Developers ConferenceBinding of Isaac Style Room Generator in Unity [Full Tutorial] - Our third part in the series - making the rooms!Game Dev GarnetIntroduction to Unity Behavior | Unity Tutorial - In this video you'll become familiar with the core concepts of Unity Behavior, including a live example.LlamAcademyHow I got my demo ready for Steam Next Fest - It's Steam Next Fest, and I've got a game in the showcase. So here are 7 tips for making the most of this demo sharing festival.Game Maker's ToolkitOptimizing lighting in Projekt Z: Beyond Order - 314 Arts studio lead and founder Justin Miersch discuss how the team used the Screen Space Global Illumination feature in Unity’s High Definition Render Pipeline (HDRP), along with the Unity Profiler and Timeline to overcome the lighting challenges they faced in building Projekt Z: Beyond Order.UnityMemory Arenas in Unity: Heap Allocation Without the GC - In this video, we explore how to build a custom memory arena in Unity using unsafe code and manual heap allocation. You’ll learn how to allocate raw memory for temporary graph-like structures—such as crafting trees or decision planners—without triggering the garbage collector. We’ll walk through the concept of stack frames, translate that to heap-based arena allocation, and implement a fast, disposable system that gives you full control over memory layout and lifetime. Perfect for performance-critical systems where GC spikes aren’t acceptable.git-amendCloth Animation Using The Compute Shader - In this video, we dive into cloth simulation using OpenGL compute shaders. By applying simple mathematical equations, we’ll achieve smooth, dynamic movement. We'll explore particle-based simulation, tackle synchronization challenges with double buffering, and optimize rendering using triangle strips for efficient memory usage. Whether you're familiar with compute shaders or just getting started, this is the perfect way to step up your real-time graphics skills!OGLDEVHow we're designing games for a broader audience - Our games are too hardBiteMe GamesAssetsLearn Game Dev - Unity, Godot, Unreal, Gamemaker, Blender & C# - Make games like a pro.Passionate about video games? Then start making your own! Our latest bundle will help you learn vital game development skills. Master the most popular creation platforms like Unity, Godot, Unreal, GameMaker, Blender, and C#—now that’s a sharp-lookin’ bundle! Build a 2.5D farming RPG with Unreal Engine, create a micro turn-based RPG in Godot, explore game optimization, and so much more.__Big Bang Unreal & Unity Asset Packs Bundle - 5000+ unrivaled assets in one bundle. Calling all game devs—build your worlds with this gigantic bundle of over 5000 assets, including realistic and stylized environments, SFX packs, and powerful tools. Perfect for hobbyists, beginners, and professional developers alike, you'll gain access to essential resources, tutorials, and beta-testing–ready content to start building immediately. The experts at Leartes Studios have curated an amazing library packed with value, featuring environments, VFX packs, and tutorial courses on Unreal Engine, Blender, Substance Painter, and ZBrush. Get the assets you need to bring your game to life—and help support One Tree Planted with your purchase! This bundle provides Unity Asset Store keys directly with your purchase, and FAB keys via redemption through Cosmos, if the product is available on those platforms.Humble Bundle AffiliateGameplay Tools 50% Off - Core systems, half the price. Get pro-grade tools to power your gameplay—combat, cutscenes, UI, and more. Including: HTrace: World Space Global Illumination, VFX Graph - Ultra Mega Pack - Vol.1, Magic Animation Blend, Utility Intelligence (v2): Utility AI Framework for Unity 6, Build for iOS/macOS on Windows>?Unity AffiliateHi guys, I created a website about 6 years in which I host all my field recordings and foley sounds. All free to download and use CC0. There is currently 50+ packs with 1000's of sounds and hours of field recordings all perfect for game SFX and UI. - I think game designers can benefit from a wide range of sounds on the site, especially those that enhance immersion and atmosphere.signaturesounds.orgSmartAddresser - Automate Addressing, Labeling, and Version Control for Unity's Addressable Asset System.CyberAgentGameEntertainment Open SourceEasyCS - EasyCS is an easy-to-use and flexible framework for Unity, adopting a Data-Driven Entity & Actor-Component approach. It bridges Unity's classic OOP with powerful data-oriented patterns, without forcing a complete ECS paradigm shift or a mindset change. Build smarter, not harder.Watcher3056 Open SourceBinding-Of-Isaac_Map-Generator - Binding of Isaac map generator for Unity2DGarnetKane99 Open SourceHelion - A modern fast paced Doom FPS engineHelion-Engine Open SourcePixelationFx - Pixelation post effect for Unity UrpNullTale Open SourceExtreme Add-Ons Bundle For Blender & ZBrush - Extraordinary quality—Extreme add-ons Get quality add-ons for Blender and ZBrush with our latest bundle! We’ve teamed up with the pros at FlippedNormals to deliver a gigantic library of powerful tools for your next game development project. Add new life to your creative work with standout assets like Real-time Hair ZBrush Plugin, Physical