Though not technically officially confirmed yet, it looks like a remake of The Elder Scrolls 4: Oblivion is nonetheless set to launch imminently, with all rumours pointing to a Virtuos-developed Unreal Engine 5 recreation of the 2006 classic being simultaneously revealed and released at some point in the final days of the month. Given the stature of both The Elder Scrolls as a franchise and of Oblivion as a game, there’s obviously palpable excitement surrounding the remake, even in the absence of an official announcement, and you can bet we count ourselves in that excited group as well. In particular, it’s the prospect of an enhanced and improved version of the open world action RPG masterpiece that has turned heads. The Elder Scrolls 4: Oblivion has admittedly aged poorly in some regards, even though it overall remains a great game to this day, so the idea of a remake that sands out those rough edges and helps bring the entire game up the modern standards while not deviating from its biggest core strengths is an exciting one, to say the very least. Here, we’re going to talk about a few such improvements that we’re hoping to see in the purported remake. VISUALS AND TECH Let’s start things off with the obvious pick. Any remake of a 20 year old title is going to make visual and technical improvements, and what we have already seen and heard of the Oblivion remake suggests that that is very much the case here as well. The question, of course, is just how significant of an uplift the game will have received in this area, because arguably, this is where Oblivion is at its weakest. That’s not just true in terms of what the passage of time has done to the original title, but also in the sense that, really, Oblivion was a pretty janky and rough-around-the-edges experience even back when it first came out (as is so often the case with Bethesda Game Studios titles). By all accounts, it sounds like the allegedly upcoming remake has made some major visual changes, at the very least, which is pretty much exactly what most would have hoped for. From dated visuals (with excessive bloom effects, which we couldn’t not mention) to, in typical Bethesda Game Studios fashion, a laundry list of bugs and glitches (including some outright game-breaking ones), there is a lot that holds the original title back in the technical department, so hopefully that’s something that is addressed by the remake to the fullest extent possible. Of course, it would probably be silly to expect a cutting edge visual masterpiece even by 2025 standards, but we’d be surprised if the remake didn’t tout noticeable upgrades nonetheless. SEAMLESS MAP Speaking of technical upgrades that Oblivion needs to make, a seamless map would go down rather well with the Elder Scrolls fanbase (to put it mildly), though it does seem like a somewhat unrealistic dream, simply given the history of the developer. Back when it launched in 2006, Cyrodil not being a completely seamless open world wasn’t that big of a deal, because totally seamless maps were something of a rarity at the time. The fact that the map was broken up into multiple separate instances that had been stitched together with loading screens was, as such, not something that stood out to anyone in any meaningful way. That is conclusively no longer the case, of course, and hasn’t been for years at this point. Seamless open worlds are very much the norm- and yet, Bethesda Game Studios RPGs haven’t got that memo, it seems, because they continue to do things their way (frustratingly enough). This supposed Oblivion remake, however, isn’t being made by Bethesda Game Studios. Allegedly, Virtuos has remade the title in Unreal Engine 5 while bringing over a lot of the original assets, and while it remains to be seen how much of that is accurate, if it is accurate, that may well have been the perfect opportunity to finally make that change and turn Oblivion’s map into what it always should have been. BETTER DUNGEONS There’s more in Oblivion’s open world that a potential remake could try and improve with smart improvements, beyond attempting to stitch it together into a single, contiguous whole. Take, for instance, the original game’s dungeons, which were probably among its biggest weaknesses. From the dungeons being procedurally generated to even the Oblivion Gates randomly selected between a limited number of layout templates that quickly became repetitive, there were some pretty notable issues with the game in this department. Now, we are obviously not expecting the Oblivion remake to completely redo the dungeons from head to toe – we’re actually not even sure if it is going to be that significant of an overhaul – but at the very least, targeted improvements are in order. Just as an example, getting rid of procedural generation and instead delivering actual, designed dungeons would help make exploration significantly more compelling. It’d be a real shame to see that opportunity passed up. IMPROVED COMBAT Combat has never really been Bethesda Game Studios’ strong suit, to say the least, which stands doubly true for The Elder Scrolls games in particular. Typically, Oblivion’s combat was something of a jankfest- floaty movements and animations, shoddy hit detection, spotty AI, and more issues combined to make the whole affair feel disappointingly weightless and lackluster. Ideally, that should be one of the first things any hypothetical remake looks at when considering which areas of the original game are most in need of improvement. Thankfully, it does seem like rumured developer Virtuos is making some much needed changes. If reports are to be believed, the Oblivion remake will feature a new stamina system that will take cues from the Souls series, as well as overhauled blocking and archery mechanics, among other things. Our hope is that those are actual, major improvements, because put together, such changes could have a big impact on how the combat fares on a moment to moment basis. No, we don’t expect the combat to suddenly turn into one of the game’s highlights, but it would be disappointing to see it brought over as is, flaws and all. FINETUNED PROGRESSION The Elder Scrolls 4: Oblivion is counted to this day as one of the finest RPGs ever made, and a game truly deserving of that moniker (as Oblivion obviously is) is obviously not going to have any major deficiencies where its progression and customization systems are concerned. There is still room for improvement, however, and we fully expect to see some tweaking here and there, should the remake actually prove to be real. Specifically, it’s likely that the remake will make it less easy for players to break the progression curve in the manner that the original allowed players to, especially if you had spent enough time paying the game. We don’t expect – or hell, even want – that side of the experience to completely go away – there’s something undeniably appealing about RPGs that let you do that, after all – but this remake is a great chance for Virtuos and Bethesda to root out the more prominent balancing issues that do hold the progression systems back, and we’d fully expect them to take that chance.

