The Tocopilla railway was a mountain railway built to serve the sodium nitrate mines in the Toco area of the Antofagasta Region in Chile. With a gauge of 3ft 6in (1,067mm), it ran from the port of Tocopilla on the Pacific coast up to a height of 4,902 feet (1,494 metres), with gradients up to 1 in 24. The railway was built by a joint-stock company founded in London and was designed by William Stirling of Lima, with a detailed description of the initial operation of the railway published by his brother Robert in 1900.The line was electrified in the mid-1920s and expanded in 1930 with the addition of lines serving new areas of mining. It continued operating into the 21st century, but was forced to close in 2015 when flash flooding caused numerous washouts on the electrified section of the railroad. With the declining prospects for nitrate, it was not economical for the line to be repaired. This photograph taken in 2013 shows a boxcab on the Tocopilla railway, leading a train down towards the coast.Photograph credit: David GublerRecently featured: Colias croceusMichael William BalfeLimeArchiveMore featured pictures

html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd"Originally posted on 3 July 2024. Scroll down for news of the Mansur-Rig 2.6 update.Former MPC Lead Rigging TD Assaf Ben Zuf has made Mansur-Rig, his modular toolset for building Maya character rigs, available for free.The software streamlines the process of building character rigs for animation, visual effects and games animation, and comes with an online library of free character rigs.Assaf Ben Zuf says that he intends to continue actively developing the software, and that he hopes that making it free for commercial projects will increase its user base.A flexible high-performance system for building character rigs in MayaFirst released in 2021, and updated regularly, Mansur-Rig is an advanced modular rig builder for Maya, based on a dedicated C++ node library.It can be used to build joint-based full-body and facial rigs for human and humanoid characters, although it isnt an auto-skinning tool.Key selling points cited in the online FAQs include the high performance of the rigs generated, its flexible, non-linear workflow, and its ability to rig non-T-posed models.Other notable features include a readymade animation picker UI and a LODs Tool for managing switches between Level of Detail states of a character.The software comes with a detailed user guide and video tutorials, plus an accompanying online library of character and vehicle rigs, also available free.Now free for commercial use, including for selling character rigsMansur-Rig is now free to download, although you will need to register an account on the product website to install it, as shown in the video above.The software is licensed for use on commercial projects, and you can even sell the rigs created with it, although users also need Mansur-Rig to use the rigs.However, animations created using Mansur-Rig can be exported from Maya in FBX format: for example, for use in game engines.Still in active developmentAccording to Assaf Ben Zuf, making the software free in no way means that development and support is [being discontinued].The hope is that this change will urge the community to use Mansur-Rig more, [so that] more bugs can be found and fixed, and new needs found and addressed.You can see features currently in development on the Mansur-Rig Trello board.Updated 10 January 2025: Assaf Ben Zuf has released Mansur-Rig 2.6.The release updates the toolsets Wing module to support bat as well as bird wings, and adds a new readymade peregrine falcon rig to the accompanying online library.It comes with controls for the wing and tail feathers, as well as the head, body and legs.The update also makes Mansur-Rig compatible with Maya 2025 on Windows, at least: Linux support is still in progress and removes its dependency on the PyMEL module.PyMEL used to be installed by default with Maya, but became a separate installation in Maya 2022+.System requirementsMansur-Rig 2.6 is compatible with Maya 2020+ on Windows and Maya 2020-2024 on Linux. The software is now free, including for use on commercial projects. You can also sell character rigs created with Mansur-Rig.Download modular Maya rig-building system Mansur-Rig for free(Registration required)Read the FAQs about Mansur-RigHave your say on this story by following CG Channel on Facebook, Instagram and X (formerly Twitter). As well as being able to comment on stories, followers of our social media accounts can see videos we dont post on the site itself, including making-ofs for the latest VFX movies, animations, games cinematics and motion graphics projects.

Previously, OpenAI partnered with robotics startup Figure to provide the models powering the latter's humanoid robots.Read More

Based on a new benchmark, Google DeepMind found Gemini 2.0 Flash to be the most factual LLM, with a score of 83.6%.Read More

