Twitch watch time broadly stable in 2024 at 18.9bn hours"Expect 2025 to be a big year for multistreaming as the tools evolve and become more accessible" News by Vikki Blake Contributor Published on Dec. 19, 2024 Watch time on Twitch has remained stable year-over-year, with 18.9bn hours watched in 2023 compared to 18.5bn this year.StreamElements' State of the Stream 2024 yearly summary - compiled from data courtesy of Rainmaker.gg - suggests possible reasons for the "lack of growth are creators migrating to other platforms or merely splitting their time with the growing popularity of multistreaming in order to broaden their fanbase."Daily hours watched in Q4 are also consistent across the year, with the Q2 dip in hours watched extending throughout the rest of the year.Interestingly, Twitch's top 9 games remain essentially unchanged over the last five years, with Grand Theft Auto 5, League of Legends, and Valorant being the first, second, and third most popular games, respectively. CSGO, Fortnite, Dota 2, Minecraft, World of Warcraft, and Call of Duty: Warzone round out the rest of the top 10, although the platform's biggest category remains Just Chatting, with more than twice the hours watched to second-place GTA 5, presumably due to KaiCenat's subathon.Despite this, hours watched for League of Legends are down 34% in November, with GTA 5 and Dota 2 dropping 10% and 7%, too. World of Warcraft is up 84% in November 2024, though, and Fortnite up 38%.KaiCenat - described as a "powerhouse" on the monthly and yearly charts - clocked up 184m hours watched across the year, over twice as much as second place ibai, although StreamElements warns that shouldn't "overshadow the impressive numbers put up by the other chart members," with Jynxzi, Caudrel, caseoh_, and Papaplatte all cracking the top 10 of the year for the first time.Here's StreamElement's full breakdown."Twitch has shuffled through several top streamers over the years. There was the Ninja era, then xQc's reign, and now we are in the KaiCenat age," said Or Perry, CEO of StreamElements."The driver behind the changing of the guards is platform diversification. Ninja began multistreaming while xQc is leveraging different platforms for unique content, both of which have remained successful. It's only a matter of time before KaiCenat tests the water."This is because being everywhere that gamers are is the best way to grow, whether you are using some platforms for discovery and others for community building. Expect 2025 to be a big year for multistreaming as the tools evolve and become more accessible to the broader creator ecosystem."Twitchs monthly hours watched "rebounded" back in September, hitting 1.653bn hours watched, up from 1.568bn in June 2024.

Australian games sector employment "steady" as industry generates $211.9m in 2024Plus 81% of Australian studios expect headcount to remain stable or grow in 2025 News by Vikki Blake Contributor Published on Dec. 19, 2024 The Australian video game industry generated $339.1m AUD ($211.9m USD) in revenue in 2024.Despite a "challenging" year for game development globally, Australia's full-time employment was "steady" at 2465 developers, with 61% of studios planning to hire next year. 81% of respondents also said they expected their studio's headcount to either remain stable or grow in 2025.In its Australian Game Development Industry Snapshot report, the Interactive Games & Entertainment Association (IGEA) revealed that of the 137 studios that responded, 85% developed games based on their own IP, and 93% of their revenue was generated from outside of Australia.The survey also found that Victoria "continues to be a popular hub for game development," with 52% of game studio head offices located in the state. Employee distribution illustrated that many studios "have a presence in more than one state and tend to follow talent and incentives across the country," however, 82% of workers are located on the east coast (36% Victoria, 27% Queensland, 18% New South Wales, 1% ACT)."This year has shown the local games sector continues to build capability in the face of global headwinds. 93% of the revenue generated in Australia comes from overseas sources, highlighting the popularity of Australian-made games to a global audience," said Ron Curry, CEO of IGEA."Generous and globally competitive tax rebates plus direct funding from the federal and state governments continues to ensure Australian game developers can deliver amazing game projects. 25% of our survey respondents accessed the Digital Games Tax Offset last year, and just under 40% intend to do so this financial year. We must maintain this level of support to ensure the local industry continues to grow and contribute to Australia's creative and digital exports."The full report is downloadable from IGEA's website.