Reinforcement learning (RL) is a powerful technique for enhancing the reasoning capabilities of LLMs, enabling them to develop and refine long Chain-of-Thought (CoT). Models like OpenAI o1 and DeepSeek R1 have shown great performance in text-based reasoning tasks, however, they face limitations on tasks that require precise numerical calculations or symbolic manipulations, such as geometric reasoning, complex computations, or equation solving. Recent research has explored prompting and supervised fine-tuning methods to equip LLMs with tool-use capabilities, but they are constrained by their reliance on imitating curated data distributions. This often results in poor generalization beyond seen patterns and an inability to determine when and how to invoke external tools. Recent advancements in LLMs show progress toward human-like metacognition through CoT prompting. Research has evolved from train-time scaling to test-time scaling, allocating additional computational resources during inference to generate intermediate reasoning steps. Techniques like stepwise preference optimization, Monte Carlo Tree Search, and RL have improved multi-step mathematical reasoning, as evidenced by models like OpenAI-o1 and DeepSeek-R1. In addition to CoT, Program-of-Thought reasoning integrates external computational tools such as Python interpreters to simplify complex reasoning steps. Further, Tool-integrated reasoning was initially introduced to help LLMs solve computationally intensive problems through programming strategies. Researchers from ByteDance Seed have proposed ReTool, a CI-powered RL framework designed to address math problem-solving tasks. It enhances long-form reasoning with tool-integrated learning through two key features. First, it enables dynamic interleaving of real-time code execution within natural language reasoning processes. Second, it implements an automated RL technique that allows policy rollouts with multi-turn real-time code execution, teaching the model when and how to invoke tools based on outcome feedback. ReTool employs a systematic training framework that begins with synthetic cold-start data generation to produce code-augmented long-form reasoning traces for fine-tuning base models. The ReTool consists of two primary stages, cold-start supervised fine-tuning followed by RL with interleaved code execution rollout. The pipeline designed for collecting and curating high-quality data begins with collecting high-quality mathematical reasoning data from diverse sources, including open-source datasets like OpenThoughts. A dual-verification approach combining human expert curation and Deepseek-R1 evaluation filters invalid data. From this foundation, code-integrated reasoning data is automatically constructed. The VeRL framework is employed with PPO as the RL method for training. The maximum sequence length is set to 16384 tokens, with a 512 mini-batch size and a KL coefficient of 0.0, using Qwen2.5-32B-Instruct as the main backbone. ReTool enables the LLM to utilize the code interpreter flexibly during the RL stage, leading to substantial performance improvements. ReTool (Qwen2.5-32B-Instruct) achieves accuracies of 67.0% on AIME2024 and 49.3% on AIME2025 with only 400 training steps. This outperforms the text-based RL baseline (Qwen2.5-32B-Instruct), which attains 40.0% and 36.7% on the respective benchmarks despite using over 1000 training steps. Moreover, on AIME2024, ReTool (Qwen2.5-32B-Instruct) surpasses the competitive baseline s1-32B by 10.3%. Similarly, on AIME2025, it achieves an 11.4% gain over OpenAI’s o1-preview. When combined with a more advanced backbone, ReTool (DeepSeek-R1-Distill-Qwen-32B) further improves performance with scores of 72.5% on AIME2024 and 54.3% on AIME2025. In conclusion, researchers introduced ReTool, a novel RL framework that empowers LLMs to self-enhance their mathematical reasoning capabilities through effective Code Interpreter utilization. Experiments on AIME2024 and AIME2025 show that ReTool achieves superior accuracy compared to conventional text-based RL approaches and converges with significantly fewer training steps. Through careful data curation and a specialized tool-using pipeline, ReTool enables models to develop complex computational intervention strategies, paving the way for more efficient and powerful tool-augmented reasoning in LLMs. The results demonstrate that tool-integrated RL represents a promising direction for advancing mathematical reasoning capabilities in LLMs for tasks requiring precise computation and symbolic manipulation. Check out the Paper. Also, don’t forget to follow us on Twitter and join our Telegram Channel and LinkedIn Group. Don’t Forget to join our 90k+ ML SubReddit. Sajjad AnsariSajjad Ansari is a final year undergraduate from IIT Kharagpur. As a Tech enthusiast, he delves into the practical applications of AI with a focus on understanding the impact of AI technologies and their real-world implications. He aims to articulate complex AI concepts in a clear and accessible manner.Sajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Fourier Neural Operators Just Got a Turbo Boost: Researchers from UC Riverside Introduce TurboFNO, a Fully Fused FFT-GEMM-iFFT Kernel Achieving Up to 150% Speedup over PyTorchSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Model Compression Without Compromise: Loop-Residual Neural Networks Show Comparable Results to Larger GPT-2 Variants Using Iterative RefinementSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/Underdamped Diffusion Samplers Outperform Traditional Methods: Researchers from Karlsruhe Institute of Technology, NVIDIA, and Zuse Institute Berlin Introduce a New Framework for Efficient Sampling from Complex Distributions with Degenerate NoiseSajjad Ansarihttps://www.marktechpost.com/author/sajjadansari/NVIDIA AI Releases UltraLong-8B: A Series of Ultra-Long Context Language Models Designed to Process Extensive Sequences of Text (up to 1M, 2M, and 4M tokens)

