Google Beam: AI-Powered 3D Video Platform for Enterprise

Proud moment for the Industrious team today. Google and HP are bringing their new 3D video platform, Google Beam, right into our workspaces. 💼✨This means our members and guests will have access to next-gen, true-to-life meeting rooms without dealing with bulky AR/VR headsets. Distributed teams, startups, and clients can finally meet and actually feel like they're sitting at the exact same table. Who’s booking a room first? #WorkplaceInnovation #Coworking #Google #HybridWork

Really enjoyed this blog by Tomer Badug on the future of Audio AI at the edge. As AI expands beyond phones and PCs, enabling tiny devices to understand sound locally is becoming a key differentiator. Faster response times, lower power consumption, and improved privacy are just the beginning. #AudioAI #EdgeAI #IoTInnovation #Ceva #Semiconductors

As Meta's annual Connect event approaches, industry speculation points to significant hardware and AI announcements. We anticipate a focus on refreshed smart glasses, including a new camera-free model, alongside updates to its mixed reality roadmap. https://lnkd.in/en3syhMB

AR glasses are racing to become the next mobile computing platform. The hardware is real. It's shipping. And they're all arriving from nearly every major tech company within months of each other. From Snap Inc. (SPECS), Meta, Google, Samsung Electronics, Qualcomm, XREAL, RayNeo, Rokid Corporation Ltd, Even Realities, and more. This is one of the most important moments in computing since the smartphone. The devices that will connect people, physical AI, and robotics to the real world are here. What comes next is making sure what gets built for these devices is open, interoperable, and built to last. New standards are evolving web browsers and websites for spatial computing, physical AI, robotics, and digital twins. That is the work we are doing through OMBI at the Metaverse Standards Forum. The hardware is arriving. The open software layer is what makes it all work together. Learn more - link in comments. #spatialcomputing #physicalAI #robotics #digitaltwins #SPECS #SnapSpecs #metaverse

Congrats to the Snap Inc. team on SPECS! 😎 What stood out to me is Specs for Enterprise. Field service, retail, remote support. Salesforce, Amazon Web Services (AWS), and NVIDIA are building real workplace tools for a pair of glasses. That is a big step. It also raises the question enterprises always ask next: 𝐝𝐨𝐞𝐬 𝐭𝐡𝐢𝐬 𝐨𝐧𝐥𝐲 𝐰𝐨𝐫𝐤 𝐡𝐞𝐫𝐞, 𝐨𝐫 𝐜𝐚𝐧 𝐰𝐞 𝐭𝐚𝐤𝐞 𝐢𝐭 𝐰𝐢𝐭𝐡 𝐮𝐬 𝐰𝐡𝐞𝐧 𝐭𝐡𝐞 𝐧𝐞𝐱𝐭 𝐝𝐞𝐯𝐢𝐜𝐞 𝐬𝐡𝐨𝐰𝐬 𝐮𝐩? We are building new open standards that evolve web browsers and websites for spatial computing so those workflows do not have to be rebuilt for each device. That work happening now at the Metaverse Standards Forum, through the Open Metaverse Browser Initiative. If you want to see how we are making that possible, start with the links in the comments. #spatialcomputing #physicalAI #robotics #digitaltwins #SnapSpecs

The first thing a meeting agent gets wrong is the name next to each task. The summary is the easy half. Any decent model can turn an hour of talk into five bullet points. The hard half is attribution: who committed to what, and who only said it sounded like a good idea. On Sept 23, NVIDIA released Nemotron 3 Diarization, an open-weight model that does only that job. It takes 16 kHz audio and outputs, frame by frame, the probability that each of up to 8 speakers is talking. It doesn't transcribe anything. You pair it with a separate speech-to-text model. About 100M parameters. Streaming latency options go down to 0.32 seconds. NVIDIA says it ranked first on VoiceArena's diarization benchmark with a 14.72% error rate. What I take from it: → Attribution is its own component. Asking a big model to guess who spoke from a flat transcript is how action items land on the wrong person. → Streaming means an agent can tag decisions while the call is still going, instead of rebuilding them from a recording afterwards. The tradeoff is written on the model card itself. The cap is 8 speakers, and error rates climb noticeably in the 5 to 9 speaker range compared with 1 to 4. The meetings where attribution matters most, lots of people and lots of crosstalk, are exactly where it's weakest. And 14.72% is an average. One wrong name on the one task that mattered costs more than the average suggests. So before shipping a meeting agent, I'd ask how it shows the attributions it wasn't sure about. When I facilitate sessions with five or six people, the follow-up that goes sideways is almost always ownership: two people who each thought the other one had it. That's the line I'd want the agent to flag with a question mark, and leave for a human to confirm. #AIAgents #SpeechAI #EnterpriseAI

