Microsoft is a little sneaky. Sure, theres just one big update for Windows 11 each year. But Microsofts Windows team is always working on something, and new Windows 11 features are arriving on PCs every month even outside of those high-profile updates.So as we arrive at the end of 2024, lets review the most interesting and useful new features that have shown up on Windows 11 in the past year. I bet youll find at least a few features you havent yet discovered!And be sure to sign up for my free Windows Intelligence newsletter for even more useful knowledge. Ill keep you up to date on all the interesting new features you can find and explore as we get into the new year.New Windows 11 features #13: Phone integration powersUse an Android phone as a webcam, wirelesslyWindows now lets you use your Android phone as a webcam. It all happens entirely wirelessly, if you like no cables! The setup process is quick, and its particularly useful if youre not a big fan of your PCs on-board webcam quality.Get PC notifications when you take new photos or screenshots on AndroidWindows can now pop up a notification whenever you take a new photo or screenshot on your Android phone. Then you can click that notification to immediately transfer the photo and open it for viewing, editing, or sharing on your PC.Transfer files to and from Android phones, wirelesslyOnce youve connected your Android phone to your PC for the above wireless features, youll also see your phone pop up in File Explorer. You can transfer files to and from your Android phone right from File Explorer wirelessly!New Windows 11 features #45: Windows taming tricksWhy Widgets on Windows 11 are finally worth usingWindows 11s built-in Widgets menu got some big updates in 2024. Its worth giving it a second chance now that you can tweak various options to hide the viral article feed and customize it further.Remove stocks and sports from your PCs lock screenWindows now shows stock prices and sports updates alongside the weather on your lock screen its a recent update, too! (That said, many people wont be fans of this.)New Windows 11 features #68: New Microsoft appsTry the new Sticky Notes appOut with the old and in with the new! Windows has a new Sticky Notes app thats more closely integrated with OneNote.The ultimate Windows app launcherMicrosofts PowerToys package isnt included with Windows, but its an honorary part of the operating system as far as Im concerned. In 2024, Microsoft released an especially powerful application-launcher-and-arranger named PowerToys Workspaces. Its a big productivity upgrade for many people.Activate and use the PowerToys New+ menuSpeaking of PowerToys, Microsoft also launched an even more powerful and flexible New+ menu that lets you quickly create new files and folders from templates right in File Explorer. Im already putting this one to good use myself.New Windows 11 features #912: AI grab bagErase objects from photos with AI on your PCMicrosofts Photos app includes a new Generative Erase feature that works on all Windows 11 PCs and on Windows 10 PCs, too! You can select objects in photos and use this AI-powered erase to get rid of them.Microsoft wants Copilot to be your new AI best friendWindows got a brand-spanking-new Copilot app this year. Beyond productivity, Copilot now wants to be your best buddy!5 Copilot+ PC AI features you can use at launchSome of the newest PCs are branded Copilot+ PCs, and they have access to new AI-based Windows features. Again, these dont work on most Windows 11 PCs these need a new Copilot+ PC. Here are the AI features you can use on those PCs.Hands on with Microsofts Windows RecallRecall is the most controversial Windows feature of the year. Its still available only in testing form right now but Im experimenting with it, and you can see how it works.Bonus: More Windows AI features10 Windows 11 AI features you can use todayOkay technically this one is from December 2023, but whos counting? All Windows 11 PCs include AI features, even if they arent Copilot+ PCs. This guide reveals what you can use today on any Windows 11 PC.Stay tuned: 2025 promises to be even more of a busy year when it comes to Windows development. Well explore it all together, every step of the way.Dont wait for next years December review! Sign up for my free Windows Intelligence newsletter now, and Ill send you three new things to know and try every single Friday.