The following article contains spoilers for The Last of Us season 2, episode 2. If the first episode of The Last of Us season 2 was the scene setter, then episode 2 is the catalyst for everything that comes next. It features – SPOILERS! – the death of Joel Miller, and it takes place in pretty much the exact same way as it did in the game. The devil is in the details, though, and not everything about that scene is a 1:1 recreation. That’s where our TV show vs game comparison comes in. We’ve taken the major scenes of episode 2 that are taken from the game and compared them against the original source material, analysing what’s changed and what’s stayed the same. You can see both versions in the video above, or read on below for our written explanations. Joel’s DeathAs the instigating incident of the game and the big main event of the season’s second episode, Joel’s death was naturally going to be a meticulously recreated sequence. Both the broad strokes and many of the small details are all here; Abby blasts his leg with a shotgun and then proceeds to lay into him with a golf club. Kaityn Dever delivers the “You don’t get to rush this” line exactly as her game counterpart did. When Ellie arrives on the scene, the direction largely follows in the footsteps of the original cutscene, using the same floor-level camera angles and high-pitched sound break as Abby deals the final blow. Lurking between the many game-accurate details are several changes, though. First and foremost is that Joel is with Dina for this scene, not Tommy. Furthermore, Dina is drugged for the whole event, which means Ellie is the only Jackson resident to witness Joel’s murder. Abby’s behaviour is also slightly different; in the show she reveals to Joel who she is and why she’s about to kill him. In the game there’s no such scene, and we’re left to believe that Joel died without truly knowing why Abby hated him so much.Oh, and then there’s the matter of Abby’s hole in… well, not quite one. The game depicts the blow that finally kills Joel as a horrible bit of blunt force trauma to the skull, using the actual head of the golf club. In the show, the club is broken in Abby’s assault, and so all that’s left is the sharp broken end. Abby uses this as a shiv, stabbing it into Joel’s neck. Abby’s FlashbackEpisode two features a sequence in which a young Abby searches the Fireflies’ hospital for her dad. It’s a recreation of the end of Tracking Lessons, the chapter of the game where the perspective shifts from Ellie to Abby. While the setting of this scene remains the same – the hospital corridor bathed in red emergency lighting – the actual events are rather different. In the show, older Abby confronts her younger self, cementing that this is a dream sequence rather than a flashback. Abby tells herself not to look inside the operating room, as she’ll have to see her dead father. Younger Abby does go into the operating room, but the camera does not follow, and so we’re left with just the older Abby’s restrained tears to relay the horror found inside. In the game, since you are in control of Abby, you get to see the inside of the room and Abby’s reaction to finding her father.This reframing of the scene is likely due to its shifted position; in the game, this is the moment you learn that Abby’s father was killed in Joel’s attack on the Fireflies, and so the raw emotion of seeing Abby cry on the operating room flaw is necessary to humanise a character who has been portrayed as a villain for the prior 10 or so hours. The show reveals Abby’s motive from the very start and moves this sequence to much earlier in the story, and so it serves a different purpose. Ellie’s Awkward MorningThe most faithful scene recreation of episode two arrives when Jesse comes knocking at Ellie’s door to go out on patrol. What follows is an awkward exchange regarding Ellie and Dina’s kiss the previous night at the barn dance – Dina, of course, had only recently split up with Jesse. The dialogue here is a 95% match to the game’s script, and the camera work also does its best to recreate the conversation in exact detail. The main difference here is context. In the game, the barn dance isn’t shown until right near the end of the campaign, and so when playing this sequence you’re using the information provided by the characters to piece together an event you’ve not witnessed. In the show, you know exactly what Jesse and Ellie are talking about, as you watched it happen in the previous episode.Bigot SandwichesSimilar to Ellie’s conversation with Jesse, her encounter with Seth the morning after the dance is largely a 1:1 replication of the same scene in the game. Seth has once again prepared steak sandwiches as an apology, and much of the dialogue around this awkward exchange is taken straight from the game’s script. The main difference here is that Jesse is now part of the scene, and he thanks Seth for the sandwiches rather than Maria. Additionally, the building itself is visually very different to that in the game, looking more like a canteen than a timber-constructed bar. Eugene’s Weed FarmThe show sees Jesse and Ellie head out on a patrol that recreates the middle section of the game’s first chapter. Fans of the game will instantly notice a key difference, as Ellie’s discovery of Eugene’s weed farm actually takes place during the same patrol depicted in episode one where Ellie and Dina explore the supermarket. These events have been split up and changed, as it’s now Jesse, not Dina who accompanies Ellie. That means the show removes the sequence in which Ellie and Dina smoke weed and are implied to have sex.Despite this, there are still several key elements of the sequence that are kept intact. Ellie still discovers Eugene’s Firefly pendant, as well as his once-impressive marijuana operation (which is far less well hidden in the show.) Among the belongings scattered about, Ellie finds Eugene’s bong gas mask, which can also be found in the game. Jesse, however, is much less impressed with Eugene’s ingenuity than Dina was. Abby’s EscapeAbby’s fateful encounter with the infected horde plays out much like it does in the game’s first chapter. A chase sequence results in Abby becoming trapped behind a chainlink fence that begins to collapse under the weight of the clawing runners. If anything, this sequence is even nastier than it was in the game thanks to a shot of a hand being pushed through the fence, the wire cutting through the flesh. Aside from that, the broad direction of this scene is very close to the framing of the game, right up to the way Joel’s revolver appears from the side of the shot to blast the infected that pins Abby to the ground. For more from The Last of Us, check out our spoiler-free season two review and our spoiler-filled review of the second episode. We’ve also asked the show’s creators about how canon can change, and what that means for the show's biggest plot points.Matt Purslow is IGN's Senior Features Editor.

