Neil Druckmann of Naughty Dog announced that Sony is working on a location-based physical entertainment experience based on The Last of Us.Read More

If you were looking for your own personal AI supercomputer, Nvidia has you covered.The chipmaker announced at CES its launching a personal AI supercomputer called Project Digits in May. The heart of Project Digits is the new GB10 Grace Blackwell Superchip, which packs enough processing power to run sophisticated AI models while being compact enough to fit on a desk and run from a standard power outlet (this kind of processing power used to require much larger, more power-hungry systems). This desktop-sized system can handle AI models with up to 200 billion parameters, and has a starting price of $3,000. The product itself looks a lot like a Mac Mini.AI will be mainstream in every application for every industry. With Project Digits, the Grace Blackwell Superchip comes to millions of developers, Nvidia CEO Jensen Huang said in a press release. Placing an AI supercomputer on the desks of every data scientist, AI researcher and student empowers them to engage and shape the age of AI.Project Digits looks like a mini PC. Image: NvidiaEach Project Digits system comes equipped with 128GB of unified, coherent memory (by comparison, a good laptop might have 16GB or 32GB of RAM) and up to 4TB of NVMe storage. For even more demanding applications, two Project Digits systems can be linked together to handle models with up to 405 billion parameters (Metas best model, Llama 3.1, has 405 billion parameters).The GB10 chip delivers up to 1 petaflop of AI performance (which means it can perform 1 quadrillion AI calculations per second) at FP4 precision (which helps make the calculations faster by making approximations), and the system features Nvidias latest-generation CUDA cores and fifth-generation Tensor Cores, connected via NVLink-C2C to a Grace CPU containing 20 power-efficient Arm-based cores. MediaTek, known for their Arm-based chip designs, collaborated on the GB10s development to optimize its power efficiency and performance.The Digits supercomputer specs. Image: NvidiaUsers will also get access to Nvidias AI software library, including development kits, orchestration tools, and pre-trained models available through the Nvidia NGC catalog. The system runs on Linux-based Nvidia DGX OS and supports popular frameworks like PyTorch, Python, and Jupyter notebooks. Developers can fine-tune models using the Nvidia NeMo framework and accelerate data science workflows with Nvidia RAPIDS libraries.Users can develop and test their AI models locally on Project Digits, then deploy them to cloud services or data center infrastructure using the same Grace Blackwell architecture and Nvidia AI Enterprise software platform.Nvidia offers a range of similar devices in the same accessibility style in December, it announced a $249 version of its Jetson computer for AI applications, targeting hobbyists and startups, called the Jetson Orin Nano Super (it handles models up to 8 billion parameters).

Alienware is joining Asus, Samsung, and MSI in making a silly worlds first claim all four of them are now set to launch a genuinely exciting new wave of 27-inch 4K QD-OLED gaming monitors with an excellent 240Hz refresh rate. Remember when I called Alienwares 32-inch version the best monitor of CES last year because it finally offered the best of all worlds? Now, youll be able to buy a smaller 27-inch version, without a curved screen, in a far more subtle design that no longer dominates your desk. Like competitors that are using the same Samsung panel, the Alienware AW2725Q has a technically 26-inch screen that offers 166 pixels per inch and 250 nits of typical brightness (1,000 nits of HDR at peak), while displaying 99 percent of the DCI-P3 color gamut. Unlike some competitors, though, Alienwares G-Sync and FreeSync Premium Pro screen only offers DisplayPort 1.4 rather than DisplayPort 2.1, which could theoretically mean dealing with the occasional Display Stream Compression (DSC) hiccup to get your 4K at 240Hz. It also features a pair of HDMI 2.1 ports that offer 4K at 240 with HDR, VRR, ALLM, and eARC for Dolby Atmos passthrough. Theres no USB-C video in, but the 15-pound monitor does offer a very basic USB-A hub and a single USB-C port with 15W charging.The ports. Tap here to enhance. Image: DellThe monitor supports Dolby Vision HDR and offers a three-year burn-in warranty with a graphite sheet placed between the panel layers to help fight burn-in and image retention, plus AI-based technology that actively monitors on-screen images and makes adjustments to help prevent burn-in.For now, the best thing it has going for it over the competition is a firm price and release window: March for $900 in the US or $1,230 CAD in Canada. (Alienwares competitors for the 32-inch version took a good bit longer to hit the market, and I wonder if thatll be the same this time.) The monitor will actually hit China this month and arrive in EMEA territories in April.