In 2024, the surge in generative AI (genAI) pilot projects sparked concerns over high experimentation costs and uncertain benefits. That prompted companies to then shift their focus to delivering business outcomes, enhancing data quality, and developing talent.In 2025, enterprises are expected to prioritize strategy, add business-IT partnerships to assist with genAI projects and move from large language model (LLM) pilots to production instances. And small language models will also likely come into their own, addressing specific tasks without overburdening data center processing and power.Organizations will also adopt new technologies and architectures to better govern data and AI, with a return to predictive AI, according to Forrester Research.Predictive AI uses historical data and techniques such as machine learning and statistics to forecast future events or behaviors, said Forrester analyst Jayesh Chaurasia. GenAI, on the other hand, creates new content such as images, text, videos, or synthetic data leveraging deep learning methods such as generative adversarial networks (GANs). Chaurasia predicts the AI pendulum will swing back to predictive AI for over 50% of use cases.LLMs are, of course, central to genAI, helping enterprises tackle complex tasks and improve operations. Forrester reported that 55% of US genAI decision-makers with a strategy use LLMs embedded in applications, while 33% purchase domain-specific genAI apps. Meanwhile, SLMs are quickly gaining attention.The rise of small and mid-sized language models should enable customers to better meet the trade-offs on accuracy, speed and costs, said Arun Chandrasekaran, a distinguished vice president analyst with Gartner Research, noting that Most organizations are still struggling to realize business value from their genAI investment.GartnerIn the coming year, SLM integration could surge by as much as 60%, according to a Forrester report.As nearly eight-in-10 IT decision makers report software costs rising over the past year, theyre looking to SLMs because theyre more cost-effective and offer better accuracy, relevance, and trustworthiness by training on specific domains. Theyre also easier to integrate and excel in specialized industries such as finance, healthcare, and legal services.By 2025, 750 million apps are expected to use LLMs, underscoring the genAI markets rapid growth. Forrester predicts the market will grow in value from $1.59 billion in 2023 to $259.8 billion by 2030,.Even with that growth, many AI experts argue that LLMs may be excessive for automating workflows and repetitive tasks, both in terms of performance and environmental impact. A Cornell University study found that training OpenAIs GPT-3 LLM consumed 500 metric tons of carbon, the equivalent of 1.1 million pounds.As enterprises face challenges meeting expectations, gen AI investments in 2025 will likely shift toward proven predictive AI applications like maintenance, personalization, supply chain optimization, and demand forecasting. Forward-thinking organizations will also recognize the synergy between predictive and generative AI, using predictions to enhance generative outputs. That approach is expected to boost the share of combined use cases from 28% today to 35%, according to Forrester.What are small language models?In the simplest of terms, an SLM is a lightweight genAI model. The small in this context refers to the size of the models neural network, the number of parameters, and the volume of data on which it is trained, according toRosemary Thomas, a senior technical researcher in the AI lab at Version 1, a management consulting and software development firm.SLMs use fewer computational resources, enabling on-premises or private cloud deployment, which natively enhances privacy and security.While some SLM implementations can require substantial compute and memory resources, several models can have more than 5 billion parameters and run on a single GPU, Thomas said.Gartner Research defines SLMs differently, as language models with 10 billion parameters or less. Compared to LLMs, they are two to three orders of magnitude (around 100-1,000x) smaller, making them significantly more cost-efficient to use or customize.SLMs includeGoogle Gemini Nano, MicrosoftsOrca-27bandOrca-213b, MetasLlama-213b,andothers, Thomas noted in arecent post, arguing that SLM growth is being driven by the need for more efficient models and the speed at which they can be trained and set up.GartnerSLMs have gained popularity due to practical considerations such as computational resources, training time, and specific application requirements, Thomas said. Over the past couple of years, SLMs have become increasingly relevant, especially in scenarios where sustainability and efficiency are crucial.SLMs enable most organizations to achieve task specialization, improving the accuracy, robustness, and reliability of genAI solutions, according to Gartner. And because deployment costs, data privacy, and risk mitigation are key challenges when using genAI, SLMs offer a cost-effective and energy-efficient alternative to LLMs for most organizations, Gartner said.Three out of four (75%) of IT-decision makers believe SLMs outperform LLMs in speed, cost, accuracy and ROI, according to a Harris Poll of more than 500 users commissioned by the start-up Hyperscience.Data is the lifeblood of any AI initiative, and the success of these projects hinges on the