In another twist on the OpenAI saga that raged over the weekend, Microsoft has swooped in and hired Sam Altman and Greg Brockman just after OpenAI confirmed Altman's firing, CEO Satya Nadella announced in a post on X. The pair, along with colleagues are joining Microsoft "to lead a new advanced AI research team." Nadella added that Microsoft "remain[s] committed to our partnership with OpenAI," but the move looks a giant hedge on that bet. It also means the much of OpenAI's talent may be moving to Microsoft, which may stall the latter's progress with its widely-used ChatGPT product.
We remain committed to our partnership with OpenAI and have confidence in our product roadmap, our ability to continue to innovate with everything we announced at Microsoft Ignite, and in continuing to support our customers and partners. We look forward to getting to know Emmett…
The appoint was announced just after it was suggested that Altman may return to OpenAI, following his sudden firing on Friday. That was preceded by an outpouring of support for Altman from OpenAI employees, many of whom shared hearts on social media.
In a bit of a surprising development, though, talks between the parties broke down and OpenAI subsequently said that Altman (and co-founder Greg Brockman) would not return, according to reports. Instead, Twitch co-founder Emmett Shear was appointed as the new CEO.
Developing...
This article originally appeared on Engadget at https://www.engadget.com/microsoft-snatches-sam-ultman-and-former-openai-colleagues-to-form-its-own-ai-research-team-082755226.html?src=rss
In May 2022, just months before Elon Musk acquired Twitter (now X), the Federal Trade Commission (FTC) hit the platform with a $150 million fine and consent decree that imposed stipulations for safeguarding users' personal data. The ruling was imposed after the FTC found that Twitter had violated a previous consent decree by inappropriately sharing user phone numbers and email addresses with advertisers.
After acquiring Twitter, Elon Musk went to court in an attempt to overturn that consent decree, arguing that the FTC had increased scrutiny on X and pressured a third-party assessor to find fault with its security practices. In a post on X at the time, Musk called the FTC's actions "a shameful case of weaponization of a government agency for political purposes and suppression of the truth."
However, a federal judge just rejected that motion, ruling that his court lacked the authority to block the FTC's court order, while also blocking Musk's request to avoid a deposition that's part of a separate investigation, The Washington Post reported. In other words, X must still adhere to the strict privacy reporting standards imposed by the FTC last year.
The judge listed other issues with X's argument, noting that the FTC had good reason to increase scrutiny on X after Musk took over. "The government says this increase in investigative activity should not be surprising because Musk directed at least five rounds of terminations, layoffs or other reductions in X Corp.’s workforce, which affected the security, governance, risk and compliance team," he wrote in the ruling.
He added that the third-party assessor said he faced steady delays due to executive turnover at X. Finally, the judge ruled that Musk can't avoid being deposed by the FTC, as "the government argues that the major changes to the company appear to have been initiated by Musk himself."
Musk's FTC problems started shortly after his acquisition of X, when key executives on Twitter's privacy and security teams departed. Still, Musk seemed willing to comply with the rules, telling employees at the time that "Twitter will do whatever it takes to adhere to both the letter and spirit of the FTC consent decree."
Back in March, though, the FTC stepped up its probe into Musk's actions in an effort to get more information about the company’s handling of layoffs, X Premium subscriptions, the "Twitter Files" and other issues.
Meanwhile, the platform can't seem to stay out of the news, as IBM recently said it would pull advertising on X after discovering its ads were being placed next to pro-Nazi content. Musk himself faced criticism after he appeared to endorse an antisemitic post on the platform.
This article originally appeared on Engadget at https://www.engadget.com/judge-rejects-attempt-by-x-to-nullify-ftc-privacy-settlement-111536396.html?src=rss
Kia's concept EVs often get more than the usual amount of attention because it tends to incorporate a lot of the ideas in the final production models — with the EV9 electric SUV being a good example of that. With that in mind, the company just unveiled the Concept EV3 and Concept EV4 representing its "Opposites Unite" design philosophy, with an emphasis on interior vibes, geometric shapes and sustainable materials.