Alienware has built some of the easiest-to-open toolless gaming PCs around but for years, the Dell-owned brand has stifled their upgrade potential by limiting them to Dells own proprietary power supplies and motherboards.But the 2025 Alienware Area-51, an 80-liter tower just introduced at CES in Las Vegas, finally ditches the proprietary parts in favor of standard ATX components. Even though the tempered glass sided chassis features fancy compartments for liquid cooling and power supplies, its no longer a hexagonal monster or even a proprietary tower: itll come with a standard power supply, standard based motherboard, and even feature standard fan mounting locations. And even though theres a dedicated daughterboard to easily control and cable manage its lighting, fans, I/O, and power switch, Dell will offer a conversion kit to make it work with third-party motherboards.You can see the daughterboard here. Photo by Antonio G. Di Benedetto / The VergeWith this edition, we are returning to our roots with a machine that caters to the desires of PC gaming enthusiasts and longtime Alienware fans who have a deep appreciation for technology and a can-do attitude for manually customizing their build to their needs, writes brand manager Matt McGowan, promising the ability to make serious upgrades for years to come.Alienware Area-51 key detailsBeyond the shift in strategy, whats this gaming PC all about? Like the Alienware Aurora, its ditched the spaceship look for that new stadium loop design language with tempered glass sideIntel Z890 motherboard comes with up to 24-core Intel Core Ultra 9 285KFits ATX and mATX motherboardsNvidias next-gen GeForce GPUs; room for 4-slot, 450mm cardUp to 64GB DDR5 RAM and up to 8TB of PCIe SSD storagePCIe x16 Gen 5 GPU slot and PCIe Gen 5 SSD slot (2TB Gen 5 SSD is an option)850W Gold or 1500W Platinum ATX PSUsNo exhaust fans; active intake with at least four fans, passive exhaust, gaskets to prevent leakage240mm or 360mm liquid cooling options for CPU, room for aftermarket 420mm heat exchangerThree SSD slots, two 2.5-inch drive bays, one 3.5-inch drive bayWi-Fi 7 (2x2), 2.5G ethernet standard2x 10Gbps USB-C ports, 2x 5Gbps USB-A ports, and audio jacks up top2x Thunderbolt 4 ports, 2x 10Gbps USB-C ports, and an assortment of six slower USB-A ports around back, plus audio and ethernetSeven lighting zonesWhy now? I asked McGowan, and his primary answer is that hes listening to his customers. Im talking with customers, reading reviews, understanding what the sentiment is in the market and where things are going, he tells The Verge, and what hes hearing is demand for standard mounting locations a demand so loud that Alienware decided to make a wholesale shift toward a fully upgradable computer.Thats not to say there werent reasons to go proprietary, or that Alienware is promising to do this on every PC. In fact, Alienware built its own nonproprietary ATX motherboard for this Alienware Area-51, and the new 2025 version of its smaller Aurora (a spec bump with new Nvidia and Intel chips) will still feature proprietary motherboards and power supplies, at least for now.McGowan says thats because of the leverage Dell gets with proprietary parts. If you go back years and years, there was a decision to take the power supply unit and go and drive commonality between our Dell Precision products and Alienware products, he explains. Dell got better prices that way and, he argues, more efficient, higher-quality power supplies, too, by unifying its supply chain and taking advantage of those economies of scale. And, he says, it allowed Dell to shrink the size of its PC cases at a time its commercial customers, in particular, valued a form factor aggressive chassis. Images: DellSo, does that mean the Alienware Aurora, the smaller and less expensive desktop that Dell is more likely to sell in volume, will get the ATX treatment, too? Were evaluating that for Aurora as well, McGowan tells me, but he isnt promising anything today. We have to hit an inflection point ... where we apply resources to go and redesign the internals of that chassis, he says. But Dell would need to see the numbers add up not just in terms of price, but the ratios of price, performance, size, and quality that would allow a new Aurora to compete. Theres a clear customer advantage around how much power we can put into a compact mini tower. The other [consideration] is cost related; when we get economies of scale across other Dell product and its something we can adopt with little impact on the gaming side, were going to take that and pass that savings on to the user, he says. Photo by Antonio G. Di Benedetto / The VergeThis is far from Dells only shift in strategy this year in an attempt to become more consumer-friendly: its also killing off the XPS brand for Apple-like Pro and Pro Max product lines instead, and those Pro laptops will now all feature consumer-replaceable USB-C ports in addition to user-replaceable batteries.Alienware hasnt shared the entry price or configuration of the Area-51 quite yet but says itll ship later in Q1 starting at $4,499 with a high-end, next-gen Nvidia GPU. The company is also announcing a pair of new Area-51 laptops.

