At CES 2025, Oshkosh launched a number of specialty vehicles with better electric power, safety, quiet and cleanliness.Read More

Maingear has launched two more models in its Apex line of liquid-cooled desktop gaming PCs: the Apex Force and the Apex Rush.Read More

Rocksteady cut further jobs at the end of 2024It's unclear how many people were impacted by this second redundancy round in just three months News by Vikki Blake Contributor Published on Jan. 7, 2025 Suicide Squad: Kill the Justice League developer Rocksteady made another round of layoffs just before the close of 2024.It marks the second round of job cuts in just three months.As reported by Eurogamer, half a dozen developers impacted by the cuts confirmed to our sister site that they had lost their jobs, including employees working in the programming, artistry, and QA teams.The cuts came just weeks after the Warner Bros. subsidiary Rocksteady Studios laid off a separate round of staff in September 2024.It's unclear how many people were impacted by the layoffs and as yet, Warner Bros. has not responded to press requests for comments.Warner Bros acknowledged shortly after launch that Suicide Squad "[fell] short of expectations]," added to the tough comparison with the launch of Hogwarts Legacy in the same window the previous year.Shortly thereafter came reports detailing various problems during Suicide Squad's development, including frequent changes in vision, perfectionist leadership, and Rocksteady's lack of experience with both multiplayer and live service games.

Justin Carter, Contributing EditorJanuary 7, 20252 Min ReadImage via Jar of Sparks/NetEase.At a GlanceJar of Sparks was founded in 2022 by 343 Industries alum Jerry Hook, with the intent of making narrative-driven action games.Jar of Sparks, an indie developer spun up in 2022, is pausing development on its debut project as it searches for funding.On LinkedIn, founder Jerry Hook explained current staff would be allowed to look for new opportunities during this process. "If youre searching for exceptional professionals who are driven and think outside the box, please reach out," he wrote. "Wed be more than happy to connect you with some of the best developers I have had the pleasure of working beside."Jar of Sparks was founded as a NetEase subsidiary, and Hook said in 2022 the studio intended to make "a new generation of narrative-driven action games." This marks one of the first developers to cease operations and cut staff in 2025.Throughout 2022 and 2023, NetEase courted veterans from developers like Ubisoft, Riot Games, and Cryptic Studios to spin up their own teams. As of last year, the status of those other studios is unknown. By December, the tech giant was reportedly preparing to shut down Ouka Studios, which had released Visions of Mana in August.Roadblocks for other new studios from veteran developersBefore 2024 ended, two other developers were in similar positions as Jar of Sparks. Worlds Unknown, a studio headed up by BioWare alum Mac Walters, also ceased operations to hunt for a new publishing partner. Similarly, Tales of Kenzera creator Surgent Studios paused activities and offered a public first look at its current project, 'Uso,' to lure prospective partners.Of Jar's first game, Hook said the team "strived to create something truly new and exciting for the industry. We couldnt be prouder of the groundwork weve laid together. [...] From the day we formed to this very moment, this team has accomplished remarkable featsproving that a small, focused, and dedicated group can achieve great results. We thank each of them for their courage, skill, and friendship.""Thank you to everyone who supported us throughout our journey here at Jar of Sparks," he concluded, "and we look forward to updating you all with our next step in our journey."Read more about:LayoffsAbout the AuthorJustin CarterContributing Editor, GameDeveloper.comA Kansas City, MO native, Justin Carter has written for numerous sites including IGN, Polygon, and SyFy Wire. In addition to Game Developer, his writing can be found at io9 over on Gizmodo. Don't ask him about how much gum he's had, because the answer will be more than he's willing to admit.See more from Justin CarterDaily news, dev blogs, and stories from Game Developer straight to your inboxStay UpdatedYou May Also Like