quality of the data that feeds the models, said Andrew Joiner, CEO of Hyperscience, which develops AI-based office work automation tools. Alarmingly, three out of five decision makers report their lack of understanding of their own data inhibits their ability to utilize genAI to its maximum potential. The true potentiallies in adopting tailored SLMs, which can transform document processing and enhance operational efficiency.Gartner recommends that organizations customize SLMs to specific needs for better accuracy, robustness, and efficiency. Task specialization improves alignment, while embedding static organizational knowledge reduces costs. Dynamic information can still be provided as needed, making this hybrid approach both effective and efficient, the research firm said.In highly regulated industries, such as financial services, healthcare and pharmaceuticals, the future of LLMs is definitely small, according to Emmanuel Walckenaer, CEO of Yseop, a vendor that offers pre-trained genAI models for the BioPharma industry.Smaller, more specialized models will reduce wasted time and energy spent on building large models that arent needed for current tasks, according to Yseop.Agentic AI holds promise, but its not yet matureIn the year ahead, there is likely to be a rise in domain-specific AI agents, although it is unclear how many of these agents can live up to the lofty expectations, according to Gartners Chandrasekaran.While Agentic AI architectures are a top emerging technology, theyre still two years away from reaching the lofty automation expected of them, according to Forrester.While companies are eager to push genAI into complex tasks through AI agents, the technology remains challenging to develop because it mostly relies on synergies between multiple models, customization through retrieval augmented generation (RAG), and specialized expertise. Aligning these components for specific outcomes is an unresolved hurdle, leaving developers frustrated, Forrester said in its report.A recent Capital One survey of 4,000 business leaders and technical practitioners across industries found that while 87% believe their data ecosystem is ready for AI at scale, 70% of technologists spend hours daily fixing data issues.Still, Capital Ones survey revealed strong optimism among business leaders about their companies AI readiness. Notably, 87% believe they have a modern data ecosystem for scaling AI solutions, 84% report having centralized tools and processes for data management, 82% are confident in their data strategy for AI adoption, and 78% feel prepared to manage the increasing volume and complexity of AI-driven data.And yet, 75% of enterprises attempting to build AI agents in-house next year are expected to fail, opting instead for consulting services or pre-integrated agents from existing software vendors. To address the mismatch between AI data preparedness and real-world complexities in 2025, 30% of enterprise CIOs will integrate Chief Data Officers (CDOs) into their IT teams as they lead AI initiatives, according to Forrester Research. CEOs will rely on CIOs to bridge the gap between technical and business expertise, recognizing that successful AI requires both solid data foundations and effective stakeholder collaboration.Forresters 2024 survey also showed that 39% of senior data leaders report to CIOs, with a similar 37% reporting to CEOs and that trend is growing. To drive AI success, CIOs and CEOs must elevate CDOs beyond being mere liaisons, positioning them as key leaders in AI strategy, change management, and delivering ROI.A growing interest in multi-modality and upskillingEmerging use cases for multi-modality, particularly image and speech as modalities in both genAI inputs and outputs, will also see more adoption in 2025.Multimodal learning, a subfield of AI, enhances machine learning by training models on diverse data types, including text, images, videos, and audio. The approach enables models to identify patterns and correlations between text and associated sensory data.By integrating multiple data types, multimodal AI expands the capabilities of intelligent systems. These models can process various input types and generate diverse outputs. For example, GPT-4, the foundation of ChatGPT, accepts both text and image inputs to produce text outputs, while OpenAIs Sora model generates videos from text.Other examples include medical imaging, patient history, and lab results that can be integrated to enhance pateitn diagnosis and treatment. In financial services, multimodal AI can analyze customer phone queries to assist contact center employees in resolving issues. And in the automotive industry inputs from cameras, GPS, and LiDAR can be integrated by AI to enhance autonomous driving, emergency response, and navigation for companies, such as Tesla, Waymo and Li Auto.AI leaders have also realized they need to prioritize business outcomes, clean their data houses, and start building AI talent. The latter is especially important given the growing gap between enterprise genAI needs and the workers with skills to meet those needs.In the year ahead, youll need to put your nose to the grindstone to develop an effective AI strategy and implementation plan, Forrester said in its report. In 2025, organizational success will depend on strong leadership, strategic refinement, and recalibration of enterprise data and AI initiatives commensurate with AI aspirations.