In a time when global health faces persistent threats from emerging pandemics, the need for advanced biosurveillance and pathogen detection systems is increasingly evident. Traditional genomic analysis methods, while effective in isolated cases, often struggle to address the complexities of large-scale health monitoring. A significant challenge is identifying and understanding the genomic diversity in environments such as wastewater, which contains a rich mix of microbial and viral DNA and RNA. The rapid advancements in biological research have further emphasized the importance of scalable, accurate, and interpretable models to analyze vast amounts of metagenomic data, aiding in the prediction and mitigation of health crises.Researchers from the University of Southern California, Prime Intellect, and the Nucleic Acid Observatory have introduced METAGENE-1, a metagenomic foundation model. This 7-billion-parameter autoregressive transformer model is specifically designed to analyze metagenomic sequences. METAGENE-1 is trained on a dataset comprising over 1.5 trillion DNA and RNA base pairs derived from human wastewater samples, utilizing next-generation sequencing technologies and a tailored byte-pair encoding (BPE) tokenization strategy to capture the intricate genomic diversity present in these datasets. The model is open-sourced, encouraging collaboration and further advancements in the field.Technical Highlights and BenefitsMETAGENE-1s architecture draws on modern transformer models, including GPT and Llama families. This decoder-only transformer uses a causal language modeling objective to predict the next token in a sequence based on preceding tokens. Its key features include:Dataset Diversity: The training data encompasses sequences from tens of thousands of species, representing the microbial and viral diversity found in human wastewater.Tokenization Strategy: The use of BPE tokenization enables the model to process novel nucleic acid sequences efficiently.Training Infrastructure: Advanced distributed training setups ensured stable training on large datasets despite hardware limitations.Applications: METAGENE-1 supports tasks like pathogen detection, anomaly detection, and species classification, making it valuable for metagenomic studies and public health research.These features enable METAGENE-1 to generate high-quality sequence embeddings and adapt to specific tasks, enhancing its utility in the genomic and public health domains.Results and InsightsThe capabilities of METAGENE-1 were assessed using multiple benchmarks, where it demonstrated notable performance. In a pathogen detection benchmark based on human wastewater samples, the model achieved an average Matthews correlation coefficient (MCC) of 92.96, significantly outperforming other models. Additionally, METAGENE-1 showed strong results in anomaly detection tasks, effectively distinguishing metagenomic sequences from other genomic data sources.In embedding-based genomic analyses, METAGENE-1 excelled on the Gene-MTEB benchmark, achieving a global average score of 0.59. This performance underscores its adaptability in both zero-shot and fine-tuning scenarios, reinforcing its value in handling complex and diverse metagenomic data.ConclusionMETAGENE-1 represents a thoughtful integration of artificial intelligence and metagenomics. By leveraging transformer architectures, the model offers practical solutions for biosurveillance and pandemic preparedness. Its open-source release invites researchers to collaborate and innovate, advancing the field of genomic science. As challenges related to emerging pathogens and global pandemics continue, METAGENE-1 demonstrates how technology can play a crucial role in addressing public health concerns effectively and responsibly.Check out the Paper, Website, GitHub Page, and Model on Hugging Face. All credit for this research goes to the researchers of this project. Also,dont forget to follow us onTwitter and join ourTelegram Channel andLinkedIn Group. Dont Forget to join our60k+ ML SubReddit. Asif RazzaqAsif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences. [Recommended Read] Nebius AI Studio expands with vision models, new language models, embeddings and LoRA (Promoted)