TL;DR: we launched SPECS for a human-first approach to tech. We don't want virtual worlds and living in computers. We want humans living in the "the only reality where I can get a decent steak” - Woody Allen. Also, of course we're going for enterprise as well, see https://lnkd.in/dnX3AqG9

The internet is starting to feel less like a place and more like a layer on top of real life. For most of our lives, going online meant reaching for a screen. You opened your phone, opened an app, searched for something, watched something or sent a message. The internet was somewhere you deliberately went. That relationship is starting to change. Smart glasses are bringing cameras, AI assistants, translation, navigation and other digital experiences into the physical world. Samsung introduced its intelligent eyewear this year, while Snap is also preparing its own standalone Specs with on-device AI. The interesting part isn’t really the glasses themselves. It’s what happens when technology stops asking us to look at a screen. Imagine walking into a city you’ve never visited and getting directions without pulling out your phone. Seeing a product and immediately getting information about it. Having something translated while you’re looking directly at it. Capturing a moment without reaching for a camera. The internet starts becoming less of a destination and more of an invisible layer sitting on top of everyday life. And that could change digital experiences in ways we’re only beginning to understand. Because if people don’t have to open a browser, app or social platform to interact with the digital world anymore, the question isn’t just what the next website looks like. It’s what the internet looks like when you don’t have to look at it. #DigitalMarketing #Technology #DigitalExperience #Innovation #FutureOfTheInternet

Two projects landed on my desk this month, both labeled “AI smart glasses.” But their chipset requirements had almost nothing in common. One was an audio and translation glasses project. Sitting down with the team, what actually mattered was straightforward: clear calls, stable audio, fast translation response, low power draw, slim temples, and a full-day battery target. Most of the translation and AI workload could run on the phone or in the cloud. In that architecture, a more powerful on-device chipset was not adding enough user value to justify more heat, a larger battery, thicker temples, and a higher BOM. The other was a camera-based AI smart glasses project. The moment image capture and local processing entered the product definition, the conversation changed completely. Camera pipeline, local AI workloads, storage and data transfer, thermal management, privacy, and app/SDK/OTA integration all became system-level requirements. One detail from that second project stayed with me. Early on, a lower-cost board paired with removable storage looked like an easy way to reduce the BOM. But once we worked through storage compatibility, supply quality, batch validation, and after-sales ownership, much of that “saving” had simply moved from the bill of materials into execution risk. Someone still carries that risk later. It just does not appear on the BOM line. At this point, I no longer see these as one product category with different specifications. They are different systems built to do different jobs. Before selecting a chipset, I would first ask: What must run on the glasses themselves — and what can the phone or cloud carry instead? What are you building, and what is non-negotiable to keep on-device? #AISmartGlasses #WearableODM #SmartGlasses #ProductDevelopment

Snap may have just drawn the line between “smart glasses” and actual AR glasses. On Wednesday Snap Inc. gave a much deeper look at its $2,195 SPECS - and the most interesting part wasn't another hardware specification. It was the ecosystem being built around the optics. SPECS are standalone, binocular see-through AR glasses running Snap's own OS. Now Snap is adding SPECS Intelligence, an “anticipatory” AI layer designed to understand context and bring information into your field of view before you necessarily ask for it. The use cases are starting to look much broader: → real-time translation across 60 languages → AR navigation → shared multiplayer AR → spatial workspaces → 3D shopping → sports training → enterprise field service and remote support And Snap is bringing in some serious partners: NVIDIA, AWS, Salesforce, Hololight, Shopify, Verizon and others. This is where things become interesting for the AR industry. For years, the discussion around smart glasses has largely been about cameras, audio and AI. SPECS move the discussion back toward true display AR - digital information spatially registered with the real world through a binocular waveguide display. But the launch also highlights how difficult that remains. Early hands-on reports point to the familiar AR trade-offs: a 51° FoV, limited outdoor visibility, relatively bulky optics and interaction that is not yet completely seamless. And perhaps that is the real significance of SPECS. Snap isn't waiting for AR glasses to become indistinguishable from normal eyewear before trying to build the platform. It is putting the hardware into users' hands and attempting to solve optics + AI + operating system + applications + connectivity simultaneously. The question may no longer be whether AI glasses become a major computing platform. It may be whether the industry reaches the necessary optical performance and form factor fast enough to make full AR preferable to today's much simpler camera-and-audio smart glasses. #AugmentedReality #AR #SmartGlasses #Optics #Snap #AI #XR #SpatialComputing