Starting in 2025, Elder Scrolls Online will go from releasing annual expansions (or "Chapters") to a more seasonal model.In a recent blog from ZeniMax Online's studio director Matt Firor, he explained the Seasons are expected to run for "three or six" months, and will feature "a mix of themed story content,events, store items, dungeons and more." According to him, the move from yearly Chapters is easier on the studio.Similar live-service games like Destiny 2 and Apex Legends have comparable three-month seasons. ZeniMax Online will gradually move toward making "smaller bite-sized content," in addition to "some larger items" that will release throughout the year."We'll be able to have teams launch content when it is ready throughout the year and not work to a date in Junethis will let us focus on a greater variety of content spread over the year," he said. "This supports the new Seasons model, and will enable us to release content, updates, fixes, and systems in a more efficient manner."Live-service is getting a refreshEarlier this year, Bungie outlined a series of changes to its own content model for Destiny 2. Like with Elder Scrolls Online, these changes will hit in 2025 and see the sci-fi shooter "explore and innovate" through medium-sized content drops rather than larger expansions in the vein of Destiny 2: The Final Shape.The first of these is Codename: Apollo, which will be split into a pair of paid updates that launch in the summer and fall. Codename: Behemoth will then follow Apollo with updates in winter 2025 and spring 2026. There will also be two free, individual seasons per update.According to Firor, ZeniMax is looking at doing quest content in Elder Scrolls Online in "different ways" with its upcoming model. Some of the plans he outlined included using existing zones and areas to tell new stories (or continue already existing ones), split large zones across multiple updates, and provide the team with "the space to find a new cadence that can mix new content and systems with addressing feedback, fixing issues, and adding new quests and activities to existing zones and stories."Firor's full outlook on Elder Scrolls Online and what's to come can be read here.

Two Tencent directors have departed from Epic Games' board of directors, according to the U.S. Justice Department. In a statement, the department explained these exits came about in light of concerns about their positions on the board being anti-competitive and in violation of Section 8 of the Clayton Act.Similarly, Tencent has amended a part of its shareholder agreement with Epic and "relinquished its unilateral right to appoint directors or observers to the Epic board in the future."As the statement explains, Section 8 stops directors and officers from "serving simultaneously on the board of competitors." Tencent has had a minority stake in Epic since 2012, and is the parent company of League of Legends creator Riot Games. The two studios are major players in the industry, and Tencent's investment in both previously drew scrutiny from the U.S.' Committee of Foreign Investment.The Department's inspection into Tencent and Epic's relationship "is the latest of the divisions ongoing Section 8 enforcement efforts, which to date have unwound or prevented interlocks involving at least two dozen companies.""Scrutiny around interlocking directorates continues to be an enforcement priority for the Antitrust Division," said the Antitrust Division's deputy director Miriam Vishio. "Due to the hard work of our tremendous staff, our increased enforcement around Section 8 over the last few years has achieved substantial results and become part of our fabric."Epic Games, epic legal battlesIn recent years, Epic Games has been subject to scrutiny by the U.S. government. Earlier this month, the FTC started sending refunds to Fortnite players who had been tricked into making unwanted purchases between 2017 and 2018.The entire saga first began in December 2022. At the time, the FTC and Epic reached a $520 million settlement after the latter had been accused of using various "design tricks" to "dupe millions of players into making unintentional purchases."Epic was also ordered to give $245 million of that settlement back to customers, and the FTC began the refund process in September 2023. The amount of refunds reportedly total over $72 million.The Fortnite maker has also actively used the country's court system to settle its affairs, namely with Apple. Its legal clash with the tech company led to a court ruling that Apple would lift some restrictions for third-party payments on its App Store. Last year, Apple successfully argued to the Ninth Circuit that it had not engaged in antitrust practices against Epic.Months later, Epic appealed that new ruling to the Supreme Court, which ultimately rejected both companies' appeals back in January.This story has been updated to clarify the Justice Department's concerns about Epic and Tencent were not a full-on investigation, and Epic's $245 million fine was already baked into its initial FTC settlement.

