Microsoft launched a new artificial intelligence model today that achieves remarkable mathematical reasoning capabilities while using far fewer computational resources than its larger competitors. The 14-billion-parameter Phi-4 frequently outperforms much larger models like Google’s Gemini Pro 1.5, marking a significant shift in how tech companies might approach AI development.
The breakthrough directly challenges the AI industry’s “bigger is better” philosophy, where companies have raced to build increasingly massive models. While competitors like OpenAI’s GPT-4o and Google’s Gemini Ultra operate with hundreds of billions or possibly trillions of parameters, Phi-4’s streamlined architecture delivers superior performance in complex mathematical reasoning.
Small language models could reshape enterprise AI economics
The implications for enterprise computing are significant. Current large language models require extensive computational resources, driving up costs and energy consumption for businesses deploying AI solutions. Phi-4’s efficiency could dramatically reduce these overhead costs, making sophisticated AI capabilities more accessible to mid-sized companies and organizations with limited computing budgets.
This development comes at a critical moment for enterprise AI adoption. Many organizations have hesitated to fully embrace large language models due to their resource requirements and operational costs. A more efficient model that maintains or exceeds current capabilities could accelerate AI integration across industries.
Mathematical reasoning shows promise for scientific applications
Phi-4 particularly excels at mathematical problem-solving, demonstrating impressive results on standardized math competition problems from the Mathematical Association of America’s American Mathematics Competitions (AMC). This capability suggests potential applications in scientific research, engineering, and financial modeling — areas where precise mathematical reasoning is crucial.
The model’s performance on these rigorous tests indicates that smaller, well-designed AI systems can match or exceed the capabilities of much larger models in specialized domains. This targeted excellence could prove more valuable for many business applications than the broad but less focused capabilities of larger models.
Microsoft emphasizes safety and responsible AI development
The company is taking a measured approach to Phi-4’s release, making it available through its Azure AI Foundry platform under a research license agreement, with plans for a wider release on Hugging Face. This controlled rollout includes comprehensive safety features and monitoring tools, reflecting growing industry awareness of AI risk management.
Through Azure AI Foundry, developers can access evaluation tools to assess model quality and safety, along with content filtering capabilities to prevent misuse. These features address mounting concerns about AI safety while providing practical tools for enterprise deployment.
Phi-4’s introduction suggests that the future of artificial intelligence might not lie in building increasingly massive models, but in designing more efficient systems that do more with less. For businesses and organizations looking to implement AI solutions, this development could herald a new era of more practical and cost-effective AI deployment.
Alphabet Inc.’s Google and Samsung Electronics Co. unveiled a joint push into the mixed-reality market, introducing a new operating system and headset in a bid to challenge devices from Apple Inc. and Meta Platforms Inc.
In what they called a collaboration as “one team,” the two companies announced a version of Google’s Android software for XR — shorthand for extended reality, which refers to a range of virtual- and augmented-reality technologies. They also showed off a Samsung-built headset code-named Project Moohan, taken from the Korean word for “infinite.”
The two tech giants look to jump-start a market that’s been slow to take off. Apple’s $3,499 Vision Pro headset, released this year, remains a niche product — held back by its burdensome weight and hefty price tag. And Meta has had more success with smart glasses and cheaper VR headsets than higher-end mixed-reality devices.
The new Android will allow a range of companies to design their own XR devices — both headsets and lighter-weight glasses — while also taking advantage of the latest artificial intelligence advances. The hope is to replicate the success Google had with the original version of Android, which is used by most major smartphones. Companies like Sony Group Corp., Xreal Inc., and Lynx Mixed Reality have committed to build devices running the new operating system, Google said.
“The time for XR is now,” Sameer Samat, a Google executive who oversees the Android ecosystem, said in an interview. “We’re not strangers to this space,” he said, referring to Google Glass, a precursor to today’s devices that flopped a decade ago. “The technology wasn’t quite ready at the time, but we never stopped believing in the vision of what XR could be.”