Now that WrestleMania 41 is over, you can watch it all again, in bits and pieces or start to finish.

SHARM EL SHEIKH, EGYPT - NOVEMBER 09: Former U.S. Vice President Al Gore (C) and Gavin McCormick, ... More CEO of WattTime, present the new Climate TRACE platform, a highly detailed facility-level global inventory of greenhouse gas emissions, at the UNFCCC COP27 climate conference on November 09, 2022 in Sharm El Sheikh, Egypt. Climate TRACE combines open-source data and imagery from multiple satellite systems with artificial intelligence to determine actual emissions of greenhouse gases, including CO2 and methane, from sites across the globe. According to the data the single highest emitter is the oil-producing Permian Basin in Texas. The COP27 conference is bringing together political leaders and representatives from 190 countries to discuss climate-related topics including climate change adaptation, climate finance, decarbonisation, agriculture and biodiversity. The conference is running from November 6-18. (Photo by Sean Gallup/Getty Images)Getty Images Last month, something revolutionary quietly took place. On March 28, 2025, the world received its first-ever monthly Climate Data update on global greenhouse gas emissions—derived not from self-reported pledges or slow-moving government reports, but from direct, verified observation. Using satellites, sensors, and artificial intelligence, Climate TRACE delivered precise emissions data with just a 60-day lag. This groundbreaking achievement comes courtesy of Climate TRACE, a coalition backed by former U.S. Vice President Al Gore. At the COP29 Climate Summit in Baku, Azerbaijan, Gore outlined a transformative vision: give climate data the urgency and transparency that financial markets take for granted. “You can only manage what you can measure.” A Market Signal for the Planet This is more than data—it’s the birth of a global climate dashboard, something akin to a Bloomberg Terminal for Earth’s health. Investors, regulators, insurers, and policymakers now have regular insight into exactly who is polluting, how much, and where. The era of vague commitments and unchecked greenwashing just got a powerful new opponent. Global monthly GHG emissions (Jan 2015-Jan 2025)Climate TRACE Climate TRACE’s January 2025 update shows global emissions totaled 5.26 billion tonnes CO₂e, a modest yet significant 0.59% drop compared to January 2024. Methane, an especially potent greenhouse gas, remained stable at roughly 32.24 million tonnes, hinting that the global emissions curve might finally be starting to flatten. From Static Reports to Live Dashboards with Climate Data Historically, global emissions data has suffered from delays of one to two years—far too late for effective action. Managing climate policy with outdated data is akin to managing a financial portfolio using last year's market prices. Investors wouldn’t tolerate that. Why should we accept it for the planet? Climate TRACE is not an organization—it's a coalition of more than 100 collaborating organizations. Gavin McCormick, Co-founder, Climate TRACE on stage from COP29Wedonthavetime.org Climate TRACE transforms this outdated approach. Its advanced network monitors over 660 million emission sources worldwide, including power plants, factories, farms, and shipping vessels. AI-powered algorithms analyze satellite heat signatures, spectral imagery, and operational data, providing accurate monthly updates. The Climate TRACE coalition uses cross-referenced satellite imagery, machine learning algorithms, and independent validation with ground-based data sources to ensure accuracy. While not perfect, the methodology represents a major leap beyond self-reported inventories. This emissions inventory’s accuracy is akin to quarterly financial reports, transforming climate tracking from guesswork into rigorous accounting.“Worldwide, we’re seeing the rapid rise in clean energy jobs overtake dirty energy jobs, and we’re seeing people all around the world continue to demand that we respect their future. And that’s why we are so excited to share this data from Climate TRACE with you.” Sector-by-Sector InsightsGlobal GHG emissions by sector (Jan 2025)Climate TRACE The January 2025 breakdown reveals where emissions are beginning to bend and where momentum is still missing. Greenhouse gas emissions increased year over year in waste and manufacturing and decreased in transportation, power, and fossil fuel operations. Transportation saw the greatest change in emissions year over year, with emissions decreasing by 1.6% Nations and Cities in Real-Time Competition The data doesn’t just show sectors—it also reveals national performance, turning global emissions into a monthly scoreboard that clearly tracks year-over-year changes (as shown below, comparing January 2025 with January 2024; China: -1.1% (a