Private Division, the indie game label from Take-Two, has been reportedly acquired by staff from last year's mass exodus at Annapurna Interactive.Per Bloomberg, the new, still-unnamed studio will inherit games and franchises previously published under the Private Division banner. The titles that will fall under this new developer's purview include the Kerbal Space Program series, and an unannounced project from Pokmon developer Game Freak.Last November, it was reported Take-Two had sold the label to a then-unknown company. According to Bloomberg, the buyer was the Austin-based private equity firm Haveli, which reportedly struck a deal to fund the Annapurna alums' studio. As a result of this purchase, some of the near-20 Private Division staffers will be laid off.What happened with Private Division and Annapurna InteractiveLast summer, the Grand Theft Auto publisher planned to either close (or shut down) Private Division. That was preceded by the full closure of Roll7 (Rollerdrome, OlliOlli) and layoffs at Kerbal creator Intercept Games. Yellow Brick Games and Bloober Team, two developers whose projects were to be published by Private Division, respectively switched to self-publishing or had their deal with the label fully dissolved.Over at Annapurna Interactive, an alleged power struggle primarily between studio leadership led to a resignation wave among staff. Among its many issues, the video game division of the popular film studio was reportedly set to launch its own offshoot called Verset, which would handle current and future Annapurna Interactive deals, and indie work of its own.Instead, discussions between Verset's planned leaders, James Masi and Nathan Gary, and Interactive co-founder Megan Ellison hit a standstill. Despite attempted negotiations, the 25-person Interactive staff all departed. The reported plan at the time for the staff was to launch its own publisher. What remains of Annapurna Interactive began hiring staff so it could make good on pre-existing deals.Bloomberg's full report on Private Division's new situation can be read here.

RCA has announced a new line of bright QLED TVs designed to be installed and blend into outdoor settings with bezels finished in a spirit of wilderness and Mossy Oak camouflage design. Theyre dust and water-resistant, and will work in temperatures ranging from minus 22 to 122 degrees Fahrenheit.The company hasnt announced when its new outdoor TV lineup will be available. When they are, the TVs will be offered in four sizes: a $3,999 43-inch model, a $4,999 55-inch model, a $5,999 65-inch model, and the largest, a 75-inch model for $7,999.Why does RCAs 43-inch quantum dot TV cost almost four grand when similarly-sized alternatives can be found for less than $300? The TV in your living room almost certainly doesnt come with an IP55 rating. The RCAs are built with a scratch-resistant aluminum case that can withstand dust and moisture ingress. You wont want to leave them out in a downpour, but they can survive being sprayed with a hose or pummeled with wind-blown rain.Other features include up to 2,000 nits of brightness that helps keep the TVs viewable in bright sunlight, Dolby Atmos support, and Google TV to provide access to various streaming apps.

At CES this week, Savant Systems announced Savant Smart Budget, a feature of its Smart Power system of modular relays and equipment that integrates with your existing circuit breaker box.If youre already at the limits of your breaker boxs capacity, Smart Budget lets you get around that with automated control of individual circuits. That way, you can add more high-draw connections, like appliances or EV chargers, than your electrical box can supply at once. For instance, you could set it so that power only goes to your EV overnight after youre done using your oven. That sort of control can also be useful if youre using a house battery or running on solar power.Savants Smart Budget software. Image: SavantSavant says its system, which starts at $1,500 and requires installation by a licensed electrician, is more affordable than the alternative of working with your electric utility provider to upgrade to higher amperage service, which could cost in the tens of thousands of dollars. Those parts fit into most major electrical panels that standardize on 1 breaker spacing, company CMO J.C. Murphy tells The Verge, including panels from Schneider, Eaton, GE, ABB, Siemens, and others.The Smart Budget kit will include two 30-amp single-pole circuit breakers, which Savant calls Power Modules, along with a double-pole 60-amp one and a current tracker for circuits you only want to monitor, according to Murphy. It also includes a Savant Director hub and sensors. The company sells additional Power Modules that cost $120 for dual 20-amp or single-pole 30-amp versions and $240 for a 60-amp double-pole module.