Mixed-reality devices blend real-world inputs with virtual reality, an immersive experience that’s long been associated with gaming. Augmented reality refers to the overlaying of data and graphics on the things users are seeing around them.
Samat said Google had been developing Android XR for several years, but “took a pause” to rebuild the software around AI. The company’s AI features are now core to the operating system. For instance, glasses with Android XR can use external cameras to analyze a newly bought piece of furniture and then provide instructions to assemble it. On a headset, it can control actions like mapping out a vacation in Google Maps.
The companies are showcasing Android XR to potential partners at a developer-focused event in New York on Thursday. The first Android XR product to market will be the Samsung headset, which is set to debut sometime next year.
In contrast with Google’s approach, Apple has no plans to offer the Vision Pro’s software — visionOS — to outside device makers. For its part, Meta said this year that it’s developing a version of its Horizon operating system for third-party headsets, including ones from Asustek Computer Inc. and Lenovo Group Ltd. Meta also has previewed a pair of next-generation AR glasses called Orion. That’s a category Apple has yet to commit to publicly.
“The Apple Vision Pro is an amazing device, but it’s one device,” Samat said. “We see a range of devices that will solve different problems in the ecosystem.”
He believes Android could power specialized headsets for gaming, watching videos, performing surgery, or even professional cycling. “We expect Android XR to cover all of this,” Samat said.
During a series of demonstrations at Google’s offices in Mountain View, California, the two companies showed the capabilities of Samsung’s headset as well as several Google-developed glasses prototypes.
The Samsung device resembles the Vision Pro, including its curved front, built-in speakers, and high-resolution displays. It also has a similar ability to toggle between VR and AR modes, and an external battery pack that connects over a cable.
The hardware itself is encased in gray plastic with cushions, and it’s able to be worn with or without a light seal — a component that blocks outside light from hindering the experience. The battery pack is interchangeable and connects over USB-C. Samsung said that a range of different packs could be available with varying levels of battery life.
The major difference from the Apple Vision Pro is that the Samsung device feels lighter and more comfortable to wear for extended periods. It also features a large cushion for the back of a user’s head, plus an adjustable plastic piece that goes around the head. And the field of view on the Samsung device appears larger than on the Vision Pro, allowing a user to see more content at once.
The idea is to better distribute the weight, said Won-Joon Choi, head of research and development at Samsung’s mobile division. “We’ve done tons of studies” using people’s head data and simulations, he said, adding that “first impressions are very important.” The company believes comfort should be one of its top priorities when building a wearable device, Choi said.
With the Samsung product, users can double-tap the side to move between AR and VR mode. That means they can switch between seeing the outside world and more immersive environments, such as a mountain range or other VR backgrounds. Likewise, the Vision Pro allows users to choose their level of immersion by turning a watch-like crown.
The Android XR operating system itself is highly reminiscent of visionOS, allowing users to fill their space with multiple windows that can be moved and resized. It includes many of the core Android apps and Google services, including Maps, YouTube, Google TV, Photos, Docs and Chrome. The software can be used like a standard computer and controlled by a mouse and keyboard.
The Samsung device can also be operated by hand-and-eye control. The big difference from the Vision Pro — at least with the prototype device — is a user needs to lift their hands up to control certain items while looking at them. Vision Pro users are able to leave their hands in their laps.
Overall, the interface was smooth to operate and included features that users expect from VR headsets, including immersive video playback. Like the latest version of visionOS, the software also includes a feature that can turn photos into 3D memories. It has a leg up on the Apple device, though, by providing the same feature for previously recorded videos.
When it comes to providing content for the device, Google is planning to rely heavily on stereoscopic video that’s already available on YouTube. Apple has been funding its own original VR video for the Vision Pro and has been slowly making more material available. Like Apple, Google also will offer a software development kit for third-party applications. And it will run existing Android phone and tablet apps that users can access from the Google Play Store.
Samsung wouldn’t say how much the headset will cost or exactly when it will go on sale next year, though the company hinted that it would undercut the Vision Pro’s $3,499 price.