Need ideas or motivation to help you build a spreadsheet in Google Sheets? You can browse through the templates that are included in this office app and select one to customize. But a more intriguing option is to use the tool in Sheets called Help Me Organize. Powered by Googles generative AI technology, Gemini, you can use it to generate a template thats more tailored for you.Based on a brief description that you write (referred to as a prompt), Help Me Organize generates a table with headings, placeholder text, and possible formulas in its cells that you can then adjust to your needs. Its mainly designed to create templates for project management. But you can tease it to make templates that include some formulas and tables that can be used to create charts.This guide explains how to use Help Me Organize and provides tips for getting best results.Who can use Gemini AI in Google SheetsIf you have a Google Workspace account, the Gemini AI tools that include Help Me Organize are available as an add-on called Gemini for Google Workspace for an extra subscription charge. If you have a regular Google personal account, you can pay for a Google One AI Premium subscription to have access to these tools. Or, for no cost, you can sign up for access to Workspace Labs with your Google account to be permitted to try out Help Me Organize.How to access Help Me Organize in Google SheetsYou access the Help Me Organize tool from a right side panel that you open while in a spreadsheet in Google Sheets. The spreadsheet can have existing data on it. But for generating templates, its best to use Help Me Organize on a new, blank spreadsheet or on a new sheet in an existing spreadsheet. You can add a new sheet to a spreadsheet by clicking the + sign thats toward the lower-left corner of the opened spreadsheet.To launch the Help me organize panel, click Insert and select Help me organize at the very bottom of the menu that opens.In the Help me organize panel that opens to the right of the page, a large text entry box invites you to write a prompt inside it. Some example prompts that are meant to show you how you can write your own cycle through this box.When you open the Help me organize panel, its entry box shows example prompts.Howard Wen / IDGHow to use Help Me OrganizeClick inside the entry box on the Help me organize panel, type a description of the kind of template you want Gemini to generate, and click Create.Type a prompt into the box on the Help me organize panel and click Create.Howard Wen / IDGDepending on the complexity of your prompt, it may take several seconds for the AI to generate a template but it may not be able to generate anything. If its unable to, try entering your description again but use fewer words.Gemini may think for a while as it generates a template.Howard Wen / IDGHow to insert a template generated by GeminiIf Gemini produces a result, the template will appear over your spreadsheet. Itll start from the upper-leftmost cell, with the templates columns and rows spreading out from here.Review the template Gemini generated, then insert it in your spreadsheet or start over.Howard Wen / IDGYou can scroll through the template to see what you think of it. Keep in mind that you should always consider what Gemini generates as a rough draft that youll need to modify to make it more suitable for your use (such as replacing placeholder text and scrutinizing and modifying any formulas). It is a template, after all.Scroll to the bottom of the template youll find a small toolbar attached to it. If you like this template, click Insert. Itll then be inserted into your spreadsheet.If you dont, click the X. The template will be removed from your spreadsheet. You can try writing another prompt in the Help me organize panel. Note that if you create a new template, you cant go back to the previous version.Optionally, you can rate if you like this template or not by clicking the thumbs up or thumbs down icon. Your feedback is used to help train Gemini to produce results in the future that may be more preferable.Once youve inserted a template in your spreadsheet, you can tweak it however you like: change heading names, add rows or columns, adjust formulas, enter real data, and so on. See How to use Google Sheets for project management