SteamOS was always supposed to be bigger than Valves own Steam Deck, and 2025 is the year it finally expands. Not only will Lenovo ship the first third-party SteamOS handheld this May, Valve has now revealed it will let you install a working copy of SteamOS on other handhelds even sooner than that. Pierre-Loup Griffais, one of the lead designers on the Steam Deck and SteamOS, tells me a beta for other handhelds is slated to ship after March sometime, and that you might discover the OS just starts working properly after that happens! Griffais and his co-designer Lawrence Yang would not confirm which handhelds might just start working, though there are some obvious candidates: the company confirmed to us in August that it had been adding support for the Asus ROG Allys controls. Also, quite a few PC gamers have also discovered that Bazzite, a fork of Valves Steam Deck experience that I loved testing on an Ally X and vastly preferred to Windows, also works wonderfully on the Lenovo Legion Go. There still arent that many handhelds out there at the end of the day, and I would think Valve would take advantage of work the Linux gaming community has already done on both.RelatedSpeaking of Bazzite, Valve seems to be flattered! We have nothing against it, says Yang. Its a great community project that delivers a lot of value to people that want a similar experience on devices right now, says Griffais, adding later In a lot of ways Bazzite is a good way to kind of get the latest and greatest of what weve been working on, and test it. But he says Bazzite isnt yet in a state where a hardware manufacturer could preload it on a handheld, nor would Valve allow that. While users can freely download and install the SteamOS image onto their own devices, companies arent allowed to sell it or modify it, and must partner with Valve first.There are some non-selfish reasons for that. Among other things, Griffais explains that the Lenovo Legion Go S will run the same SteamOS image as the Steam Deck itself, taking advantage of the same software updates and the same precached shaders that let games load and run more smoothly, just with added hardware compatibility tweaks. Valve wants to make sure SteamOS is a single platform, not a fragmented one.In general, we just want to make sure we have a good pathway to work together on things like firmware updates and you can get to things like the boot manager and the BIOS and things like that in a semi-standardized fashion, right? says Griffais, regarding what Valve needs to see in a partnership that would officially ship SteamOS on other devices.Valve isnt currently partnered with any other companies beyond Lenovo to do that collaboration Yang tells me the company is not working with GPD on official SteamOS support, despite that manufacturers claim. Valves also not promising that whichever Windows handheld you have will necessarily run SteamOS perfectly in a new blog post, Valve only confirms that a beta will ship before Lenovos Legion Go S, that it should improve the experience on other devices, and that users can download and test this themselves.As far as other form factors, like possible SteamOS living room boxes, Valve says you might have a good experience trying that. And partnerships are a possibility there too: if someone wants to bring that to the market and preload SteamOS on it, wed be happy to talk to them.Valve wouldnt tell me anything about the rumors that its developing its own Steam Controller 2, VR headset with wands, and possibly its own living room box, but did tell me that we might expect more Steam Input compatible controllers in the future.

