Latest news with #machinelearning
Wall Street Journal
a day ago
- Science
- Wall Street Journal
It's Known as ‘The List'—and It's a Secret File of AI Geniuses
All over Silicon Valley, the brightest minds in AI are buzzing about 'The List,' a compilation of the most talented engineers and researchers in artificial intelligence that Mark Zuckerberg has spent months putting together. Lucas Beyer works in multimodal vision-language research and describes himself as 'a scientist dedicated to the creation of awesomeness.' Yu Zhang specializes in automatic speech recognition and barely has an online presence besides his influential papers. Misha Bilenko is an expert in large-scale machine learning who also enjoys hiking and skiing—or, as he puts it on his website, 'applying hill-climbing search and gradient descent algorithms to real-world domains.'
Phone Arena
a day ago
- Phone Arena
Google Photos just got an update that changes how HDR photos are edited
Google has rolled out a new update for the Photos app that improves how both HDR and SDR images are edited. The goal is to help users keep more of the original detail in their photos, especially when editing images taken in high dynamic range. Photos shot in HDR now keep their full dynamic range and HDR metadata, even after using editing tools like Photo Unblur, Magic Eraser, or Portrait Light. This means the contrast and brightness in the photo should still look accurate after editing, which wasn't always the case before. Google is also adding a new tool called "Ultra HDR" that lets users control how bright an image appears. At the same time, the previous "HDR" slider has been renamed to "Tone," which better describes its role in adjusting contrast and exposure. These changes should make the editing process a bit clearer for users who want to fine-tune their photos. The new "Ultra HDR" and "Tone" tools in Google Photos. | Images credit — PhoneArena One of the more notable changes is the ability to enhance SDR (Standard Dynamic Range) photos to appear more like HDR. This feature uses machine learning to boost highlights, deepen contrast, and improve the overall clarity of older photos or those not originally taken in HDR. While it doesn't turn them into true HDR files, the result is designed to look more vivid on compatible screens. When comparing this to the Photos app on iOS, Google's approach offers more manual control. On iPhones, HDR processing happens automatically with limited editing tools afterward. Apple does preserve HDR metadata, but it doesn't let users convert SDR to HDR or adjust the look of HDR photos to the same extent. In that sense, Google's update gives users more flexibility in how their photos appear. These improvements follow a broader trend of using AI in mobile photo editing, especially on Pixel devices. While some features like Magic Editor aim to make big changes to a photo, this HDR update is more about keeping photo quality consistent after minor edits. In my opinion, this update adds more options without changing how the app works at its core. Some users may need time to get used to the renamed tools, but for those with HDR-capable displays, these improvements could make a noticeable difference in how photos look after editing. Secure your connection now at a bargain price! We may earn a commission if you make a purchase Check Out The Offer
Globe and Mail
a day ago
- Business
- Globe and Mail
Can Advanced Micro Devices Aid Its Data Center Revenues With GPUs?
Advanced Micro Devices AMD is benefiting from strong demand for its Graphics Processing Units (GPUs), driven by the increasing adoption of artificial intelligence (AI) and machine learning techniques. This growing demand is reflected in AMD's financial performance. In the first quarter of 2025, Data Center revenues surged 57.2% year over year to $3.674 billion, accounting for 49.4% of total revenues. The surge was primarily driven by growth in AMD EPYC CPU and AMD Instinct GPU sales. AMD is strengthening its footprint in the artificial intelligence market through an expanding portfolio. AMD recently introduced its comprehensive, end-to-end AI platform at the 2025 Advancing AI event, introducing the powerful Instinct MI350 Series GPUs with 4x generational AI compute gains. In June 2025, AMD unveiled its new Instinct MI350 Series GPUs and open rack-scale AI infrastructure, showcasing significant AI performance and energy efficiency advancements alongside major industry partners. Further strengthening its GPU ecosystem in the first quarter of 2025, AMD enhanced its support for frontier AI models on Instinct GPUs with ROCm software, offering day-zero support for the latest Meta AI Llama 4 and Google Gemma 3 models. In the reported quarter, Core42, G42's digital infrastructure arm, began large-scale deployment of AMD Instinct GPUs. This move powers one of France's most advanced AI compute facilities. Advanced Micro Devices Faces Stiff Competition AMD faces stiff competition in the GPU market from the likes of NVIDIA NVDA and Intel Corporation INTC. NVIDIA is benefiting from the strong growth of artificial intelligence (AI), high-performance and accelerated computing. The growing demand for generative AI and large language models, which utilize GPUs based on NVIDIA's Hopper and Blackwell architectures, is driving data center revenues. In the first quarter of fiscal 2026, Data Center revenues jumped 73.3% year over year and 9.9% from the previous quarter to $39.1 billion. Intel is actively expanding its presence in the GPU market with the introduction of new Arc Pro B-series GPUs for professionals and workstations, particularly for AI inference and demanding workloads. With features such as high memory capacity, multi-GPU scalability, and extensive software compatibility, Intel aims to support creators and AI developers with powerful, reliable solutions. AMD's Share Price Performance, Valuation and Estimates Advanced Micro Device shares have gained 18.9% year to date, outperforming the broader Zacks Computer & Technology sector's return of 4.4%. However, the company has underperformed the Zacks Computer - Integrated Systems industry's increase of 28.1%. AMD Stock Performance Image Source: Zacks Investment Research AMD stock is trading at a premium, with a forward 12-month Price/Sales of 6.78X compared with the industry's 3.92X. AMD has a Value Score of F. AMD Valuation The Zacks Consensus Estimate for second-quarter 2025 earnings is currently pegged at 54 cents per share, representing an 8.4% decline over the past 30 days. This indicates a 21.74% decrease year over year. The consensus mark for 2025 earnings is pegged at $3.92 per share, which has declined 2.4% over the past 30 days, suggesting 18.43% year-over-year growth. AMD currently carries a Zacks Rank #3 (Hold). You can see the complete list of today's Zacks #1 Rank (Strong Buy) stocks here. Zacks' Research Chief Picks Stock Most Likely to "At Least Double" Our experts have revealed their Top 5 recommendations with money-doubling potential – and Director of Research Sheraz Mian believes one is superior to the others. Of course, all our picks aren't winners but this one could far surpass earlier recommendations like Hims & Hers Health, which shot up +209%. See Our Top Stock to Double (Plus 4 Runners Up) >> Want the latest recommendations from Zacks Investment Research? Today, you can download 7 Best Stocks for the Next 30 Days. Click to get this free report Intel Corporation (INTC): Free Stock Analysis Report NVIDIA Corporation (NVDA): Free Stock Analysis Report