Starlight and Atmosphere, Easy Mesh ZBrush Plugin, and more. Get the add-ons you need to bring color and individuality to your next project—and help support Extra Life with your purchase!Humble Bundle AffiliateShop up to 50% off Gabriel Aguiar Prod - Publisher Sale - Gabriel Aguiar Prod. is best known for his extensive VFX assets that help many developers prototype and ship games with special effects. His support and educational material are also invaluable resources for the game dev community. PLUS get VFX Graph - Stylized Fire - Vol. 1 for FREE with code GAP2025Unity AffiliateSpotlightDream Garden - Dream Garden is a simulation game about building tiny cute garden dioramas. A large selection of tools, plants, decorations and customization awaits you. Try all of them and create your dream garden.[You can find it on Steam]Campfire StudioMy game, Call Of Dookie. Demo available on SteamYou can subscribe to the free weekly newsletter on GameDevDigest.comThis post includes affiliate links; I may receive compensation if you purchase products or services from the different links provided in this article.

AI researchers have recently been asking themselves a version of the question, “Is that really Zuck?”As first reported by Bloomberg, the Meta CEO has been personally asking top AI talent to join his new “superintelligence” AI lab and reboot Llama. His recruiting process typically goes like this: a cold outreach via email or WhatsApp that cites the recruit’s work history and requests a 15-minute chat. Dozens of researchers have gotten these kinds of messages at Google alone. For those who do agree to hear his pitch (amazingly, not all of them do), Zuckerberg highlights the latitude they’ll have to make risky bets, the scale of Meta’s products, and the money he’s prepared to invest in the infrastructure to support them. He makes clear that this new team will be empowered and sit with him at Meta’s headquarters, where I’m told the desks have already been rearranged for the incoming team.Most of the headlines so far have focused on the eye-popping compensation packages Zuckerberg is offering, some of which are well into the eight-figure range. As I’ve covered before, hiring the best AI researcher is like hiring a star basketball player: there are very few of them, and you have to pay up. Case in point: Zuckerberg basically just paid 14 Instagrams to hire away Scale AI CEO Alexandr Wang. It’s easily the most expensive hire of all time, dwarfing the billions that Google spent to rehire Noam Shazeer and his core team from Character.AI (a deal Zuckerberg passed on). “Opportunities of this magnitude often come at a cost,” Wang wrote in his note to employees this week. “In this instance, that cost is my departure.”Zuckerberg’s recruiting spree is already starting to rattle his competitors. The day before his offer deadline for some senior OpenAI employees, Sam Altman dropped an essay proclaiming that “before anything else, we are a superintelligence research company.” And after Zuckerberg tried to hire DeepMind CTO Koray Kavukcuoglu, he was given a larger SVP title and now reports directly to Google CEO Sundar Pichai. I expect Wang to have the title of “chief AI officer” at Meta when the new lab is announced. Jack Rae, a principal researcher from DeepMind who has signed on, will lead pre-training. Meta certainly needs a reset. According to my sources, Llama has fallen so far behind that Meta’s product teams have recently discussed using AI models from other companies (although that is highly unlikely to happen). Meta’s internal coding tool for engineers, however, is already using Claude. While Meta’s existing AI researchers have good reason to be looking over their shoulders, Zuckerberg’s $14.3 billion investment in Scale is making many longtime employees, or Scaliens, quite wealthy. They were popping champagne in the office this morning. Then, Wang held his last all-hands meeting to say goodbye and cried. He didn’t mention what he would be doing at Meta. I expect his new team will be unveiled within the next few weeks after Zuckerberg gets a critical number of members to officially sign on. Tim Cook. Getty Images / The VergeApple’s AI problemApple is accustomed to being on top of the tech industry, and for good reason: the company has enjoyed a nearly unrivaled run of dominance. After spending time at Apple HQ this week for WWDC, I’m not sure that its leaders appreciate the meteorite that is heading their way. The hubris they display suggests they don’t understand how AI is fundamentally changing how people use and build software.Heading into the keynote on Monday, everyone knew not to expect the revamped Siri that had been promised the previous year. Apple, to its credit, acknowledged that it dropped the ball there, and it sounds like a large language model rebuild of Siri is very much underway and coming in 2026.The AI industry moves much faster than Apple’s release schedule, though. By the time Siri is perhaps good enough to keep pace, it will have to contend with the lock-in that OpenAI and others are building through their memory features. Apple and OpenAI are currently partners, but both companies want to ultimately control the interface for interacting with AI, which puts them on a collision course. Apple’s