LatestMachine LearningAI Innovations and Insights 16: ASSISTRAG 0 like January 6, 2025Share this postAuthor(s): Florian June Originally published on Towards AI. The Dual RAG Engine of Thinking and MemoryThis member-only story is on us. Upgrade to access all of Medium.This article is the 16th in this promising series. Today, we will explore the advancement of RAG.The video contains a mind map:Training code and data: https://github.com/smallporridge/AssistRAG.ASSISTRAG works as a helpful assistant that helps organize your work materials and keeps track of whats important. When you have a difficult problem to solve, ASSISTRAG helps by breaking it into smaller, easier parts and finding the information you need.Now, lets dive into the detailed introduction.Early RAG methods like the Retrieve-Read framework were inadequate for complex reasoning tasks. While subsequent prompt-based RAG strategies and Supervised Fine-Tuning (SFT) methods improved performance, they required frequent retraining and risked altering foundational LLM capabilities.Figure 1: Comparisons of Naive, Prompt-based, SFT-based and our Assistant-based RAG frameworks. Source: ASSISTRAG.By introducing an intelligent information assistant that integrates memory and knowledge management, ASSISTRAG compensates for LLMs shortcomings in information accuracy and reasoning depth. As shown in Figure 1, this approach consists of a trainable assistant that manages information and a static main LLM that handles task execution.Figure 2: Overview of ASSISTRAG. ASSISTRAG enhances LLMs by providing an intelligent information assistant. Endowed with the ability of tool usage, action execution, memory building Read the full blog for free on Medium.Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming asponsor. Published via Towards AITowards AI - Medium Share this post

At CES 2025, Nvidia has just revealed the GPUs behind the next generation of gaming laptops. The RTX 5000 mobile series is here, led, of course, by the GeForce RTX 5090.Unlike its desktop lineup, Nvidia has revealed a full stack of mobile graphics cards, bookended by the RTX 5060 on the low-end and the RTX 5090 on the high-end. However, keep in mind that these mobile GPUs will be much less powerful than their desktop counterparts. These new mobile GPUs are powering a wide swath of next-generation gaming laptops, also revealed at CES, from the likes of HP, MSI, Alienware and Asus. Nvidia is claiming that these new gaming laptops are significantly faster than laptops sporting last-generation parts, which is something I'll have to test once I get these devices in the lab. If you want to get your hands on a gaming laptop with an RTX 5090, however, you should start to see them on store shelves in March. Developing...The Age of 4K Gaming Laptops?While 4K gaming laptops have been a thing for a while, mobile-class graphics chips are usually a little too weak to handle the most demanding games at that resolution. While 4K performance on gaming laptops has certainly improved over the last couple of generations, there are still times where I try to boot up Cyberpunk 2077 with path tracing on an RTX 4090 laptop, only to remember that it's a bit too ambitious. With the RTX 5000 series, particularly the 5090, it looks like gaming laptops sporting 4K displays which is becoming more common on high-end devices will be much better equipped to play pretty much any game at native resolution. Either way, I won't know exactly how these new GPUs will perform at 4K or any resolution until I'm able to benchmark it. Luckily, I won't have to wait long to see what the next generation of gaming laptops will look like. Developing...

After months of waiting, Nvidia has finally announced its new line of graphics cards, led by the Nvidia GeForce RTX 5090 at CES 2025. And while there are only a couple graphics cards at the Las Vegas tech show, the rest of the lineup should filter out over the next few months. The star of the show is obviously the Nvidia GeForce RTX 5090, succeeding the massively powerful RTX 4090. And even if you can't fork over the cash for the RTX 5090, you can grab the RTX 5070, which allegedly provides RTX 4090 performance for $549.The Nvidia GeForce RTX 5090 will come out January 30, for $1,999. The RTX 5090 will be accompanied by the RTX 5080 for $999, the RTX 5070 Ti for $749 adn the RTX 5070 for $549. Developing...Kept Us Waiting, Huh?While we all kind of knew a new graphics card is on the way, the RTX 4000 series graphics cards had one of the longest runs in a while. Usually, new graphics cards come out on a roughly 2-year cycle, coming out every other fall. However, Nvidia totally missed that mark as did AMD and delivered its cards about four months later than I would expect. Nvidia RTX 5000 Blackwell ImagesLikely a big part of that wait came down to the competition or lack thereof. Right up until Nvidia CEO Jensen Huang took the stage at CES, the RTX 4090 reigned as the most powerful dekstop graphics card on the market, with nothing from AMD or Intel even coming close. Perhaps that's why it remained a $2,000 graphics card through its entire lifespan. Of course, with the RTX 5090 getting even more expensive, it seems like Nvidia is betting that this generation is going to follow the same shape as the last, with this flagship card having a wide lead on anything else on the market. Unfortunately, that also means more expensive graphics cards all around. It's possible that we'll see Nvidia launch an affordable 50-series graphics card over the next year or so, likely with the RTX 5060, but keep in mind that the RTX 4060 came out in June 2023 a whopping 8 months after the RTX 4090 hit the market in October 2022. Either way, with games like Indiana Jones and the Great Circle adopting high-end tech like full ray tracing, a graphics card like the RTX 5090 is going to go a long way to making sure your gaming PC is up to snuff for years to come or at least until the next Nvidia graphics cards come out. Jackie Thomas is the Hardware and Buying Guides Editor at IGN and the PC components queen. You can follow her @Jackiecobra