The smallish crossover Concept EV3 looks more production-ready, and Kia says it "offers the technology, practicality and design of [our] flagship SUV, the EV9." That may be true, but the EV3 carries a bold, futuristic design as well. Like the EV9 concept, it has distinct geometric patterns including octagonal wheel wells that flare out considerably from the body, and echoes those shapes in the side skirts and taillights.
Kia
The windshield is pushed forward to open up interior space, while the glass roofline slopes front to back and appears to float thanks to the disconnected C pillar. "The unexpected treatment of volumes and graphics simultaneously interlock with each other, creating a logical yet emotional design," the company added in some top-tier concept car PR fluff.
Kia
The interior is more fanciful, with a nearly button-free layout designed to emphasize the voluminous interior and create a "transformative cabin ambience," Kia said. It features a horizontal screen mounted on top that functions as both gauge cluster and infotainment system, along with an oval steering wheel and real mirrors. The wide center console "table" converts to four modes, including "focus, social, refreshing and storage." Meanwhile, the fiber weave materials and bioplastics make the cabin inviting and modern.
Kia
The Concept EV4 is a clean-sheet design with an emphasis on sportiness that also uses some strong geometric shapes. To that end, it has a low nose and long-tail silhouette with a wide stance, along with a "technical roof spoiler" — two small extensions on either side of the roof. Again, we're seeing bulgy fenders, lots of bold lines and creases, a sloping glass roof and geometric lighting front and back.
Kia
The interior is broadly the same as the Concept EV3, but an "HVAC user interface incorporates a control panel that can be neatly stored in the center console when not in use," according to Kia. It incorporates handwoven fabrics in the vehicle's storage area and dashboard to create a 3D effect, and features natural dies and recycled cotton in the interior.
Kia
Kia didn't get into the power, battery or other powertrain features of either car, and painted both as test cases to see what people think. "That’s why we came to Los Angeles and America’s savviest EV consumers, to gauge reaction to two potential opportunities within the Kia lineup," said Kia America's COO Steven Center. It looks like Kia has nailed these designs once again, so don't be surprised to see them appear in some form as real production vehicles.
This article originally appeared on Engadget at https://www.engadget.com/kias-latest-ev-concepts-go-big-on-geometrics-and-cabin-vibes-190529735.html?src=rss
Tags are a key way to seek out content on social media, but so far they've been missing on Meta's fledgling Threads platform. That's changing soon, however, as the feature is now in testing on Threads in Australia "with more countries coming soon," Mark Zuckerberg wrote in a new thread.
You create a tag in the usual way by placing a hash before a word, which then displays in blue text without the hash — much as mentions work in Facebook. To seek out topics, type a hash plus a keyword into the search field to see a list of relevant posts, as one does on Instagram. So far, it's limited to a single tag per post, likely to discourage hashtag spamming — though that may change, as Meta said the feature is still a work in progress.
Despite still not being in Europe, Threads has shown consistent growth and now counts nearly 100 million monthly active users, Zuckerberg wrote last month. The app recently gained a few key features like the ability to delete your threads profile without killing your Instagram account and avoid automatically sharing Threads posts with Facebook and Instagram. It also added pinned posts, and Instagram boss Adam Mosseri hinted that DMs may (or may not) be done via Instagram's inbox.
This article originally appeared on Engadget at https://www.engadget.com/threads-introduces-tags-to-help-users-categorize-posts-092650879.html?src=rss
NVIDIA has launched its next-generation of AI supercomputer chips that will likely play a large role in future breakthroughs in deep learning and large language models (LLMs) like OpenAI's GPT-4, the company announced. The technology represents a significant leap over the last generation and is poised to be used in data centers and supercomputers — working on tasks like weather and climate prediction, drug discovery, quantum computing and more.
The key product is the HGX H200 GPU based on NVIDIA's "Hopper" architecture, a replacement for the popular H100 GPU. It's the company's first chip to use HBM3e memory that's faster and has more capacity, thus making it better suited for large language models. "With HBM3e, the NVIDIA H200 delivers 141GB of memory at 4.8 terabytes per second, nearly double the capacity and 2.4x more bandwidth compared with its predecessor, the NVIDIA A100," the company wrote.