Building Large Action Models: Insights from Microsoft 0 like January 7, 2025Share this postAuthor(s): Jesus Rodriguez Originally published on Towards AI. Created Using MidjourneyI recently started an AI-focused educational newsletter, that already has over 175,000 subscribers. TheSequence is a no-BS (meaning no hype, no news, etc) ML-oriented newsletter that takes 5 minutes to read. The goal is to keep you up to date with machine learning projects, research papers, and concepts. Please give it a try by subscribing below:TheSequence | Jesus Rodriguez | SubstackThe best source to stay up-to-date with the developments in the machine learning, artificial intelligence, and datathesequence.aiAction execution is one of the key building blocks of agentic workflows. One of the most interesting debates in that are is whether actions are executed by the model itself or by an external coordination layer. The supporters of the former hypothesis have lined up behind a theory known as large action models(LAMs) with projects like Gorilla or Rabbit r1 as key pioneers. However, there are still only a few practical examples of LAM frameworks. Recently, Microsoft Research published one of the most complete papers in this area outlining a complete framework for LAM models. Microsofts core idea is to simply bridge the gap between the language understanding prowess of LLMs and the need for real-world action execution.From LLMs to LAMs: A Paradigm ShiftThe limitations of traditional LLMs in interacting with and manipulating the physical world necessitate the development of LAMs. While LLMs excel at generating intricate textual responses, their inability to translate understanding into tangible actions restricts their applicability in real-world scenarios. LAMs address this challenge by extending the expertise of LLMs from language processing to action generation, enabling them to perform actions in both physical and digital environments. This transition signifies a shift from passive language understanding to active task completion, marking a significant milestone in AI development.Image Credit: Microsoft ResearchKey Architectural Components: A Step-by-Step ApproachMicrosofts framework for developing LAMs outlines a systematic process, encompassing crucial stages from inception to deployment. The key architectural components include:Data Collection and PreparationThis foundational step involves gathering and curating high-quality, action-oriented data for specific use cases. This data includes user queries, environmental context, potential actions, and any other relevant information required to train the LAM effectively. A two-phase data collection approach is adopted:Task-Plan CollectionThis phase focuses on collecting data consisting of tasks and their corresponding plans. Tasks represent user requests expressed in natural language, while plans outline detailed step-by-step procedures designed to fulfill these requests. This data is crucial for training the model to generate effective plans and enhance its high-level reasoning and planning capabilities. Sources for this data include application documentation, online how-to guides like WikiHow, and historical search queries.Task-Action CollectionThis phase converts task-plan data into executable steps. It involves refining tasks and plans to be more concrete and grounded within a specific environment. Action sequences are generated, representing actionable instructions that directly interact with the environment, such as select_text(text=hello) or click(on=Button(20), how=left, double=False). This data provides the necessary granularity for training a LAM to perform reliable and accurate task executions in real-world scenarios.Image Credit: Microsoft ResearchModel TrainingThis stage involves training or fine-tuning LLMs to perform actions rather than merely generate text. A staged training strategy, consisting of four phases, is employed:Phase 1: Task-Plan Pretraining: This phase focuses on training the model to generate coherent and logical plans for various tasks, utilizing a dataset of 76,672 task-plan pairs. This pretraining establishes a foundational understanding of task structures, enabling the model to decompose tasks into logical steps.Phase 2: Learning from Experts: The model learns to execute actions by imitating expert-labeled task-action trajectories. This phase aligns plan generation with actionable steps, teaching the model how to perform actions based on observed UI states and corresponding actions.Phase 3: Self-Boosting Exploration: This phase encourages the model to explore and handle tasks that even expert demonstrations failed to solve. By interacting with the environment and trying alternative strategies, the model autonomously generates new success cases, promoting diversity and adaptability.Phase 4: Learning from a Reward Model: This phase incorporates reinforcement learning (RL) principles to optimize decision-making. A reward model is trained on success and failure data to predict the quality of actions. This model is then used to fine-tune the LAM in an offline RL setting, allowing the model to learn from failures and improve action selection without additional environmental interactions.Image Credit: Microsoft ResearchIntegration and GroundingThe trained LAM is integrated into an agent framework, enabling interaction with external tools, maintaining memory, and interfacing with the environment. This integration transforms the model into a functional agent capable of making meaningful impacts in the physical world. Microsofts UFO, a GUI agent for Windows OS interaction, exemplifies this integration. The AppAgent within UFO serves as the operational platform for the LAM.EvaluationRigorous evaluation processes are essential to assess the reliability, robustness, and safety of the LAM before real-world deployment. This evaluation involves testing the model in a variety of scenarios to ensure generalization across different environments and tasks, as well as effective handling of unexpected situations. Both offline and online evaluations are conducted:Offline Evaluation: The LAMs performance is assessed using an offline dataset in a controlled, static environment. This allows for systematic analysis of task success rates, precision, and recall metrics.Online Evaluation: The LAMs performance is evaluated in a real-world environment. This involves measuring aspects like task completion accuracy, efficiency, and effectiveness.Image Credit: Microsoft ResearchKey Building Blocks: Essential Features of LAMsSeveral key building blocks empower LAMs to perform complex real-world tasks:Action Generation: The ability to translate user intentions into actionable steps grounded in the environment is a defining feature of LAMs. These actions can manifest as operations on graphical user interfaces (GUIs), API calls for software applications, physical manipulations by robots, or even code generation.Dynamic Planning and Adaptation: LAMs are capable of decomposing complex tasks into subtasks and dynamically adjusting their plans in response to environmental changes. This adaptive planning ensures robust performance in dynamic, real-world scenarios where unexpected situations are common.Specialization and Efficiency: LAMs can be tailored for specific domains or tasks, achieving high accuracy and efficiency within their operational scope. This specialization allows for reduced computational overhead and improved response times compared to general-purpose LLMs.Agent Systems: Agent systems provide the operational framework for LAMs, equipping them with tools, memory, and feedback mechanisms. This integration allows LAMs to interact with the world and execute actions effectively. UFOs AppAgent, for example, employs components like action executors, memory, and environment data collection to facilitate seamless interaction between the LAM and the Windows OS environment.The UFO Agent: Grounding LAMs in Windows OSMicrosofts UFO agent exemplifies the integration and grounding of LAMs in a real-world environment. Key aspects of UFO include:Architecture: UFO comprises a HostAgent for decomposing user requests into subtasks and an AppAgent for executing these subtasks within specific applications. This hierarchical structure facilitates the handling of complex, cross-application tasks.AppAgent Structure: The AppAgent, where the LAM resides, consists of:Environment Data Collection: The agent gathers information about the application environment, including UI elements and their properties, to provide context for the LAM.LAM Inference Engine: The LAM, serving as the brain of the AppAgent, processes the collected information and infers the necessary actions to fulfill the user request.Action Executor: This component grounds the LAMs predicted actions, translating them into concrete interactions with the applications UI, such as mouse clicks, keyboard inputs, or API calls.Memory: The agent maintains a memory of previous actions and plans, providing crucial context for the LAM to make informed and adaptive decisions.Image Credit: Microsoft ResearchEvaluation and Performance: Benchmarking LAMsMicrosoft employs a comprehensive evaluation framework to assess the performance of LAMs in both controlled and real-world environments. Key metrics include:Task Success Rate (TSR): This measures the percentage of tasks successfully completed out of the total attempted. It evaluates the agents ability to accurately and reliably complete tasks.Task Completion Time: This measures the total time taken to complete a task, from the initial request to the final action. It reflects the efficiency of the LAM and agent system.Object Accuracy: This measures the accuracy of selecting the correct UI element for each task step. It assesses the agents ability to interact with the appropriate UI components.Step Success Rate (SSR): This measures the percentage of individual steps completed successfully within a task. It provides a granular assessment of action execution accuracy.In online evaluations using Microsoft Word as the target application, LAM achieved a TSR of 71.0%, demonstrating competitive performance compared to baseline models like GPT-4o. Importantly, LAM exhibited superior efficiency, achieving the shortest task completion times and lowest average step latencies. These results underscore the efficacy of Microsofts framework in building LAMs that are not only accurate but also efficient in real-world applications.LimitationsDespite the advancements made, LAMs are still in their early stages of development. Key limitations and future research areas include:Safety Risks: The ability of LAMs to interact with the real world introduces potential safety concerns. Robust mechanisms are needed to ensure that LAMs operate safely and reliably, minimizing the risk of unintended consequences.Ethical Considerations: The development and deployment of LAMs raise ethical considerations, particularly regarding bias, fairness, and accountability. Future research needs to address these concerns to ensure responsible LAM development and deployment.Scalability and Adaptability: Scaling LAMs to new domains and tasks can be challenging due to the need for extensive data collection and training. Developing more efficient training methods and exploring techniques like transfer learning are crucial for enhancing the scalability and adaptability of LAMs.ConclusionMicrosofts framework for building LAMs represents a significant advancement in AI, enabling a shift from passive language understanding to active real-world engagement. The frameworks comprehensive approach, encompassing data collection, model training, agent integration, and rigorous evaluation, provides a robust foundation for building LAMs. While challenges remain, the transformative potential of LAMs in revolutionizing human-computer interaction and automating complex tasks is undeniable. Continued research and development efforts will pave the way for more sophisticated, reliable, and ethically sound LAM applications, bringing us closer to a future where AI seamlessly integrates with our lives, augmenting human capabilities and transforming our interaction with the world around us.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