reduction of 17.4 million tonnes CO₂e) USA: -0.28% India: -0.03% Russia: -0.18% EU bloc: -0.53% China’s absolute reduction stands out as particularly notable. At the city level, Dortmund (Germany) and Pohang-si (South Korea) led significant cuts, while emissions in Ma'anshan and Anshan (China) rose notably. Just as financial markets watch national GDP, inflation, and employment closely, we can now track emissions monthly, providing continuous incentives for improvement. Financial Markets Thrive on Timely Data—Now Climate Action Can Too Financial markets rely on speed, transparency, and precision. Quarterly reports delayed by years would be absurd. Now, climate policy can finally mirror this urgency. Real-time emissions data allows investors to evaluate companies based on actual performance, not public relations. Regulators can pinpoint precise interventions, and cities can compete on measurable impacts, not vague promises. A Planetary MRI Climate TRACE’s latest dataset refines tracking across key sectors, adds monitoring for 918 shipping ports, and distinguishes fossil from biogenic methane using IPCC guidelines. Moreover, the entire dataset is open-source, inviting journalists, researchers, policymakers, and citizens alike to hold polluters accountable. A New Era for Climate Data—But Will It Be Allowed?Jay Inslee, Governor of Washington state, speaks to the media beside during a tour of the Los ... More Angeles Department of Transportation bus depot in Los Angeles, California on May 3, 2019.AFP via Getty Images"This is a tool to show the threat to our own children. And when they see this tool in city council races and community races all across the United States, politicians are going to have to yield to this technology. Seeing is believing, that's the news." This unprecedented level of transparency promises a sea change in climate accountability—empowering real-time decision-making, sharpening risk assessments, and helping the world act with the precision this crisis demands. Climate TRACE gives actuaries, insurers, regulators, and investors something they’ve never had before: verified, independent data updated every month. As I wrote earlier this year in '50% GDP Collapse Ahead?', actuaries warn that unchecked climate risk could halve the global economy within decades. This data may be our last, best shot at changing course. But with great clarity comes great resistance The current U.S. administration has shown open hostility toward climate science. Plans to cancel the lease for Hawaii’s Mauna Loa Observatory—a cornerstone of atmospheric CO₂ monitoring for over 60 years—were confirmed by recent reporting from Reuters and the Honolulu Star-Advertiser. EPA budget proposals, criticized by over 1,900 scientists in an open letter published by The Guardian, threaten to dismantle environmental protections across the U.S. Climate TRACE’s open-source, satellite-driven data doesn’t just inform, it disrupts. It exposes the emitters. It reveals the truth behind the promises. And it turns delay into a choice, not an excuse. But not everyone welcomes this kind of radical transparency. In fact, some would rather we stayed blind. Will it be allowed to thrive? Or will vested interests—political and industrial—try to suppress it before it reshapes the system? Because make no mistake: this data is power. And in the wrong hands—or kept out of public reach—it’s a power we could lose. If this data disappears, so does our ability to act before disaster strikes. Today’s near-real-time emissions tracking relies on a constellation of satellites, sensors, and AI models—many of which are controlled by governments or private entities. If access is cut off, censored, or restricted, we lose the transparency needed to hold polluters accountable. Without it, climate policy risks sliding back into the dark ages of delay, distraction, and disaster. The international community must defend it. Not just as a breakthrough in climate science but as a safeguard for truth in a time of disinformation. Because without it, we lose our clearest signal and our strongest leverage. The world must remain vigilant to ensure this powerful new tool isn't silenced. Real-Time Climate Data Has the Power to Reshape Global Climate PolicyExplore the Climate TRACE Map Climate TRACE The availability of real-time climate data marks a decisive shift in how the world can respond to the climate crisis. No longer reliant on outdated, self-reported inventories, policymakers, investors, and regulators now have access to timely, independent emissions data—updated monthly, with sector-specific and geographic granularity. What happens next will determine whether this breakthrough accelerates global climate progress—or whether it’s quietly sidelined before it can change the system. Explore the full Climata Data dashboard at climatetrace.org.