Forbes
2 days ago
- Business
- Forbes
Everyone's Talking About AI Compute—But It All Starts With Storage
Dave Friend is the cofounder and CEO of Wasabi. When you think of AI, there are probably many things that come to mind, such as how to use it, where it's headed and what powers it. The conversation typically centers around compute—A.K.A. all the CPUs and GPUs you hear about when discussing AI. While compute is critical, there is a significant aspect of AI that is often overlooked: data. Although it may not be widely discussed, the reality is that these massive, unstructured and ever-growing data sets are what are truly driving global AI growth. As AI models become larger and more sophisticated, accessing the necessary data to train them is becoming a significant challenge for users. This is due to multiple factors, including the ever-increasing amount of data needed to train AI models. To make matters worse, hyperscaler storage that many rely on is expensive, overly complex and not optimized for the accessibility and performance that AI workflows demand. Additionally, enterprise data used to train AI systems is becoming a favored target for malicious actors. All of these factors combine to make AI adoption incredibly challenging, expensive and time-consuming. The reality is that most companies aren't struggling with AI compute limitations. They're hitting walls because they can't store, access and manage the data quickly, securely and affordably enough to support real-time inference, fine-tuning or long-term retention. If AI needs to run efficiently and cost-effectively, so does the data it learns from. To address these growing problems and fully leverage the benefits AI has to offer, organizations should implement a scalable cloud storage solution that provides cost-effectiveness, security and hybrid capabilities. Best Practices And What Leaders Should Expect However, not all data storage providers are created equal. The cloud giants that dominate the industry charge exorbitant fees to access data, making it more difficult, expensive and time-consuming for users. This makes training AI and storing the data that AI gleans a costly and challenging undertaking. To address this, organizations should seek out affordable cloud storage providers that don't charge these fees. This will enable them to easily access their data in a way that makes AI training as seamless and cost-effective as these storage buckets can be easily scaled up and down depending on need. This is ideal for training AI, as the storage will need to hold both the data required to train AI models and store the resulting information. Being able to scale up easily and down will ensure that an organization is adequately prepared for AI models and can adjust as needed. Just as important as where you store the data is ensuring it is stored securely. Cyberattackers are increasingly going after enterprise data due to its vital role in AI operations. As a result, it is crucial for IT leaders to ensure that the storage solutions they choose are adequately protecting their data. When selecting a provider, organizations should remember to look for one that offers robust data protection offerings that ensure the storage sets are impenetrable. Organizations should also take notice that the data is hidden from bad actors in the event of a breach to prevent deletion or ransomware threats. These are critical for avoiding an attack and protecting critical enterprise data. Key Features And Approaches To Security An essential part of a secure data management program is immutable backups, which prevent a malicious actor from modifying or deleting the stored data. Immutable backups are an air-gapped solution that isolates data from potential threats such as ransomware or accidental deletion. IT leaders should consider immutable backups to ensure their data is impenetrable and cannot be encrypted or deleted. Additionally, no secure cloud management program would be complete without employee training on cyber protection. By regularly updating cybersecurity best practices for employees and providing training, organizations can effectively prevent malicious actors from breaching their networks and accessing critical data. An approach that combines cost effectiveness and security is hybrid storage, which involves storing data in different methods and locations. This can include one copy in the cloud, one on-premises, and one on a hard drive. Incorporating cost-effective solutions like the cloud reduces expenses, while having the data in multiple locations allows it to be readily available in case of a cyberattack. For AI training, data can be readily available in the cloud, but it can also be stored on-premises for added security. While it is easy to get caught up in the AI boom, organizations must take their time incorporating the emerging feature. Technology decision-makers should ensure they prioritize cost-effective and secure ways to store the data necessary for proper AI training. Without it, they may be left behind their competitors in the AI adoption race. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Skift
3 days ago
- Business
- Skift
Agoda CEO on Why Price Is Still the Best Loyalty Program
Agoda's playbook is clear: move fast, stay cheap, and think local. They're not betting on the next big trend. They're betting on knowing what their customers and partners want, before anyone else. While many travel companies are busy reinventing loyalty programs with layers of points and perks, Agoda CEO Omri Morgenshtern believes the best loyalty program is simple: just keep prices low. 'Our belief is that the best loyalty is price,' Morgenshtern said. That is why Agoda has intentionally kept its loyalty program, a basic tiered system with clearly visible discounts, simple. 'A lot of those loyalty programs for me as a consumer are confusing,' he told Skift. 'We believe in two principles. If I can keep you engaged and believing that we will give you the best price in the market, I'm doing well.' The strategy has worked so far. But it's not easy. Finding the lowest prices at scale, across millions of listings in dozens of countries, is a hard operational and financial challenge. Using AI to be a Trusted Travel Guide Agoda has long leaned into artificial intelligence. The company has been using machine learning behind the scenes for over a decade, long before