for details on working with templates in Sheets.How to write a prompt in Help Me OrganizeUnsure about how to write a prompt? Need inspiration? Here are some general tips that can elicit useful templates from Gemini:1. First, describe a specific project that you want to track.Examples:budget breakdownbusiness travel itinerarypayroll schedule2. Describe or specifically name headings that youd like to see in the template.budget breakdown that includes in the following order: revenue, rent, utilities, internet, expensesbusiness travel itinerary with sections for travel to airport, airline, flight number, hotel, and so onpayroll schedule for employees named Mike, Pedro, Shawna, and TashaIncluding specifics in a prompt will help Gemini generate a better template.Howard Wen / IDG3. Use numbers and math formulas.a table depicting 12 months with 3 categories per monthpayroll schedule thats monthly across one yeara table that calculates compound interest at 3.5% over 3 yearsGemini can create a template that includes formulas.Howard Wen / IDG4. Describe dropdowns, lists, task lists, or to-dos.a dropdown with selections that include Greek, Japanese, Italian for a business luncheona project tracker with task lists assigned to peoplea to-do list for onboarding a new employeeA generated to-do list with status dropdowns.Howard Wen / IDGFor more on working with dropdowns, see Google Sheets power tips: How to use dropdown lists.5. Describe a chart.Gemini cant generate charts directly, but you can prompt it to create a template (table) that you can then derive a chart from. Examples:a bar chart with 9 labelsa line chart with 4 categoriesa pie chart depicting 3 categoriesDescribe a chart youd like to create.Howard Wen / IDGInsert the generated template in your spreadsheet.Next, select the template by clicking its top-leftmost cell.Then, on the menu bar over your spreadsheet, click Insert > Chart. By default, a pie chart will be generated. The Chart editor panel will also open along the right of the page, so you can change the pie chart to another type or make other adjustments to it.A pie chart based on a template that Gemini generated from a prompt.Howard Wen / IDGIts worth noting that in this pie chart example, Gemini went beyond what was asked for, breaking each of the three categories into three sections with different colors. Thus, the resulting pie chart has nine sections instead of three. Thats unlikely to be what most people would be looking for from the original prompt a good illustration of why you always need to check and adjust Geminis output, or simply discard it and start over.6. Dont be afraid to describe something complicated.a budget for at least 12 departments in my office for one year and assign supervision to employeesa project manager for 8 salespersons who have to sell seashells to the 10 biggest cities in the US Midwest with an April deadlinea weekly restaurant employee work schedule for 10 back-of-house kitchen employees and 6 front-of-house employees over 4 weeksGemini can often handle complicated prompts.Howard Wen / IDGRemember that the best way to use Help Me Organize (or any generative AI tool) is to experiment and play around with the wording of your prompts. You never know Gemini may surprise you with a result thats better or more useful than what you originally envisioned.In summaryKeep these tips in mind to write prompts that will trigger Gemini to give you the best (or at least most interesting) results in Help Me Organize:Define exactly what you want to use the template for. How would you describe it in three words?Use headings or numbers (such as dates or math formulas). These can imply columns and rows in the template.If you want the template to have a dropdown or other list type, describe it.Use Gemini to generate a table that you can then derive a chart from.Dont be afraid to experiment even if your request sounds complicated.As with all AI-generated content, the templates created with Help Me Organize should never be seen as final but they can give you a big head start for all sorts of spreadsheet-related tasks, from setting up complex schedules to creating charts to performing time-oriented calculations.This article was originally published in February 2024 and updated in December 2024.