decision to let developers use its own, on-device foundational models for free in their apps sounds strategically smart, but unfortunately, the models look far from leading. Apple ran its own benchmarks, which aren’t impressive, and has confirmed a measly context window of 4,096 tokens. It’s also saying that the models will be updated alongside its operating systems — a snail’s pace compared to how quickly AI companies move. I’d be surprised if any serious developers use these Apple models, although I can see them being helpful to indie devs who are just getting started and don’t want to spend on the leading cloud models. I don’t think most people care about the privacy angle that Apple is claiming as a differentiator; they are already sharing their darkest secrets with ChatGPT and other assistants. Some of the new Apple Intelligence features I demoed this week were impressive, such as live language translation for calls. Mostly, I came away with the impression that the company is heavily leaning on its ChatGPT partnership as a stopgap until Apple Intelligence and Siri are both where they need to be. AI probably isn’t a near-term risk to Apple’s business. No one has shipped anything close to the contextually aware Siri that was demoed at last year’s WWDC. People will continue to buy Apple hardware for a long time, even after Sam Altman and Jony Ive announce their first AI device for ChatGPT next year. AR glasses aren’t going mainstream anytime soon either, although we can expect to see more eyewear from Meta, Google, and Snap over the coming year. In aggregate, these AI-powered devices could begin to siphon away engagement from the iPhone, but I don’t see people fully replacing their smartphones for a long time. The bigger question after this week is whether Apple has what it takes to rise to the occasion and culturally reset itself for the AI era. I would have loved to hear Tim Cook address this issue directly, but the only interview he did for WWDC was a cover story in Variety about the company’s new F1 movie.ElsewhereAI agents are coming. I recently caught up with Databricks CEO Ali Ghodsi ahead of his company’s annual developer conference this week in San Francisco. Given Databricks’ position, he has a unique, bird’s-eye view of where things are headed for AI. He doesn’t envision a near-term future where AI agents completely automate real-world tasks, but he does predict a wave of startups over the next year that will come close to completing actions in areas such as travel booking. He thinks humans will need (and want) to approve what an agent does before it goes off and completes a task. “We have most of the airplanes flying automated, and we still want pilots in there.”Buyouts are the new normal at Google. That much is clear after this week’s rollout of the “voluntary exit program” in core engineering, the Search organization, and some other divisions. In his internal memo, Search SVP Nick Fox was clear that management thinks buyouts have been successful in other parts of the company that have tried them. In a separate memo I saw, engineering exec Jen Fitzpatrick called the buyouts an “opportunity to create internal mobility and fresh growth opportunities.” Google appears to be attempting a cultural reset, which will be a challenging task for a company of its size. We’ll see if it can pull it off. Evan Spiegel wants help with AR glasses. I doubt that his announcement that consumer glasses are coming next year was solely aimed at AR developers. Telegraphing the plan and announcing that Snap has spent $3 billion on hardware to date feels more aimed at potential partners that want to make a bigger glasses play, such as Google. A strategic investment could help insulate Snap from the pain of the stock market. A full acquisition may not be off the table, either. When he was recently asked if he’d be open to a sale, Spiegel didn’t shut it down like he always has, but instead said he’d “consider anything” that helps the company “create the next computing platform.”Link listMore to click on:If you haven’t already, don’t forget to subscribe to The Verge, which includes unlimited access to Command Line and all of our reporting.As always, I welcome your feedback, especially if you’re an AI researcher fielding a juicy job offer. You can respond here or ping me securely on Signal.Thanks for subscribing.See More:

Join the event trusted by enterprise leaders for nearly two decades. VB Transform brings together the people building real enterprise AI strategy. Learn more When DeepSeek released its R1 model this January, it wasn’t just another AI announcement. It was a watershed moment that sent shockwaves through the tech industry, forcing industry leaders to reconsider their fundamental approaches to AI development. What makes DeepSeek’s accomplishment remarkable isn’t that the company developed novel capabilities; rather, it was how it achieved comparable results to those delivered by tech heavyweights at a fraction of the cost. In reality, DeepSeek didn’t do anything that hadn’t been done before; its innovation stemmed from pursuing different priorities. As a result, we are now experiencing rapid-fire development along two parallel tracks: efficiency and compute.  