In terms of benefits for AI, NVIDIA says the HGX H200 doubles inference speed on Llama 2, a 70 billion-parameter LLM, compared to the H100. It'll be available in 4- and 8-way configurations that are compatible with both the software and hardware in H100 systems. It'll work in every type of data center, (on-premises, cloud, hybrid-cloud and edge), and be deployed by Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure, among others. It's set to arrive in Q2 2024.
NVIDIA
NVIDIA's other key product is the GH200 Grace Hopper "superchip" that marries the HGX H200 GPU and Arm-based NVIDIA Grace CPU using the company's NVLink-C2C interlink. It's designed for supercomputers to allow "scientists and researchers to tackle the world’s most challenging problems by accelerating complex AI and HPC applications running terabytes of data," NVIDIA wrote.
The GH200 will be used in "40+ AI supercomputers across global research centers, system makers and cloud providers," the company said, including from Dell, Eviden, Hewlett Packard Enterprise (HPE), Lenovo, QCT and Supermicro. Notable among those is HPE's Cray EX2500 supercomputers that will use quad GH200s, scaling up to tens of thousands of Grace Hopper Superchip nodes.
NVIDIA
Perhaps the biggest Grace Hopper supercomputer will be JUPITER, located at the Jülich facility in Germany, which will become the "world’s most powerful AI system" when it's installed in 2024. It uses a liquid-cooled architecture, "with a booster module comprising close to 24,000 NVIDIA GH200 Superchips interconnected with the NVIDIA Quantum-2 InfiniBand networking platform," according to NVIDIA.
NVIDIA says JUPITER will help aid scientific breakthroughs in a number of areas, including climate and weather prediction, generating high-resolution climate and weather simulations with interactive visualization. It'll also be employed for drug discovery, quantum computing and industrial engineering. Many of these areas use custom NVIDIA software solutions that ease development but also make supercomputing groups reliant on NVIDIA hardware.
The new technologies will be key for NVIDIA, which now makes most of its revenue from the AI and data center segments. Last quarter the company saw a record $10.32 billion in revenue in that area alone (out of $13.51 billion total revenue), up 171 percent from a year ago. It no doubt hopes the new GPU and superchip will help continue that trend. Just last week, NVIDIA broke its own AI training benchmark record using older H100 technology, so its new tech should help it extend that lead over rivals in a sector it already dominates.
This article originally appeared on Engadget at https://www.engadget.com/nvidia-announces-its-next-generation-of-ai-supercomputer-chips-140004095.html?src=rss
Microsoft's Xbox leadership is starting to take shape following the company's blockbuster Activision Blizzard acquisition. The new head of Xbox Game Studios is Alan Hartman, former boss of Forza Motorsport developer Turn 10, according to a LinkedIn post seen by Game Developer. "Thrilled to be representing such an elite set of game creators in my new role leading Xbox Game Studios!" he wrote.
Hartman has led Turn 10 since 2005, with recent stints as corporate VP of Forza and Fable. He oversaw the release of Forza series games and the ForzaTech engine, which is used in both Forza Horizon and the upcoming Fable reboot.
Hartman succeeds Matt Booty, who was promoted to president of gaming content and studios. Other execs recently changing positions include Sarah Bond, who became Xbox's first black president in its 22 year history. She reports directly to Microsoft Gaming president Phil Spencer. More shakeups could be coming with Microsoft's $69 billion Activision Blizzard acquisition, as properties like Call of Duty, World of Warcraft, Overwatch and others now fall under the company's purview. Bobby Kotick remains as CEO of Activision Blizzard, but will step down at the end of 2023.
This article originally appeared on Engadget at https://www.engadget.com/turn-10s-alan-hartman-is-the-new-head-of-xbox-game-studios-095546957.html?src=rss
Meta hasn't been in China since Facebook was blocked way back in 2009, but that may be about to change. The company is reportedly planning to release an all-new, lower-priced version of its virtual reality headset in the nation via an exclusive deal with video game giant Tencent, The Wall Street Journal has reported. Sales are tentatively set to begin in late 2024, but some details must still be finalized.