Separating AI reality from hyped-up fiction isnt always easy. Thats why weve created the AI Hype Indexa simple, at-a-glance summary of everything you need to know about the state of the industry.More than 70 countries went to the polls in 2024. The good news is that this year of global elections turned out to be largely free from any major deepfake campaigns or AI manipulation. Instead we saw lots of AI slop: buff Trump, Elon as ultra-Chad, California as catastrophic wasteland. While some worry that development of large language models is slowing down, you wouldnt know it from the steady drumbeat of new products, features, and services rolling out from itty-bitty startups and massive incumbents alike. So whats for real and whats just a lot of hallucinatory nonsense?

By the time this crowd gathered on New Year's Eve, 1938, the Times Square ball drop had been an annual rite for decades Bettman / Getty ImagesOne Times Square, the billboard-covered building at the intersection of Broadway, Seventh Avenue and 42nd Street in Midtown Manhattan, is the definitive place where one year passes into the next.Each year, thousands of people pack around the building to watch the Times Square ball slowly descend a pole marking the New Yeara beloved tradition that began on this day in 1907. Viewers, including almost a billion watching on television, witness a glittering spectacle, sponsored by companies angling to cash in on New Years resolutions.The fanfare in Times Square began 120 years ago. Adolph Simon Ochs, the publisher and owner of the New York Times, wanted to increase publicity for his papers new headquarters at One Times Square. So, as 1904 passed into 1905, workers launched a pyrotechnic show from the building.From base to dome, the Times wrote, the giant structure was alighta torch to usher in the new born, a funeral pyre for the old which pierced the very heavens.The fireworks only lasted two more years. In 1907, Ochs hired sign maker and advertiser Artkraft Strauss to build an iron and wood ball weighing 700 pounds, five feet in diameter and covered in 100 lightbulbs.The first Times Square ball drop was a rousing, if messy, success. Crowds began pouring out of the subway station the night of December 31, the Times wrote, reporting, perhaps with exaggeration, that its roars were heard as far north as Yonkers, more than ten miles up the Hudson River.Hurrah for 1908, the crowd cheered when the ball dropped down the pole at midnight. As the Times went to press later that early morning, broken horns, demolished and cast-aside rattlers, soiled confetti, forlorn ticklers and dented and deserted derby hats littered Times Square.The history of time balls dates to 1818, when future Royal Navy admiral Robert Wauchope developed a visible signal to help ships crews reliably tell time while at sea. At first he tried flags, but settled on a ball that descended a pole at a given time of day. In 1833, the Royal Observatory at Greenwich installed a time ball on its roof so ships on the Thames could coordinate their times. Every day since then, the ball raises halfway up the mast at 12:55 p.m., reaches the top at 12:58 p.m. and falls back down exactly at 1:00 p.m., delighting visitors even today.Ochs may have found inspiration for the Times Tower in a ball atop the Western Union Building in Manhattans Financial District that dropped every day at noon, synchronizing with the Naval Observatory in Washington, D.C. via telegraph to ensure militaristic accuracy.Since 1907, the ceremony at One Times Square has taken place every year except 1942 and 1943, when wartime dimouts prohibited excessive light and pomp. During those dark midnights, recordings of bells followed a moment of silence.The Times left its flagship building in 1961 after selling the structure to an advertising firm. Allied Chemical bought it in 1963 and stripped Ochs ornate building into a modernist glass skyscraper. Through countless negotiations and purchases since then, the Times Square ball evolved, varying in weight, size and illumination. In the 1980s, it became a red apple with a green stem to support a New York tourism campaign.After more than a century, the ball is at its most opulent today. According to its official website, the Big Ball is covered in 32,256 LEDs. Its 12 feet in diameter, weighs nearly six tons and sparkles all year long.Get the latest stories in your inbox every weekday.Filed Under: New Year's Day, New York City, On This Day in History, Rituals and Traditions

