AI News Feed
Market watch
Products & Applications

Tech Roundup: Nvidia Unveils DLSS 5, Dyson Launches CameraJet, GoPro to Be Acquired

Nvidia released DLSS 5 with neural rendering, Dyson launched the CameraJet smart toothbrush in China, and Starman Optical agreed to acquire GoPro for $285 million, among other tech news.

Nvidia's DLSS 5 is built on 3D-oriented neural rendering, adopting a deterministic one-frame-in, one-frame-out model that processes motion vectors, color, albedo, and normal data in real time on a per-frame basis. The computation requirement has dropped from dual GPUs, as originally announced in March, to a single card, achieving a fivefold performance improvement in six months. It runs on the tensor cores of GeForce RTX 50-series desktop and laptop GPUs, supporting real-time rendering up to 4K resolution. Developers can access features such as structural and tone intensity adjustments, plus semantic and engine-level mask controls, through the NVIDIA Streamline framework or the Unreal Engine 5 plugin. At 4K with ultra preset and ray tracing, an RTX 5090 can reach up to 370 frames per second; at 2560x1440, the RTX 5090 hits 590 fps, RTX 5080 410 fps, RTX 5070 Ti 350 fps, and RTX 5070 260 fps. The first game to use DLSS 5, NBA 2K27, along with the corresponding GeForce Game Ready driver, will launch on September 3 at 9 p.m. Pacific time, and will also be available on GeForce NOW Ultimate, which uses cloud servers with RTX 5080 GPUs.

Dyson has opened reservations on JD.com for its CameraJet smart toothbrush in China. The device comes in dawn pink and sea blue, featuring a transparent gravity-defying water tank that stores mouthwash for automatic spraying, and a 100,000-pixel camera that, with built-in algorithms, analyzes 28 oral photos per second to lock onto teeth gaps and guide water flow. The toothbrush's motor provides 58,000 variable-amplitude sonic vibrations per minute, and the spray system uses a 21-degree cone to clean between teeth. The brush head has dual bristle structures, with harder inner bristles for surface stains and softer outer ones for gaps. It offers three modes: auto spray and brush, vibration brushing, and spray-only. An RFID module monitors usage time and reminds users to replace the brush head, while the companion app provides guided brushing and generates trend analysis based on personal habits. The CameraJet is priced at 3,899 yuan and goes on sale September 6.

Sony also expanded its ULT Tower party speaker line with three new models. The ULT Tower 5, priced at $450, is the most compact; the ULT Tower 7, at $650, adds higher output and 360-degree surround audio; and the flagship ULT Tower Max, at $1,500, features a high-power amplifier, a 360-degree audio system, two front woofers, one rear woofer, and four tweeters. All three support ULT1, ULT2, FLAT, and LIVE sound profiles, and come with XLR, USB, and line inputs plus microphone jacks. They are slated for release in September 2026.

Starman Optical has agreed to merge with GoPro through an acquisition worth $285 million. The transaction, expected to close by the end of 2026, will fully repay GoPro's $92 million outstanding debt. GoPro will remain a publicly traded company, with Starman Optical paying $1.14 per share to GoPro shareholders, who will end up owning about 10 percent of the combined company. Starman Optical is an entity under Starman Holding, which owns accessory brands Incase, Incipio, and Griffin. Registered in Delaware on August 31, 2026, it operates through Starman New Photonics, a business established in 2025 that develops and manufactures optical transceivers and photonics technology in the United States. After the merger, GoPro will expand beyond action cameras into photonic infrastructure and defense, incorporating Starman's optical transceivers to enter the optical infrastructure market for AI data centers. Combined with GoPro's more than 2,500 U.S. patents and optical IP, the new company will target defense, government, robotics, and aerospace, while pushing U.S.-based manufacturing. GoPro's existing consumer cameras, cloud platform, and subscription services will continue to be supported.

Google has introduced agentic video understanding for its Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite models. Instead of extracting media at a fixed frame rate of 1 FPS, the models now dynamically retrieve and examine frames, audio, and subtitle text through built-in native tools. This enables sub-second event retrieval, long-video search, high-frame-rate resampling for anomaly detection, and accurate counting of actions and objects. In benchmarks, Gemini 3.7 Flash with this feature cut token consumption by up to 88 percent, lowered analysis costs by as much as 66 percent, and improved accuracy by up to 7 percent. The capability is available today in Google AI Studio's Gemini API and the Gemini Enterprise Agent Platform, supporting uploaded videos and YouTube processing; developers can enable it by setting the API processing parameter to "agentic," and it is billed at standard Gemini API token rates with no extra feature fee. Consumer access will arrive later in the Gemini app and within months in the Ask YouTube feature on YouTube video pages.

South Korea's government has launched "AI for All," a program that provides free generative AI to all citizens, following a decision on August 29. Three technology consortia, led by the country's two largest telecom companies and Kakao, will deliver the service. The government will allocate up to 512 Nvidia B200 chips to the consortia and help cover operating costs, offering unlimited token access at no charge. The service will connect directly to government systems to support medical appointments, housing searches, tax calculations, and educational recommendations. South Korea's 2026 AI budget is about 10 trillion won, triple last year's allocation; roughly a quarter of the population currently pays for AI services, and over 20 million people use free generative AI tools. A beta test is set to begin in September 2026, with the full launch planned within the year.

In other news, Google removed all remaining Manifest V2 extensions from the Chrome Web Store, including uBlock Origin. Users on Chrome 138 or earlier can still use installed MV2 extensions but will not receive updates, and uninstalled extensions cannot be reinstalled from the store. Brave has responded by hosting four commonly used MV2 extensions on its own backend—AdGuard, uBlock Origin, uMatrix, and NoScript. Google also rolled out its September Android Drop, adding features like Find Hub support for unlabeled items via Gemini on Android 16+ and Motion Assist on Android 17 to reduce motion sickness. TP-Link's Archer 8 Ultra BN19000 and Deco 8 Ultra routers, the world's first Wi-Fi 8 routers, went on sale August 30, according to Tom's Hardware. WeChat Pay announced on August 31 that its AI-exclusive card now supports DeepSeek Harness and OpenClaw, enabling users to complete purchases through AI conversations. Jiemian News reported that OPPO plans to raise prices on its Reno 16 and OnePlus Turbo 6X Pro series by 300 to 500 yuan this month.

MapQuest's iOS app rose to No. 1 in navigation, No. 4 in apps, and No. 8 overall in the U.S. App Store, and No. 2 overall in Canada, after the company refused to rename Lake Ontario to "Lake America" as ordered by a presidential executive order. Google Maps complied for U.S. users on August 29, while Apple Maps had not changed the name or issued a statement as of August 31. Phil Schiller stepped down from running the App Store and became an Apple Fellow, with day-to-day operations passed to Carson Oliver, a 14-year Apple veteran; the business was also moved from marketing into Eddy Cue's services division. In the trade-secret lawsuit against OpenAI, Apple submitted new evidence accusing former engineer Chang Liu, now at OpenAI, of using confidential circuit schematics and internal-tool names, and of allegedly conspiring with colleague Yu-Ting Peng to destroy evidence after learning of the investigation in June. Apple also said Liu had exploited a rare unknown authentication vulnerability to retain access, contrary to OpenAI's claims of residual access and providing assistance to former colleagues.