PlayStation maker Sony is set to become the largest shareholder in FromSoftware parent company Kadokawa Corporation.It was previously reported that Sony was in talks to purchase Kadokawa, which confirmed it received an acquisition offer from an undisclosed party in November.Now, as noted in a press release, Sony has signed a strategic capital and business alliance agreement with Kadokawa that will see it acquire 12,054,110 new shares in the company for approximately 50 billion yen ($318.5 million).Sony will hold 10 percent of Kadokawa's shares following the deal. That total includes the shares Sony purchased in 2021. Sony also owns a 14.09 percent stake in FromSoftware.Both companies said the move will maximize IP value globally and facilitate wider and deeper collaboration, including potential joint investments in the content field."In the future, the two companies plan to discuss specific initiatives for collaboration, such as initiatives to adapt Kadokawa's IP into live-action films and TV dramas globally, co-produce anime works, expand global distribution of Kadokawa's anime works through the Sony Group, further expand publishing of Kadokawa's games, and develop human resources to promote and expand virtual production," adds the press release.Back in November, Kadokawa outlined plans to expand its console and PC game publishing business, and said it was "developing a structure" to bolster its efforts in that arena.Elden Ring's significant contributionsThe company noted that Elden Ring continues to drive "high growth" within its gaming business. The title has sold over 25 million copies worldwide since launching in 2022 and has received additional content in the form of the sprawling Shadow of the Erdtree expansion.FromSoftware recently announced a multiplayer Elden Ring spin-off called Elden Ring Nightreign at The Game Awards 2024.Discussing its future in the wake of the Sony deal, Kadokawa CEO Takeshi Natsuno said the company hopes to bring its franchises to more users around the world."We are confident that this will greatly contribute to maximizing the value of our IP and increasing our corporate value in the mid to long-term," he added. "We intend to do our utmost to ensure that our collaborative efforts with Sony produce great results in the global market."

Bluesky now has a specific tab for mentions in your notifications as part of the apps just-released 1.96 update. With the mentions tab, its much easier to see your replies or conversations youve been tagged in on the platform.Speaking of replies, update 1.96 lets you easily access settings that let you control how replies on posts appear to you. Replies can be linear, meaning they show up one post after another, or threaded, which means they will appear in indented threads (kind of like how they appear on Reddit). You can also sort replies by newest, oldest, most-liked, hot, and random (which Bluesky also calls Posters Roulette).If you choose to set a custom domain as your username, with 1.96, Bluesky will also reserve your old .bsky.social name so that it cant be picked up by someone else. I wish this feature had been available when I set my custom domain when I did that, I made an alt account thats parked on my old .bsky.social name so that it doesnt get taken.Bluesky saw a surge of new users in November, and although growth has slowed as of late, the platform surpassed 25 million total users last week. The company plans to launch a subscription service early next year and, at some point, its own payment platform.And the platform has some big competition from Metas Threads, which seems to be doing everything it can to remind people that it can ship lots of features and that its much larger than Bluesky.

A Google Street View image played in role in solving a missing person investigation in Spain, according to reports from the BBC and The New York Times. The image, which can still be seen on Street View, shows what appears to be a person loading a large object into the trunk of a car, as Gizmodo points out. Another image showed someone transporting a large white bundle in a wheelbarrow, the BBC reports. Part of the human remains of the missing person have been found buried in a cemetery in a town in the province of Soria and advanced technical means had to be used to locate them.One of the clues that investigators had to solve the crime, although it was not decisive, was some images that they detected during the investigations in a location search application.This is the Google Street View image allegedly linked to the crime. Screenshot: GoogleThough the press release doesnt explicitly name the location application it used, a spokesperson for the Spanish National Police confirmed to The New York Times that it was Google Maps while also saying, The image was not the key to solving the case.Authorities were investigating the case of a 33-year-old man who disappeared last year while visiting a partner in Tajueco, Soria, according to the Spanish outlet El Pas. Police arrested a woman and her ex-partner in connection with the crime in November.Google Maps has been credited with helping investigators in the past, with Italian authorities using Street View to trace the location of a mafia fugitive in 2022.