Table of Contents Table of Contents Memory innovations headed to AI phones The interplay of storage and AI  Going beyond the RAM capacity The road to more private AI experiences?  One of the most obvious — and honestly, the dullest —trends within the smartphone industry over the past couple of years has been the incessant talk about AI experiences. Silicon warriors, in particular, often touted how their latest mobile processor would enable on-device AI processes such as video generation. We’re already there, albeit not completely. Amidst all the hype show with hit-and-miss AI tricks for smartphone users, the debate barely ever went beyond the glitzy presentations about the new processors and ever-evolving chatbots. Recommended Videos It was only when the Gemini Nano’s absence on the Google Pixel 8 raised eyebrows that the masses came to know about the critical importance of RAM capacity for AI on mobile devices. Soon, Apple also made it clear that it was keeping Apple Intelligence locked to devices with at least 8GB of RAM. Related But the “AI phone” picture is not all about the memory capacity. How well your phone can perform AI-powered tasks also depends on the invisible RAM optimizations, as well as the storage modules. And no, I’m not just talking about the capacity. Memory innovations headed to AI phones Micron / Digital Trends Digital Trends sat with Micron, a global leader in memory and storage solutions, to break down the role of RAM and storage for AI processes on smartphones. The advancements made by Micron should be on your radar the next you go shopping for a top-tier phone.  The latest from the Idaho-based company includes the G9 NAND mobile UFS 4.1 storage and 1γ (1-gamma) LPDDR5X RAM modules for flagship smartphones. So, how exactly do these memory solutions push the cause of AI on smartphones, apart from boosting the capacity?  Let’s start with the G9 NAND UFS 4.1 storage solution. The overarching promise is frugal power consumption, lower latency, and high bandwidth. The UFS 4.1 standard can reach peak sequential read and write speeds of 4100 MBps, which amounts to a 15% gain over the UFS 4.0 generation while trimming the latency numbers, too.  Another crucial benefit is that Micron’s next-gen mobile storage modules go all the way up to 2TB capacity. Moreover, Micron has managed to shrink their size, making them an ideal solution for foldable phones and next-gen slim phones such as the Samsung Galaxy S25 Edge.  Micron / Digital Trends Moving over to the RAM progress, Micron has developed what it calls 1γ LPDDR5X RAM modules. They deliver a peak speed of 9200 MT/s, can pack 30% more transistors due to size shrinking, and consume 20% lower power while at it. Micron has already served the slightly slower 1β (1-beta) RAM solution packed inside the Samsung Galaxy S25 series smartphones. The interplay of storage and AI  Ben Rivera, Director of Product Marketing in Micron’s Mobile Business Unit, tells me that Micron has made four crucial enhancements atop their latest storage solutions to ensure faster AI operations on mobile devices. They include Zoned UFS, Data Defragmentation, Pinned WriteBooster, and Intelligent Latency Tracker.  “This feature enables the processor or host to identify and isolate or “pin” a smartphone’s most frequently used data to an area of the storage device called the WriteBooster buffer (akin to a cache) to enable quick, fast access,” explains Rivera about the Pinned WriteBooster feature.  Micron / Digital Trends Every AI model – think of Google Gemini or ChatGPT — that seeks to perform on-device tasks needs its own set of instruction files that are stored locally on a mobile device. Apple Intelligence, for example, needs 7GB of storage for all its shenanigans. To perform a task, you can’t depute the entire AI package to the RAM, because it would need space for handling other critical chores such as calling or interacting with other important apps. To deal with the constraint on the Micron storage module, a memory map is created that only loads the needed AI weights from the storage and onto the RAM.  When resources get tight, what you need is a faster data swap and reading. Doing so ensures that your AI tasks are executed without affecting the speed of other important tasks. Thanks to Pinned WriteBooster, this data exchange is sped up by 30%, ensuring the AI tasks are handled without any delays. So, let’s say you need Gemini to pull up a PDF for analysis. The fast memory swap ensures that the needed AI weights are quickly shifted from the storage to the RAM module.  Next, we have Data Defrag. Think of it as a desk or almirah organizer, one that ensures that objects are neatly grouped across different categories and placed in their unique cabinets so that it’s easy to find them.  Micron / Digital Trends In the context of smartphones, as more data is saved over an extended period of usage, all of it is usually stored in a rather haphazard matter. The net impact is that when the onboard system needs access to a certain kind of files, it becomes harder to find them all, leading to slower operation.  According to Rivera, Data Defrag not only helps with orderly storage of data, but also changes the route of interaction between the storage and device controller. In doing so, it enhances the read speed of data by an impressive 60%, which naturally hastens all kinds of user-machine interactions, including AI workflows.  “This feature can help expedite AI features such as when a generative AI model, like one used to generate an image from a text prompt, is called from storage to memory, allowing data to be read faster from storage into memory,” the Micron executive told Digital Trends.  Intelligence Latency Tracker is another feature that essentially keeps an eye on lag events and factors that might be slowing down the usual pace of your phone. It subsequently helps with debugging and optimizing the phone’s performance to ensure that regular, as well as AI tasks, don’t run into speed bumps.  