As DeepSeek prepares to release its R2 model, and as it concurrently faces the potential of even greater chip restrictions from the U.S., it’s important to look at how it captured so much attention. Engineering around constraints DeepSeek’s arrival, as sudden and dramatic as it was, captivated us all because it showcased the capacity for innovation to thrive even under significant constraints. Faced with U.S. export controls limiting access to cutting-edge AI chips, DeepSeek was forced to find alternative pathways to AI advancement. While U.S. companies pursued performance gains through more powerful hardware, bigger models and better data, DeepSeek focused on optimizing what was available. It implemented known ideas with remarkable execution — and there is novelty in executing what’s known and doing it well. This efficiency-first mindset yielded incredibly impressive results. DeepSeek’s R1 model reportedly matches OpenAI’s capabilities at just 5 to 10% of the operating cost. According to reports, the final training run for DeepSeek’s V3 predecessor cost a mere $6 million — which was described by former Tesla AI scientist Andrej Karpathy as “a joke of a budget” compared to the tens or hundreds of millions spent by U.S. competitors. More strikingly, while OpenAI reportedly spent $500 million training its recent “Orion” model, DeepSeek achieved superior benchmark results for just $5.6 million — less than 1.2% of OpenAI’s investment. If you get starry eyed believing these incredible results were achieved even as DeepSeek was at a severe disadvantage based on its inability to access advanced AI chips, I hate to tell you, but that narrative isn’t entirely accurate (even though it makes a good story). Initial U.S. export controls focused primarily on compute capabilities, not on memory and networking — two crucial components for AI development. That means that the chips DeepSeek had access to were not poor quality chips; their networking and memory capabilities allowed DeepSeek to parallelize operations across many units, a key strategy for running their large model efficiently. This, combined with China’s national push toward controlling the entire vertical stack of AI infrastructure, resulted in accelerated innovation that many Western observers didn’t anticipate. DeepSeek’s advancements were an inevitable part of AI development, but they brought known advancements forward a few years earlier than would have been possible otherwise, and that’s pretty amazing. Pragmatism over process Beyond hardware optimization, DeepSeek’s approach to training data represents another departure from conventional Western practices. Rather than relying solely on web-scraped content, DeepSeek reportedly leveraged significant amounts of synthetic data and outputs from other proprietary models. This is a classic example of model distillation, or the ability to learn from really powerful models. Such an approach, however, raises questions about data privacy and governance that might concern Western enterprise customers. Still, it underscores DeepSeek’s overall pragmatic focus on results over process. The effective use of synthetic data is a key differentiator. Synthetic data can be very effective when it comes to training large models, but you have to be careful; some model architectures handle synthetic data better than others. For instance, transformer-based models with mixture of experts (MoE) architectures like DeepSeek’s tend to be more robust when incorporating synthetic data, while more traditional dense architectures like those used in early Llama models can experience performance degradation or even “model collapse” when trained on too much synthetic content. This architectural sensitivity matters because synthetic data introduces different patterns and distributions compared to real-world data. When a model architecture doesn’t handle synthetic data well, it may learn shortcuts or biases present in the synthetic data generation process rather than generalizable knowledge. This can lead to reduced performance on real-world tasks, increased hallucinations or brittleness when facing novel situations.  Still, DeepSeek’s engineering teams reportedly designed their model architecture specifically with synthetic data integration in mind from the earliest planning stages. This allowed the company to leverage the cost benefits of synthetic data without sacrificing performance. Market reverberations Why does all of this matter? Stock market aside, DeepSeek’s emergence has triggered substantive strategic shifts among industry leaders. Case in point: OpenAI. Sam Altman recently announced plans to release the company’s first “open-weight” language model since 2019. This is a pretty notable pivot for a company that built its business on proprietary systems. It seems DeepSeek’s rise, on top of Llama’s success, has hit OpenAI’s leader hard. Just a month after DeepSeek arrived on the scene, Altman admitted that OpenAI had been “on the wrong side of history” regarding open-source AI.  