The new headset would be more powerful than the Quest 2, but use lower-quality optics than the Quest 3, according to the report. The more budget-oriented headset may also be sold in other regions. Meta would reportedly take a larger share of headset sales, while Tencent pull in more of the content and service revenue, "like software subscriptions and game sales."
The deal would open up a huge market for Meta's VR division, which has been hemorrhaging money and could certainly use the boost. However, it's not clear whether Tencent would require government approval to sell the devices. Gaming is a popular VR activity, but China's strict rules have already had a significant impact on Tencent, the world's largest video game company.
China may not be a panacea for Meta's weak VR division, either. TikTok owner ByteDance is China's virtual reality leader with its Pico headset, but is struggling with sales in China just as Meta is elsewhere. And Tencent itself was reportedly on the brink of disbanding its own VR division, but supposedly built it back up once the Meta deal seemed inevitable. Headset sales across the globe fell nearly 45 percent this quarter compared to the same period last year.
If the deal goes through, Meta would regain a foothold in China after 14 years with no direct presence there. The company has a 50 percent worldwide share of the VR market, with Sony's PlayStation VR2 and Pico's VR headset in second and third place. Apple is about to enter the market with its $3,500 Vision Pro headset, but isn't likely to have a serious presence until it releases a cheaper version down the road.
This article originally appeared on Engadget at https://www.engadget.com/meta-may-return-to-china-with-the-release-of-a-new-budget-vr-headset-090522700.html?src=rss
Google is rolling out a new feature that allow advertisers to create AI generated content using the same technology as the Bard chatbot, confirming a report from earlier this year. The feature is now available in beta on Google's Performance Max advertising product, allowing US advertisers to create and scale text and image assets for campaigns using AI, the company announced in a blog post.
Performance Max is already an AI-powered product that works across multiple Google products including Youtube, search, display and others. It optimizes ads by analyzing performance data, and the new feature supplements that by using AI to assist in asset creation as well. As Google puts it, the features will allow advertisers to quickly create high-quality, personalized assets on various Google platforms.
"Asset variety is a key ingredient for a successful Performance Max campaign," wrote Google's Pallavi Naresh. "You’ve told us that creating and scaling assets can be one of the hardest parts of building and optimizing a cross-channel campaign. Now, you’ll be able to generate new text and image assets for your campaign in just a few clicks."
Google
Much like Bard or ChatGPT, users feed prompts to the AI, and it creates unique images and text for each business. Marketers can review and edit any assets created by the system prior to publication. It can be used to create versions of the same ad, or build new ads from scratch. All AI-generated imagery contains a visible watermark and is tagged as such. "We also have guardrails in place to prevent our systems from engaging with inappropriate or sensitive prompts or suggesting policy-violating creatives," Naresh wrote.
The feature should help marketers create advertising materials more quickly, while of course helping Google post those ads and make money more quickly. In that sense, it's pretty much a perfect AI use case for Google, which makes the vast majority of its revenue from advertising. The new system is currently in beta and only available in the US, but is expected to roll out more widely by the end of 2023.
This article originally appeared on Engadget at https://www.engadget.com/google-is-rolling-out-tools-that-let-advertisers-create-ai-generated-content-080255864.html?src=rss
Many Pixel owners have been left with a bad taste in their mouths after it took Google over a month to fix a serious bug, Ars Technica has reported. It first appeared with the launch of Android 14 back on October 8th, locking some users with multiple accounts out of their device's local storage. It affects multiple devices ranging from the Pixel 4 to the Pixel 8, and for many users, it was akin to being locked out of their phone by ransomware.
Some folks were unable to unlock their devices, while others were able to boot up but had no access to lock storage. However, the bug rendered some phones completely unusable, as they would continuously bootloop and never reach the home screen. Reports of the issue appeared shortly after Android 14 launched, but Google kept rolling out the buggy release and only acknowledged the flaw some 20 days after it appeared.
The November update patch is now rolling out, but the initial November 2 release notes weren't very positive. Google said users locked out of their storage may only get some data back, and those experiencing a bootloop may lose everything. Today's update, however, states that users who were unable to access media storage should get all their data back once they install the November patch.