Sumo Group CEO Carl Cavers receives OBE in New Year's Honours listFor services to the video games industry News by Christopher Dring Head of Games B2B Published on Dec. 31, 2024 The co-founder and CEO of Sumo Group Carl Cavers has been made an Officer of the Order of the British Empire in the King's New Year's Honours List.He receives the OBE for services to the video games industry.Cavers has spent 30 years in video games, including senior roles at Gremlin and Infogrammes before setting up Sumo in 2003 alongside Paul Porter, Darren Mills and James North-Hearn.Sumo is best known as a work-for-hire studio, and has created high profile games including Sonic All-Stars Racing Transformed, LittleBigPlanet 3, Sackboy: A Big Adventure and Crackdown 3. It's also supported the development of titles including Forza Horizon 2, Hogwarts Legacy and Call of Duty Vanguard, and is increasingly developing and publishing its own projects, such as Snake Pass and Still Wakes The Deep.Sumo is a frequent winner of the GamesIndustry.biz Best Places To Work Awards, particularly around its work with education and young people. In 2021, it was acquired by Tencent in a $1.27 billion deal, and despite some recent restructuring, it remains one of the largest development groups in the UK.Outside of Sumo, Cavers is a supporter of the wider games business. He is actively involved in mentorship and networking groups in the North of England and is a vice president of games industry charity SpecialEffect."I'm humbled to have been recognised in the Kings New Year Honours list with an OBE," Cavers told GamesIndustry.biz."2025 marks my 30th year in the video games industry and 22 years since co-founding Sumo Digital alongside Paul, Darren and James. The sector has changed a lot over the last three decades but the passion, enthusiasm and creativity it exudes has not waned and despite some challenging times it continues to mature into a global phenomenon. There are lots people I want to thank, too many to list here, however it is the people in our industry who make it great and to stand alongside them and be recognised for making a contribution really is an honour in itself.

Putting people first practical tips for mental health and wellbeing support | HR Summit VideoHayley Blundy of Futurlab offers a series of practical advice on supporting employees during the global mental health crisis News by Christopher Dring Head of Games B2B Published on Dec. 31, 2024 As the year comes to a close, GamesIndustry.biz has been running our videos from September's HR Summit in London.You can watch seven of the sessions on our YouTube channel right now. And the last one to go live is this great talk from Hayley Blundy of Futurlab, who hosted a talk titled: Putting people first practical tips for mental health and wellbeing support.The talk was entirely focused on a series of practical advice on supporting employees during the global mental health crisis.The HR Summit took place at the Royal Institution in London back in September, and preceded the Best Places To Work Awards 2024. To watch all the talks from the event, simply click here.Watch on YouTube

LG has announced additions to its ultra-light Gram and Gram Pro laptop lineup, adding cloud-based and on-device AI-powered features that go beyond its current Gram laptops.The 16-inch Gram Pro will also be the first Copilot Plus PC in the LG Gram lineup and is further distinguished as the only model using the Intel Lunar Lake Core Ultra V-Series processors. The 17-inch Gram Pro and 16-inch 2-in-1 use Intels Arrow Lake Core Ultra H-Series processors.The three LG Gram Pro laptops, including the 2-in-1. Image: LGThe LG Gram Pro will be available in 17-inch and 16-inch models featuring 2,560 x 1,600 displays, up to 32GB of LPDDR5X memory, and up to 2TB Gen4 NVMe SSDs. The 16-inch Gram Pro and the 16-inch LG Gram Pro 2-in-1 will have Intel Arc GPUs, while the 17-inch Gram Pro will instead feature an Nvidia GeForce RTX4050 graphics card.Gram Chat On-Device, which uses a small language model derived from LG AI Researchs EXAONE large language model, powers offline features, including Time Travel, which lets users quickly revisit web pages, documents, videos and audio files theyve recently accessed. LGs software is adding tools similar to Microsofts Copilot Plus suite and Apple Intelligence, but given the trouble Microsoft has had with Recall, well have to wait and see how it all measures up. It also might make those features available on more PCs, but LG hasnt specified which AI features will be available on which laptops in the new Gram lineup.Gram Chat Cloud is powered by OpenAIs GPT-4o and responds to inquiries using vast web-based datasets for detailed and comprehensive responses while integrating with calendar and email services. It requires an active internet connection and will only be free for the first year.All of the new Gram laptops also support LGs Gram Link 2.0, which streamlines document and file sharing with other PCs and iOS or Android-based smartphones. It also allows incoming phone calls to be answered through the Gram laptops, so you dont have to swap headsets or Bluetooth headphones to another device temporarily.LG will also introduce its entry-level Gram Book to the US market next year. Powered by an Intel Core i5 processor,it features a 15.6-inch 60Hz full HD display, a 720p webcam, and configurations of up to 1TB of SSD storage and 16GB of DDR4 memory.