Alongside a new trailer featuring Kaitlyn Dever's Abby, HBO has confirmed The Last of Us Season 2 will officially arrive in April 2025.While we don't have an exact date, this reveal narrows the window and gives us an idea of when we will see the events of 2020's The Last of Us Part II unfold on Max.This new trailer puts a big focus on Dever's Abby, as we see her walking through the hospital that was featured in Season 1 and in the first game. We also hear her voice as she shares some words as other footage from Season 2 flashes by."It doesn't matter if you have a code like me," Abby says. "There are just some things everyone agrees are just... wrong."We won't spoil what any of these words mean, but its a very dark and ominous trailer that clearly hints at some major drama headed our way in these seven new episodes. The official Season 2 logline does this as well."Five years after the events of the first season, Joel and Ellie are drawn into conflict with each other and a world even more dangerous and unpredictable than the one they left behind," the logline reads.PlayThe Last of Us Season 2 will only tell part of the story of Naughty Dog's second game in this post-apocalyptic franchise, and showrunner Craig Mazin teased that "the story may require Season 4." Have a tip for us? Want to discuss a possible story? Please send an email to newstips@ign.com.Adam Bankhurst is a writer for IGN. You can follow him on X/Twitter @AdamBankhurst and on TikTok.

Smart glasses are all trying to get smarter. In an ongoing quest to find ways on our faces, companies are pitching ever-more-normal-looking glasses promising AI onboard that can interact with our lives using built-in cameras. Meta's already laid out a path to that future with current Meta Ray-Bans and future dreams of Meta Orion full AR glasses, while Google and Samsung have plans for AR glasses using Android XR. RayNeo, another AR glasses company, has a pair of compact AR glasses announced at this year's CES show in Las Vegas and coming this year that aim for the same zone.The RayNeo X3 Pro is an update to a line of AR glasses I've tried before, but they've gotten a lot smaller this time around. The 3-ounce glasses look pretty everyday normal from RayNeo's photos, yet they have built-in dual Micro LED displays that are projected onto the transparent lenses via wave guides. RayNeo's previous glasses had a VR chip to power its AR/AI features, the same chip that was on the Quest 2. This time, the glasses use Qualcomm's AR 1 Gen 1 chipset that's focused on shrinking glasses' size and optimizing for AI. They still have dual cameras, room tracking and even hand tracking. The X3 Pro is coming later in 2025, at a still-unknown price. RayNeoThe previous RayNeo X2 glasses were bulky, had limited battery life and were heavier (4.3 ounces compared with 3 ounces on the X3 Pro). According to reviewers, they also had a very narrow field of view, limiting the bright displays to a tiny floating window.Will the X3 Pros be better and more reliable? CNET will hopefully get to demo them soon, so we'll know more. These glasses also don't seem to be part of Google's upcoming AR- and AI-focused Android XR right now, although at some point it would make sense for them to be. The RayNeo Air X3 display glasses are similar to ones made by Xreal and others, with promised audio and video improvements. RayNeoRayNeo's other new glasses follow familiar forms: the X3 Air are display glasses, similar to what's available from Xreal and others. RayNeo promises a 200,000:1 contrast ratio, a 201-inch equivalent virtual display and better audio, but they don't seem to be able to auto-anchor displays like the new Xreal One glasses can and there are no specifics on field of view or display type. The V3, meanwhile, are camera-equipped glasses with audio, similar to Meta's Ray-Bans. All of these glasses look like they're arriving toward the second half of this year, but again, with no prices available yet.What stands out to me, though, is how much smaller the X3 Pro glasses are. AR glasses approaching the type of stuff Meta's promising are starting to arrive, although RayNeo's field of view is undoubtedly far smallerand its software services are still a relative unknown.