Micron / Digital Trends The final storage enhancement is Zoned UFS. This system ensures that data with similar I/O nature is stored in an orderly fashion. This is crucial because it makes it easier for the system to locate the necessary files, instead of wasting time rummaging through all the folders and directories.  “Micron’s ZUFS feature helps organize data so that when the system needs to locate specific data for a task, it’s a faster and smoother process,” Rivera told us.  Going beyond the RAM capacity When it comes to AI workflows, you need a certain amount of RAM. The more, the better. While Apple has set the baseline at 8GB for its Apple Intelligence stack, players in the Android ecosystem have moved to 12GB as the safe default. Why so?  “AI experiences are also extremely data-intensive and thus power-hungry. So, in order to deliver on the promise of AI, memory and storage need to deliver low latency and high performance at the utmost power efficiency,” explains Rivera.  With its next-gen 1γ (1-gamma) LPDDR5X RAM solution for smartphones, Micron has managed to reduce the operational voltage of the memory modules. Then there’s the all-too-important question of local performance. Rivera says the new memory modules can hum at up to 9.6 gigabits per second, ensuring top-notch AI performance.  Micron / Digital Trends Micron says improvements in the Extreme ultraviolet (EUV) lithography process have opened the doors for not only higher speeds, but also a healthy 20% jump in energy efficiency.  The road to more private AI experiences?  Microns’s next-gen RAM and storage solutions for smartphones are targeted not just at improving the AI performance, but also the general pace of your day-to-day smartphone chores. I was curious whether the G9 NAND mobile UFS 4.1 storage and 1γ (1-gamma) LPDDR5X RAM enhancements would also speed up the offline AI processors.  Smartphone makers as well as AI labs are increasingly shifting towards local AI processing. That means instead of sending your queries to a cloud server where the operation is handled, and then the result is sent to your phone using an internet connection, the entire workflow is executed locally on your phone. Nadeem Sarwar / Digital Trends From transcribing calls and voice notes to processing your complex research material in PDF files, everything happens on your phone, and no personal data ever leaves your device. It’s a safer approach that is also faster, but at the same time, it requires beefy system resources. A faster and more efficient memory module is one of those prerequisites.  Can Micron’s next-gen solutions help with local AI processing? It can. In fact, it will also speed up processes that require a cloud connection, such as generating videos using Google’s Veo model, which still require powerful compute servers. “A native AI app running directly on the device would have the most data traffic since not only is it reading user data from the storage device, it’s also conducting AI inferencing on the device. In this case, our features would help optimize data flow for both,” Rivera tells me.  So, how soon can you expect phones equipped with the latest Micron solutions to land on the shelves? Rivera says all major smartphone manufacturers will adopt Micron’s next-gen RAM and storage modules. As far as market arrival goes, “flagship models launching in late 2025 or early 2026” should be on your shopping radar. Editors’ Recommendations

On Sunday, a 737 Max plane meant for a Chinese airline was returned to Boeing's production hub in the US. Dan Catchpole/Reuters 2025-04-21T06:12:47Z Save Saved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now. Have an account? A Boeing jet meant for Xiamen Airlines returned to Boeing's US production hub on Sunday. Another is en route from China to the US territory of Guam. Boeing has said China is a "significant market" and that trade restrictions would hurt its market share. A Boeing jet made for China's Xiamen Airlines was returned to Boeing's American production hub as the American aviation giant finds itself caught up in the trade war.According to online flight records, a 737 Max flew from Zhoushan and landed in Seattle on Sunday night local time.Reuters first reported the plane landing, saying that it was among several 737 Max jets that had been waiting at Boeing's Zhoushan completion center for final work and delivery to Chinese airlines.The Puget Sound Business Journal reported in 2020 that Xiamen Airlines cut its flights to Seattle in 2019. Previously, it flew Boeing 787s on the route.It is unclear why the plane was returned to the US and whether more aircraft bound for Chinese airlines will be sent back.Data from AirNav Radar showed another Boeing 737 Max from Zhoushan heading to Guam on Monday morning local time — a frequent stop for planes traveling back to the US.Boeing and Xiamen Airlines did not immediately respond to requests for comment from Business Insider.Last week, Bloomberg reported, citing anonymous sources, that China has ordered its airlines to stop taking deliveries of Boeing planes and American aircraft parts, as a new measure against its trade war with the US. After the Bloomberg story, President Donald Trump said on Truth SocialXiamen Airlines is a subsidiary of China Southern Airlines, a state-owned carrier that is among the country's "Big Three" airlines.On April 11, China Southern Airlines stopped the sale of 10 of its used Boeing 787-8 Dreamliner planes, per a filing with Shanghai United Assets and Equity Exchange. China Southern had planned to replace its Dreamliners with bigger and newer planes, but it reversed its decision. The state of the trade warThe US-China trade war escalated quickly inChina has curbed the import of Hollywood films and added tariffs to various US goods. The US put restrictions on Nvidia selling its H20 chips to China, a move analysts said is essentially a ban on exporting those chips.China has a 125% duty on American goods. The White House said Chinese products face a levy of up to 245%.China is an important market for Boeing, which has been recovering after a string of financial and reputational losses last year. Any imposition on deliveries threatens its market share as it competes with Europe's Airbus and newer entrants from China.In its 2024 annual report filed in February, Boeing called China a "significant market" that would be affected by "deterioration in geopolitical or trade relations." Boeing did not break down the company's revenue by region.Boeing is among the US's 100 most valuable companies and employed 172,000 people as of December. Its stock is down 8.5% this year. Recommended video