There is a new version of GitHub CoPilot that is completely free to use, no credit card or subscription required. The only requirement is a GitHub account and a compatible IDE ( Visual Studio, Visual Studio Code and JetBrains products as well as GitHub.com currently ). Details of the new free tier from the Visual Studio Code blog.Were excited to announce an all newfree plan for GitHub Copilot, available for everyone today in VS Code. All you need is a GitHub account. No trial. No subscription. No credit card required.With GitHub Copilot Free you get2000 code completions/month. Thats about 80 per working day which is a lot. You also get50 chat requests/month, as well asaccess to both GPT-4o and Claude 3.5 Sonnet models.If you hit these limits, ideally its because Copilot is doing its job well, which is to help you do yours! If you find you need more Copilot, thepaid Pro planis unlimited and provides access to additional models likeo1andGemini(coming in the new year).With this announcement, GitHub Copilot becomes a core part of the VS Code experience. The team has been hard at work, as always, improving that experience with brand new AI features and capabilities. Lets take a look at some of the newer additions to GitHub Copilot that dropped in just the past few months. This is your editor, redefined with AI.Key LinksGitHub Copilot Free AnnouncementGitHub Copilot HomepageYou can learn more about the new GitHub Copilot free tier in the video below.

The evaluation of LLMs in medical tasks has traditionally relied on multiple-choice question benchmarks. However, these benchmarks are limited in scope, often yielding saturated results with repeated high performance from LLMs, and do not accurately reflect real-world clinical scenarios. Clinical reasoning, the cognitive process physicians use to analyze and synthesize medical data for diagnosis and treatment, is a more meaningful benchmark for assessing model performance. Recent LLMs have demonstrated the potential to outperform clinicians in routine and complex diagnostic tasks, surpassing earlier AI-based diagnostic tools that utilized regression models, Bayesian approaches, and rule-based systems.Advances in LLMs, including foundation models, have significantly outperformed medical professionals in diagnostic benchmarks, with strategies such as CoT prompting further enhancing their reasoning abilities. OpenAIs o1-preview model, introduced in September 2024, integrates a native CoT mechanism, enabling more deliberate reasoning during complex problem-solving tasks. This model has outperformed GPT-4 in addressing intricate challenges like informatics and medicine. Despite these advancements, multiple-choice benchmarks fail to capture the complexity of clinical decision-making, as they often enable models to leverage semantic patterns rather than genuine reasoning. Real-world clinical practice demands dynamic, multi-step reasoning, where models must continuously process and integrate diverse data sources, refine differential diagnoses, and make critical decisions under uncertainty.Researchers from leading institutions, including Beth Israel Deaconess Medical Center, Stanford University, and Harvard Medical School, conducted a study to evaluate OpenAIs o1-preview model, designed to enhance reasoning through chain-of-thought processes. The model was tested on five tasks: differential diagnosis generation, reasoning explanation, triage diagnosis, probabilistic reasoning, and management reasoning. Expert physicians assessed the models outputs using validated metrics and compared them to prior LLMs and human benchmarks. Results showed significant improvements in diagnostic and management reasoning but no advancements in probabilistic reasoning or triage. The study underscores the need for robust benchmarks and real-world trials to evaluate LLM capabilities in clinical settings.The study evaluated OpenAIs o1-preview model using diverse medical diagnostic cases, including NEJM Clinicopathologic Conference (CPC) cases, NEJM Healer cases, Grey Matters management cases, landmark diagnostic cases, and probabilistic reasoning tasks. Outcomes focused on differential diagnosis quality, testing plans, clinical reasoning documentation, and identifying critical diagnoses. Physicians assessed scores using validated metrics like Bond Scores, R-IDEA, and normalized rubrics. The models performance was compared to historical GPT-4 controls, human benchmarks, and augmented resources. Statistical analyses, including McNemars test and mixed-effects models, were conducted in R. Results