“We’re doing a lot of research on what price point customers will feel comfortable paying,” Choi said. A Samsung representative added the company is “fully aware” that price has previously prevented this technology from being widely adopted.
The companies said the Samsung device is powered by Qualcomm Inc.’s XR2 Gen 2 chip for mixed-reality devices. Samsung representatives said the company started with a mixed-reality headset because it provides the “richest” experience, though glasses are also on its product roadmap.
Google has its own hardware engineering team, which develops Pixel phones, smartwatches ,and other gadgets, but the company has no current plans to sell another VR headset. In 2016, Google launched Daydream, a VR headset that required the insertion of a phone. It was discontinued in 2019.
Samsung, which has its own Tizen operating system for TVs, watches and certain phones, said it’s planning to work with Android XR over the long term. “We don’t see the benefit of working on another platform,” Choi said.
The first glasses prototypes demonstrated by Google are part of its Project Astra initiative. Android XR glasses can use AI to get chores done, in addition to offering features like turn-by-turn directions for pedestrians.
The prototypes rely on microLED-type technology created by Raxium, a company Google acquired two years ago to help it build augmented-reality glasses. The approach allows for bright images without using a lot of power. The glasses rely on external cameras to gather data and can sync up with a phone.
Google didn’t say when it expects the first glasses with Android XR to be ready. The company is giving prototype pairs to testers and will “bring something to market when it makes sense,” Samat said.
Throw a stone and you’ll likely hit a deepfake. The commoditization of generative AI has led to an absolute explosion of fake content online: According to ID verification platform Sumsub, there’s been a 4x increase in deepfakes worldwide from 2023 to 2024. In 2024, deepfakes accounted for 7% of all fraud, per Sumsub, ranging from impersonations and account takeovers to sophisticated social engineering campaigns.
In what it hopes will be a meaningful contribution to the fight against deepfakes, Meta is releasing a tool to apply imperceptible watermarks to AI-generated video clips. Announced on Thursday, the tool, called Meta Video Seal, is available in open source and designed to be integrated into existing software. The tool joins Meta’s other watermarking tools, Watermark Anything (re-released today under a permissive license) and Audio Seal.
“We developed Video Seal to provide a more effective video watermarking solution, particularly for detecting AI-generated videos and protecting originality,” Pierre Fernandez, AI research scientist at Meta, told TechCrunch in an interview.
Video Seal isn’t the first technology of its kind. DeepMind’s SynthID can watermark videos, and Microsoft has its own video watermarking methodologies.
But Fernandez asserts that many existing approaches fall short.
“While other watermarking tools exist, they don’t offer sufficient robustness to video compression, which is very prevalent when sharing content through social platforms; weren’t efficient enough to run at scale; weren’t open or reproducible; or were derived from image watermarking, which is suboptimal for videos,” Fernandez said.
In addition to a watermark, Video Seal can add a hidden message to videos that can later be uncovered to determine their origins. Meta claims that Video Seal is resilient against common edits like blurring and cropping, as well as popular compression algorithms.
Fernandez admits that Video Seal has certain limitations, mainly the trade-off between how perceptible the tool’s watermarks are and their overall resilience to manipulation. Heavy compression and significant edits may alter the watermarks or render them unrecoverable, he added.
Of course, the bigger problem facing Video Seal is that devs and industry won’t have much reason to adopt it, particularly those already using proprietary solutions. In a bid to address that, Meta is launching a public leaderboard, Meta Omni Seal Bench, dedicated to comparing the performance of various watermarking methods and organizing a workshop on watermarking this year at ICLR, a major AI conference.
“We hope that more and more AI researchers and developers will integrate some form of watermarking into their work,” Fernandez said. “We want to collaborate with the industry and the academic community to progress faster in the field.”
You’ll be able to pick call participants from a group chat, use new effects on calls, and take advantage of calls with higher resolution video, according to WhatsApp. The desktop app is getting a new calls tab, too.
:format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25787896/470070852_435745649585095_1517067625928730252_n.png)