J Office / RVMNSave this picture!© Yongjoon Choi Architects: RVMN Area Area of this architecture project Area:  176 m² Year Completion year of this architecture project Year:  2025 Photographs Photographs:Yongjoon Choi Lead Architects: Yonghyun Kwon , Hyoju Kim More SpecsLess Specs Save this picture! Text description provided by the architects. Where do modern people spend most of their day? Most likely, it's their workplace—the space where they work. It may well be the most dynamic and productive environment in a person's daily life. In this project, we delve into the unique rhythm of the 'J' group—its corporate values, work culture, the working styles of its members, and the time they dedicate to their tasks. This company embraces honesty and practicality, rejecting pretense and unnecessary formality. Its members work closely with one another, yet remain focused on their individual tasks, each maintaining a subtle but respectful boundary.Save this picture!We focused on this delicate sense of distance. Just like how trees grow side by side, reaching their branches without overlapping, they form natural boundaries while coexisting. In such open spaces, sunlight reaches the ground, and the trees grow alongside each other without intrusion. Inspired by this considerate behavior in nature, we interpreted the space through the concept of Crown Shyness.Save this picture!The layout is simple: work areas, meeting rooms, an executive office, a lounge, and a pantry. Rather than simply immersing in work and letting the hours pass unnoticed, we wanted people to work while feeling the warmth of sunlight and the passing of time. We placed the most frequently occupied spaces near windows and designed meeting rooms, office automation (OA) areas, the lounge, and the pantry in line with natural movement flows and rhythms. The executive office and meeting rooms, located on the windowless side, form core masses in the layout. Like trees in a forest, each with its own unique presence, these masses differ in height and finish. The light filtering in between these volumes through the louver ceiling spreads across the space, just like sunlight streaming through gaps between trees.Save this picture!Save this picture!The lounge functions both as a waiting area for visitors and a resting spot for staff. While unified as a single mass, this space exudes warmth and hospitality, like a tree fully bathed in sunlight. More than just a break room, it offers a brief pause and a chance for members to restore their rhythm—a vital shelter within the office. Beyond the lounge lies the main work area, kept as open as possible to create a spacious, shared atmosphere. Even in an open layout, we used low partitions to create gentle boundaries, allowing both concentration and collaboration to coexist. A space where everyone is together, yet each person can focus—that's our interpretation of Crown Shyness in action.Save this picture!Looking up in a forest, light filters softly through the spaces between trees. To recreate this feeling, the ceiling is left open using louvers instead of solid panels, with lighting placed between structural elements to mimic sunlight streaming through foliage. In the work area, we used perforated metal mesh to create semi-transparent surfaces, adding handmade pendant lights that brighten desks and bring a sense of playfulness to the space.Save this picture!The overall tone and manner of the interior blend white, oak, and walnut-toned HPL, combined with stone-textured tiles and wood flooring that evoke the feeling of walking outdoors. Given the office's location in the high-rise jungle of Gangnam, we incorporated metal finishes to add a cool, urban touch, creating a space that feels both warm and metropolitan.Save this picture!The meeting rooms and executive office carry the same massive design language, but are differentiated according to their function. Meeting rooms, where many practical conversations take place, are finished in oak to create a warm atmosphere, and satin glass lets the silhouette of the interior remain visible from the outside, encouraging transparency and communication. The executive office, by contrast, is finished in toned-down ash and walnut HPL with wood flooring, creating a natural yet more weighted atmosphere distinct from the rest of the office. Within the same space, different scenes unfold.Save this picture!This is not merely a place for work. Like the order of the forest, where each tree grows without disturbing another, this space respects individual rhythms and boundaries, allowing for natural coexistence and growth. The quiet consideration embedded in the concept of Crown Shyness becomes the true heart of the space.Save this picture! Project gallerySee allShow less Project locationAddress:Gangnam-gu, Seoul, South KoreaLocation to be used only as a reference. It could indicate city/country but not exact address.About this officeRVMNOffice••• MaterialSteelMaterials and TagsPublished on April 21, 2025Cite: "J Office / RVMN" 21 Apr 2025. ArchDaily. Accessed . <https://www.archdaily.com/1029220/j-office-rvmn&gt ISSN 0719-8884Save世界上最受欢迎的建筑网站现已推出你的母语版本!想浏览ArchDaily中国吗?是否 You've started following your first account!Did you know?You'll now receive updates based on what you follow! Personalize your stream and start following your favorite authors, offices and users.Go to my stream

Nature, Published online: 16 April 2025; doi:10.1038/d41586-025-00839-0The World Health Organization accord promotes sharing scientific data in exchange for more-equitable distribution of drugs and vaccines.

Character Modeler @RainMichi unveiled a stylized 3D recreation of Teen Titans' Starfire made in ZBrush, Maya, and Blender.See more: https://80.lv/articles/3d-artist-recreates-teen-titans-starfire-in-maya-zbrush-blender/