highlighted o1-previews strengths in reasoning but identified areas like probabilistic reasoning needing improvement.The study evaluated o1-previews diagnostic capabilities using New England Journal of Medicine (NEJM) cases and benchmarked it against GPT-4 and physicians. o1-preview correctly included the diagnosis in 78.3% of NEJM cases, outperforming GPT-4 (88.6% vs. 72.9%). It achieved high test-selection accuracy (87.5%) and scored perfectly on clinical reasoning (R-IDEA) for 78/80 NEJM Healer cases, surpassing GPT-4 and physicians. In management vignettes, o1-preview outperformed GPT-4 and physicians by over 40%. It achieved a median score of 97% for landmark diagnostic cases, comparable to GPT-4 but higher than physicians. Probabilistic reasoning was performed similarly to GPT -4, with better accuracy in coronary stress tests.In conclusion, The o1-preview model demonstrated superior performance in medical reasoning across five experiments, surpassing GPT-4 and human baselines in tasks like differential diagnosis, diagnostic reasoning, and management decisions. However, it showed no significant improvement over GPT-4 in probabilistic reasoning or critical diagnosis identification. These highlight the potential of LLMs in clinical decision support, though real-world trials are necessary to validate their integration into patient care. Current benchmarks, like NEJM CPCs, are nearing saturation, prompting the need for more realistic, challenging evaluations. Limitations include verbosity, lack of human-computer interaction studies, and a focus on internal medicine, underscoring the need for broader assessments.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)

NewsTransforming Biology with Generative AI: Unveiling GenBio AIs State-of-the-art Multiscale Models 1 like December 19, 2024Share this postLast Updated on December 19, 2024 by Editorial TeamTL;DR: GenBio AI is advancing biology with Generative AI by developing AI-Driven Digital Organisms (AIDO). The AIDO system integrates multiscale foundation models for DNA, RNA, proteins, and cellular systems, allowing researchers to simulate, predict, and program biological outcomes from molecular to systemic levels. These tools aim to transform drug discovery, disease understanding, and personalized medicine, setting the stage for a new era in biological research.Advancing Biology with Generative AI: Inside GenBio AIs AI-Driven Digital OrganismBiology is entering an era where artificial intelligence is redefining the way we approach research and discovery. Leading this transformation is GenBio AI with its groundbreakingAI-Driven Digital Organism (AIDO), an integrated system of multiscale foundation models that enables researchers to simulate, program, and predict complex biological outcomes. AIDO addresses critical challenges in medicine, biotechnology, and life sciences by unifying insights across molecular, cellular, and systemic levels.Professor Eric Xing, Co-Founder and Chief Scientist of GenBio AI, underscores the ambition behind AIDO:GenBio will usher in a new era of medical and life sciencethrough a paradigm shift powered by next-generation Generative AI technology beyond what has already brought us disruptive results such as ChatGPT. Our transformative technology allows biological data of all types and scales to be utilized to distill holistic and comprehensive knowledge of how living systems work. Therefore, multiscale biological complexities are no longer barriers but opportunities for breakthrough insights.Moving Beyond Silos with AIDOTraditional biological models often operate in isolation, analyzing narrow datasets like DNA or proteins without integrating broader system interactions. AIDO disrupts this approach by creating a cohesive framework where modular models interact seamlessly, enabling a comprehensive understanding of biology as an interconnected system.Key Features of AIDO:Multitasking Efficiency: Handles up to 300 tasks simultaneously, surpassing the one or two tasks most current systems manage.Interoperable Modules: Models for DNA, RNA, proteins, single cells, and evolutionary data work in concert, addressing the siloed nature of traditional approaches.Comprehensive Data Utilization: Incorporates diverse biological data types, from sequences to structures, providing unprecedented insight into complex systems.By bridging biological scales, AIDO equips researchers with tools to analyze interactions across molecular, cellular, and organismal levels.Breaking Down the AIDO Foundation ModelsGenBio AIs first phase of AIDO introduces