With OpenAI reportedly spending $7 to 8 billion annually on operations, the economic pressure from efficient alternatives like DeepSeek has become impossible to ignore. As AI scholar Kai-Fu Lee bluntly put it: “You’re spending $7 billion or $8 billion a year, making a massive loss, and here you have a competitor coming in with an open-source model that’s for free.” This necessitates change. This economic reality prompted OpenAI to pursue a massive $40 billion funding round that valued the company at an unprecedented $300 billion. But even with a war chest of funds at its disposal, the fundamental challenge remains: OpenAI’s approach is dramatically more resource-intensive than DeepSeek’s. Beyond model training Another significant trend accelerated by DeepSeek is the shift toward “test-time compute” (TTC). As major AI labs have now trained their models on much of the available public data on the internet, data scarcity is slowing further improvements in pre-training. To get around this, DeepSeek announced a collaboration with Tsinghua University to enable “self-principled critique tuning” (SPCT). This approach trains AI to develop its own rules for judging content and then uses those rules to provide detailed critiques. The system includes a built-in “judge” that evaluates the AI’s answers in real-time, comparing responses against core rules and quality standards. The development is part of a movement towards autonomous self-evaluation and improvement in AI systems in which models use inference time to improve results, rather than simply making models larger during training. DeepSeek calls its system “DeepSeek-GRM” (generalist reward modeling). But, as with its model distillation approach, this could be considered a mix of promise and risk. For example, if the AI develops its own judging criteria, there’s a risk those principles diverge from human values, ethics or context. The rules could end up being overly rigid or biased, optimizing for style over substance, and/or reinforce incorrect assumptions or hallucinations. Additionally, without a human in the loop, issues could arise if the “judge” is flawed or misaligned. It’s a kind of AI talking to itself, without robust external grounding. On top of this, users and developers may not understand why the AI reached a certain conclusion — which feeds into a bigger concern: Should an AI be allowed to decide what is “good” or “correct” based solely on its own logic? These risks shouldn’t be discounted. At the same time, this approach is gaining traction, as again DeepSeek builds on the body of work of others (think OpenAI’s “critique and revise” methods, Anthropic’s constitutional AI or research on self-rewarding agents) to create what is likely the first full-stack application of SPCT in a commercial effort. This could mark a powerful shift in AI autonomy, but there still is a need for rigorous auditing, transparency and safeguards. It’s not just about models getting smarter, but that they remain aligned, interpretable, and trustworthy as they begin critiquing themselves without human guardrails. Moving into the future So, taking all of this into account, the rise of DeepSeek signals a broader shift in the AI industry toward parallel innovation tracks. While companies continue building more powerful compute clusters for next-generation capabilities, there will also be intense focus on finding efficiency gains through software engineering and model architecture improvements to offset the challenges of AI energy consumption, which far outpaces power generation capacity.  Companies are taking note. Microsoft, for example, has halted data center development in multiple regions globally, recalibrating toward a more distributed, efficient infrastructure approach. While still planning to invest approximately $80 billion in AI infrastructure this fiscal year, the company is reallocating resources in response to the efficiency gains DeepSeek introduced to the market. Meta has also responded, With so much movement in such a short time, it becomes somewhat ironic that the U.S. sanctions designed to maintain American AI dominance may have instead accelerated the very innovation they sought to contain. By constraining access to materials, DeepSeek was forced to blaze a new trail. Moving forward, as the industry continues to evolve globally, adaptability for all players will be key. Policies, people and market reactions will continue to shift the ground rules — whether it’s eliminating the AI diffusion rule, a new ban on technology purchases or something else entirely. It’s what we learn from one another and how we respond that will be worth watching. Jae Lee is CEO and co-founder of TwelveLabs. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Meta is looking to up its weakening AI game with a key talent grab. Following days of speculation, the social media giant has confirmed that Scale AI’s founder and CEO, Alexandr Wang, is joining Meta to work on its AI efforts. Meta will invest $14.3 billion in Scale AI as part of the deal, and will have a 49% stake in the AI startup, which specializes in data labeling and model evaluation services. Other key Scale employees will also move over to Meta, while CSO Jason Droege will step in as Scale’s interim CEO. This move comes as the Mark Zuckerberg-led company goes all-in on building a new research lab focused on “superintelligence,” the next step beyond artificial general intelligence (AGI). The arrangement also reflects a growing trend in big tech, where industry giants are buying companies without really buying