Vision Transformers (ViTs) have become a cornerstone in computer vision, offering strong performance and adaptability. However, their large size and computational demands create challenges, particularly for deployment on devices with limited resources. Models like FLUX Vision Transformers, with billions of parameters, require substantial storage and memory, making them impractical for many use cases. These limitations restrict the real-world application of advanced generative models. Addressing these challenges calls for innovative methods to reduce the computational burden without compromising performance.Researchers from ByteDance Introduce 1.58-bit FLUXResearchers from ByteDance have introduced the 1.58-bit FLUX model, a quantized version of the FLUX Vision Transformer. This model reduces 99.5% of its parameters (11.9 billion in total) to 1.58 bits, significantly lowering computational and storage requirements. The process is unique in that it does not rely on image data, instead using a self-supervised approach based on the FLUX.1-dev model. By incorporating a custom kernel optimized for 1.58-bit operations, the researchers achieved a 7.7 reduction in storage and a 5.1 reduction in inference memory usage, making deployment in resource-constrained environments more feasible.Technical Details and BenefitsThe core of the 1.58-bit FLUX lies in its quantization technique, which restricts model weights to three values: +1, -1, or 0. This approach compresses parameters from 16-bit precision to 1.58 bits. Unlike traditional methods, this data-free quantization relies solely on a calibration dataset of text prompts, removing the need for image data. To handle the complexities of low-bit operations, a custom kernel was developed to optimize computations. These advances lead to substantial reductions in storage and memory requirements while maintaining the ability to generate high-resolution images of 1024 1024 pixels.Results and InsightsExtensive evaluations of the 1.58-bit FLUX model on benchmarks such as GenEval and T2I CompBench demonstrated its efficacy. The model delivered performance on par with its full-precision counterpart, with minor deviations observed in specific tasks. In terms of efficiency, the model achieved a 7.7 reduction in storage and a 5.1 reduction in memory usage across various GPUs. Deployment-friendly GPUs, such as the L20 and A10, further highlighted the models practicality with notable latency improvements. These results indicate that 1.58-bit FLUX effectively balances efficiency and performance, making it suitable for a range of applications.ConclusionThe development of 1.58-bit FLUX addresses critical challenges in deploying large-scale Vision Transformers. Its ability to significantly reduce storage and memory requirements without sacrificing performance represents a step forward in efficient AI model design. While there is room for improvement, such as enhancing activation quantization and fine-detail rendering, this work sets a solid foundation for future advancements. As research continues, the prospect of deploying high-quality generative models on everyday devices becomes increasingly realistic, broadening access to powerful AI capabilities.Check out the Paper. 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. Sana Hassan+ postsSana Hassan, a consulting intern at Marktechpost and dual-degree student at IIT Madras, is passionate about applying technology and AI to address real-world challenges. With a keen interest in solving practical problems, he brings a fresh perspective to the intersection of AI and real-life solutions. [Download] Evaluation of Large Language Model Vulnerabilities Report (Promoted)