Those stuck in a reboot may not be as lucky, though. Those folks will be able to at least get up and running again after submitting a form. However Google said that "data recovery solutions are still being investigated for devices that are repeatedly rebooting," adding that "we'll share additional updates soon."
The sordid episode shows how Google failed to properly implement its own much-touted failsafe systems, as Ars Technica noted. It kept rolling out Android 14 with the flaw despite multiple reports, and the vaunted dual partition system didn't work because it didn't accurately detect a boot failure. Finally, it took Google ages to elevate the issue to a higher priority, leaving many users stuck with bricked phones for weeks. "Little did I realize that 'seven years of updates' was not a feature, but a threat," said one disgusted user on Google's issue tracker.
This article originally appeared on Engadget at https://www.engadget.com/google-has-fixed-a-bug-in-android-14-that-locked-pixel-users-out-of-their-devices-061040556.html?src=rss
Sony has just unveiled its latest pro-oriented mirrorless camera, and we now know why it took them four years to develop it The Alpha A9 III is the first full-frame camera on the market with a global stacked sensor, a sort of holy grail in the photographic world. It allows for some wild specs, like 120fps shooting speeds with no blackout, up to a 1/80,000th of a second shutter speed and zero rolling shutter — albeit with a 24.6 megapixel resolution that may disappoint some.
The new tech opens up a lot of new possibilities for photographers. Without the constraints of a physical shutter (no, there isn't one), it can shoot full-resolution, blackout-free 14-bit RAW bursts at up to 120fps with a 1.6 second buffer (around 180 shots). Sony also installed the latest Bionz XR image processor and "high-density focal plane phase detection AF" that allows for real-time autofocus (AF) tracking.
Sony
"A designated AI processing unit uses real-time recognition AF to recognize a wide variety of subjects with high precision," Sony explains in the press release. "By combining high-speed performance of up to 120 fps with highly accurate subject recognition performance, it is possible to easily photograph scenes and moments that cannot be seen with the naked eye."
Sony
The global shutter also allows for shutter speeds of 1/80,000th of a second (1/16,000th during continuous shooting), ten times faster than most cameras. It also means that compatible flashes can be synced all the way up to the maximum shutter speed, rather than being limited to much lower speeds in electronic shutter mode — normally around 1/250th to 1/500th of a second. It also allows the shutter speed to be finely adjusted to eliminate flicker in video. And Sony is offering a 1-second pre-burst feature that can capture scenes before the shutter button is pressed, reducing the possibility of a missed shot.
Sony
The stacked global shutter provides large benefits for video, as well. It's Sony's first camera to support 4K 120p video with no cropping and does so with no rolling shutter distortion (skewing), along with 4K 60p with 6K oversampling. It also offers 10-bit recording with S-Log3 capture in all video modes (including 4K 120p), along with S-Cinetone borrowed from the company's high-end Venice lineup that "makes human skin tones and subjects stand out beautifully," Sony said. It should be one of Sony's best-focusing cameras for video, as all the photo subject tracking features work in that mode, too.
It's got a long list of other features you'd expect on a high-end Sony camera, most notably in-body stabilization with up to 8 stops of shake reduction. The electronic viewfinder is Sony's best with 9.44 million dots, and It resolves the rear display tilt vs. flip debate by doing both of those things. And to help clear the buffer as quickly as possible, it supports fast CFexpress Type A cards on top of SD UHS II, much like the Alpha A1.
Sony
Other specs show an ISO range of 250-25600 (expandable to ISO 125–51200) with a minimum ISO that's a bit on the high side. That shouldn't bother photographers on sunny days, given the extremely fast shutter speeds, but video shooters will need ND filters to block some light.
What will be a key for this camera is the image quality and low-light sensitivity, given that this is an all-new sensor and there's not a lot of data on global shutters. Sony didn't release any sample photos yet from what I've seen, so upcoming reviews of this model will be key. The Alpha A9 III goes on pre-order tomorrow for $6,000, with a vertical grip available for $400 — but won't arrive until next spring.
This article originally appeared on Engadget at https://www.engadget.com/the-sony-a9-iii-is-the-fastest-full-frame-camera-ever-thanks-to-a-global-stacked-sensor-175522418.html?src=rss