them — what’s increasingly being referred to as “acqui-hiring.” It involves recruiting key personnel from a company, licensing its technology, and selling its products, but leaving it as a private entity. “This is fundamentally a massive ‘acqui-hire’ play disguised as a strategic investment,” said Wyatt Mayham, lead AI consultant at Northwest AI Consulting. “While Meta gets Scale’s data infrastructure, the real prize is Wang joining Meta to lead their superintelligence lab. At the $14.3 billion price tag, this might be the most expensive individual talent acquisition in tech history.” Closing gaps with competitors Meta has struggled to keep up with OpenAI, Anthropic, and other key competitors in the AI race, recently even delaying the launch of its new flagship model, Behemoth, purportedly due to internal concerns about its performance. It has also seen the departure of several of its top researchers.  “It’s not really a secret at this point that Meta’s Llama 4 models have had significant performance issues,” Mayham said. “Zuck is essentially betting that Wang’s track record building AI infrastructure can solve Meta’s alignment and model quality problems faster than internal development.” And, he added, Scale’s enterprise-grade human feedback loops are exactly what Meta’s Llama models need to compete with ChatGPT and Claude on reliability and task-following. Data quality, a key focus for Wang, is a big factor in solving those performance problems. He wrote in a note to Scale employees on Thursday, later posted on X (formerly Twitter), that when he founded Scale AI in 2016 amidst some of the early AI breakthroughs, “it was clear even then that data was the lifeblood of AI systems, and that was the inspiration behind starting Scale.” But despite Meta’s huge investment, Scale AI is underscoring its commitment to sovereignty: “Scale remains an independent leader in AI, committed to providing industry-leading AI solutions and safeguarding customer data,” the company wrote in a blog post. “Scale will continue to partner with leading AI labs, multinational enterprises, and governments to deliver expert data and technology solutions through every phase of AI’s evolution.” Allowing big tech to side-step notification But while it’s only just been inked, the high-profile deal is already raising some eyebrows. According to experts, arrangements like these allow tech companies to acquire top talent and key technologies in a side-stepping manner, thus avoiding regulatory notification requirements. The US Federal Trade Commission (FTC) requires mergers and acquisitions totaling more than $126 million be reported in advance. Licensing deals or the mass hiring-away of a company’s employees don’t have this requirement. This allows companies to move more quickly, as they don’t have to undergo the lengthy federal review process. Microsoft’s deal with Inflection AI is probably one of the highest-profile examples of the “acqui-hiring” trend. In March 2024, the tech giant paid the startup $650 million in licensing fees and hired much of its team, including co-founders Mustafa Suleyman (now CEO of Microsoft AI) and Karén Simonyan (chief scientist of Microsoft AI). Similarly, last year Amazon hired more than 50% of Adept AI’s key personnel, including its CEO, to focus on AGI. Google also inked a licensing agreement with Character AI and hired a majority of its founders and researchers. However, regulators have caught on, with the FTC launching inquiries into both the Microsoft-Inflection and Amazon-Adept deals, and the US Justice Department (DOJ) analyzing Google-Character AI. Reflecting ‘desperation’ in the AI industry Meta’s decision to go forward with this arrangement anyway, despite that dicey backdrop, seems to indicate how anxious the company is to keep up in the AI race. “The most interesting piece of this all is the timing,” said Mayham. “It reflects broader industry desperation. Tech giants are increasingly buying parts of promising AI startups to secure key talent without acquiring full companies, following similar patterns with Microsoft-Inflection and Google-Character AI.” However, the regulatory risks are “real but nuanced,” he noted. Meta’s acquisition could face scrutiny from antitrust regulators, particularly as the company is involved in an ongoing FTC lawsuit over its Instagram and WhatsApp acquisitions. While the 49% ownership position appears designed to avoid triggering automatic thresholds, US regulatory bodies like the FTC and DOJ can review minority stake acquisitions under the Clayton Antitrust Act if they seem to threaten competition. Perhaps more importantly, Meta is not considered a leader in AGI development and is trailing OpenAI, Anthropic, and Google, meaning regulators may not consider the deal all that concerning (yet). All told, the arrangement certainly signals Meta’s recognition that the AI race has shifted from a compute and model size competition to a data quality and alignment battle, Mayham noted. “I think the [gist] of this is that Zuck’s biggest bet is that talent and data infrastructure matter more than raw compute power in the AI race,” he said. “The regulatory risk is manageable given Meta’s trailing position, but the acqui-hire premium shows how expensive top AI talent has become.”