six foundational models, each designed to tackle specific biological challenges:AIDO-DNA: A 7-billion-parameter model trained on data from 796 species, offering advanced insights into genomic structure and function.AIDO-RNA: The largest model of its kind with 1.6 billion parameters, tailored for RNA structure prediction, genetic regulation, and vaccine design.AIDO-Protein: A computationally efficient model that facilitates exploration of protein functionality, essential for drug discovery.AIDO-Single Cell: Processes entire human transcriptomes without truncation, uncovering complex cellular dynamics with precision.Protein Structure Model: Focuses on three-dimensional protein modeling, uncovering relationships between structure and biological activity.Evolutionary Information Model: Provides insights into molecular evolution, connecting genetic data across species.These models not only excel individually but also operate as an integrated system, making AIDO a comprehensive toolkit for biological research. You can download them on GitHub or Hugging Face.Transformative Applications of AIDOAIDOs real-world applications are poised to address some of the most pressing challenges in medicine and biotechnology:Accelerating Drug DiscoveryTraditional drug development is costly and time-intensive, often with high failure rates. AIDO allows researchers to simulate and test millions of potential compounds in hours, drastically reducing both time and costs.Advancing Personalized MedicineAdverse drug reactions remain a leading cause of mortality worldwide. By creating digital patient twins, AIDO supports the design of personalized treatments that reduce risks and improve therapeutic outcomes.Understanding Complex DiseasesFrom cancer to neurodegenerative disorders, many diseases involve systemic interactions. AIDOs multiscale approach equips researchers to study these mechanisms and identify new pathways for intervention.Source: GenBio AIGlobal Expertise, Global ImpactGenBio AIs achievements are the result of a collaborative effort among world-renowned scientists and institutions. Headquartered in Palo Alto, with labs in Paris and Abu Dhabi, the companys team includes experts from Carnegie Mellon University, Stanford, the Weizmann Institute of Science, and MBZUAI. These partnerships have resulted in six peer-reviewed papers presented at NeurIPS, showcasing the rigorous research behind AIDO.Professor Eran Segal of the Weizmann Institute of Science highlights the significance of this work:GenBio AIs six multiscale foundation models are a leap forward in our ability to understand and predict biological phenomena. We now have the capacity to uncover systemic insights into how organisms function. This is transformative for genomics research, where the ability to simulate and program at multiple scales opens new avenues for precision medicine and disease intervention.Professor Fabian Theis of Helmholtz Munich adds:GenBio AIs achievement in creating scalable state-of-the-art models on multiple scales is a game-changer. This technology not only accelerates our ability to explore cellular dynamics but also bridges the gap between molecular and systems biology, unlocking unprecedented opportunities for disease modeling and therapeutic innovation.Explore the Research:The Road AheadThe development of AIDO represents just the beginning of GenBio AIs roadmap. The company envisions deeper integration between foundational models in future phases, expanding the systems utility for synthetic biology, environmental sustainability, and longevity research.Dr. Le Song, Co-Founder and CTO of GenBio AI, encapsulates the vision:What we have built is revolutionary because our integrated system will use these state-of-the-art models to create interactive digital versions of biological systems that can be safely experimented on and precisely modified. This technology lets us program biology the way we program computers, opening up possibilities weve never had before in medicine and biotechnology.As AIDO evolves, it promises to reshape how we approach biological research, offering scientists the tools to address complex challenges with precision and efficiency. For researchers working in genomics, drug development, or systems biology, AIDO provides a unified platform to tackle the most ambitious questions in life sciences.GenBio AI is setting the stage for a future where biology is not just observed but actively designed and improved.Source: GenBio AI Releases Phase 1 of Worlds First Digital Organism to Transform Medical ResearchShare this post