In a rather unusual merging of two completely separate reported stories from last year, the former staff of Annapurna Interactive are seemingly preparing to take over the portfolio of shuttered indie label Private Division. At least some of the remaining Private Division employees are expected to be laid off in the process.This is according to a report from Bloomberg, which states that a currently unnamed company staffed by former Annapurna Interactive employees has reached a deal with private equity firm Haveli to take over the distribution of Private Division titles. Bloomberg reports and IGN can independently confirm that Haveli is the company that purchased Private Division from Take-Two Interactive last year for an undisclosed amount.The portfolio includes both current and upcoming Private Division titles, such as Tales of the Shire, which releases March 25. It also includes Project Bloom, a AAA action-adventure game developed by Game Freak that was announced back in 2023 with nothing more than a concept art teaser.Concept art for Game Freak's Project Bloom, formerly published by Private Division.As Bloomberg reports, Haveli's purchase of Private Division included not just the portfolio, but 20 employees who remained with the label following Take-Two-implemented layoffs last spring. Remaining employees have reportedly been told to explore other employment options, with the expectation that at least some of them will be laid off as part of the deal with the unnamed new publisher.Private Division was formerly Take-Two's publishing label, which the company founded back in 2017. It was intended to support independent games that were smaller than the fare typically supported by the Grand Theft Auto publisher. Over the years, Private Division produced titles such as The Outer Worlds, OlliOlli World, and Kerbal Space Program 2, but game sales repeatedly fell short of Take-Two expectations. Early last year, we reported that Take-Two was slowly shuttering operations at Private Division, first by winding down operations at its supported studios and then by selling off the label.The new owners of Private Division's portfolio are a group of former employees of Annapurna Interactive that collectively resigned last year following a leadership dispute at Annapurna. We reported last fall on the messy circumstances, which left Annapurna seeking to restaff an entire publishing team to cover its numerous obligations and roughly 25 individuals seeking new employment.Rebekah Valentine is a senior reporter for IGN. You can find her posting on BlueSky @duckvalentine.bsky.social. Got a story tip? Send it to rvalentine@ign.com.