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn MoreOrganizations interested in deploying AI agents must first fine-tune them, especially in workflows that often feel rote. While some organizations want agents that only perform one kind of task in one workflow, sometimes agents need to be brought into new environments with the hope that they adapt.Researchers from the Beijing University of Posts and Telecommunications have unveiled a new method, AgentRefine. It teaches agents to self-correct, leading to more generalized and adaptive AI agents.The researchers said that current tuning methods limit agents to the same tasks as their training dataset, or held-in tasks, and do not perform as well for held-out, or new environments. By following only the rules laid out through the training data, agents trained with these frameworks would have trouble learning from their mistakes and cannot be made into general agents and brought into to new workflows.To combat that limitation, AgentRefine aims to create more generalized agent training datasets that enable the model to learn from mistakes and fit into new workflows. In a new paper, the researchers said that AgentRefines goal is to develop generalized agent-tuning data and establish the correlation between agent generalization and self-refinement. If agents self-correct, they will not perpetuate any errors they learned and bring these same mistakes to other environments theyre deployed in.We find that agent-tuning on the self-refinement data enhances the agent to explore more viable actions while meeting bad situations, thereby resulting in better generalization to new agent environments, the researchers write.AI agent training inspired by D&DTaking their cue from the tabletop roleplaying game Dungeons & Dragons, the researchers created personas, scripts for the agent to follow and challenges. And yes, there is a Dungeon Master (DM).They divided data construction for AgentRefine into three areas: script generation, trajectory generation and verification.In script generation, the model creates a script, or guide, with information on the environment, tasks and actions personas can take.(The researchers tested AgentRefine using Llama-3-8B-Instruct, Llama-3-70B-Instruct, Mistral-7B-Instruct-v0.3, GPT-4o-mini and GPT-4o)The model then generates agent data that has errors and acts both as a DM and a player during the trajectory stage. It asses the actions it can take and then see if these contain errors. The last stage, verification, checks the script and trajectory, allowing for the potential of agents it trains to do self-correction.Better and more diverse task abilitiesThe researchers found that agents trained using the AgentRefine method and dataset performed better on diverse tasks and adapted to new scenarios. These agents self-correct more to redirect their actions and decision-making to avoid errors, and become more robust in the process.In particular, AgentRefine improved the performance of all the models to work on held-out tasks.Enterprises must make agents more task-adaptable so that they dont repeat only what theyve learned so they can become better decision-makers. Orchestrating agents not only direct traffic for multiple agents but also determine whether agents have completed tasks based on user requests.OpenAIs o3 offers program synthesis which could improve task adaptability. Other orchestration and training frameworks, like Magentic-One from Microsoft, sets actions for supervisor agents to learn when to move tasks to different agents.Daily insights on business use cases with VB DailyIf you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.Read our Privacy PolicyThanks for subscribing. Check out more VB newsletters here.An error occured.

Meta CEO Mark Zuckerberg thinks Apple [hasnt] really invented anything great in a while and that it has been coasting off of its past success. Steve Jobs invented the iPhone and now theyre just kind of sitting on it 20 years later, he said this week.Zuckerberg made the statements during a nearly three-hour long podcast with Joe Rogan where, along with discussing Metas moderation policy changes and turn against diversity and inclusion policies, they got into Metas beef with Apple and its policies.The conversation actually started with Rogans issues with Apple. Rogan said hes moving from Apple to Android in part because he doesnt like being attached to one company. He also isnt a fan of Apples App Store policies. The way they do that Apple store, where they charge people 30 percent, he said. That seems so insane that they can get away with doing that. I have some opinions about this, Zuckerberg said. While he gives credit to the iPhone as obviously one of the most important inventions probably of all time, he argued that Apple has put rules in place that feel arbitrary.Zuckerberg said that Apple has thoroughly hamstrung the ability for anyone else to build something that can connect to the iPhone in the same way as Apples own products, like the AirPods. If Apple let other people use its protocol, there would probably be much better competitors to AirPods out there, Zuckerberg said.RelatedNaturally, theres business behind Zuckerbergs gripes. Meta has had longstanding issues with Apple and the 30 percent cut it takes on some App Store transactions. Apples iOS restrictions have made it harder for Meta to compete on hardware and wiped out billions of dollars in advertising. Zuckerberg said that if Apples random rules didnt apply, Meta would make twice as much profit or something based on his back of the envelope calculation.Apple is increasingly under pressure to open up. Its made changes in the European Union in response to new laws targeting its policies, and its facing a lawsuit from the US Department of Justice for holding a monopoly over smartphones. But the company seems intent on maintaining its closed ecosystem until its forced to change.Zuckerberg believes that Apples reliance on just advantaging their stuff will ultimately hurt the company. Apple has been so off their game in terms of not really releasing many innovative things, he said. He said that the tech industry is super dynamic, and if you just dont do a good job for like 10 years, eventually, youre just going to get beat by someone. (Its easy to guess who Zuckerberg thinks that might be!)Apple did not immediately respond to a request for comment on Zuckerbergs remarks.There isnt a physical world and a digital world anymore.Zuckerberg touched on a lot of other tech topics as part of his conversation with Rogan, including AI and how he thinks about screen time with his daughter playing Minecraft. One area he spent some time on was neural interfaces and how physical and digital worlds will blend together.He thinks that its going to be a while before were really widely deploying anything that jacks into your brain, for example, and (naturally) he talked about the benefits of a wrist-based neural interface, which Meta is working on as part of its Orion augmented reality glasses.Down the line, Zuckerberg envisions a world where youll be able to use the neural interface wristband and the glasses to text a friend or an AI and have the glasses give you the answer. He also believes that as smart glasses or even contact lenses as a computing platform become more developed, the internet will be overlaid on the physical world.I think well basically be in this wild world where most of the world will be physical, but there will be this increasing amount of virtual objects or people who are beaming in or hologramming into different things to interact in different ways, he said.There isnt a physical world and a digital world anymore, he added. Were in 2025. Its one world.

Like Meta, Amazon is ending some of its diversity, equity, and inclusion (DEI) programs. In a memo sent last month, Candi Castleberry, Amazons VP of inclusive experiences and technology, said the company has been winding down outdated programs and materials related to its efforts around representation and inclusion, as reported earlier by CNBC and Bloomberg.In the memo, a copy of which Amazon provided to The Verge,Amazon spokesperson Brad Glasser declined to identify which programs had been ended.This approach where we move away from programs that were separate from our existing processes, and instead integrating our work into existing processes so they become durable is the evolution to built in and born inclusive,instead of bolted on, Castleberry wrote.News of the December memo comes just hours after Meta confirmed a report from Axios detailing the companys plans to eliminate its DEI program entirely, citing the legal and policy landscape in the US. Corporations outside of tech are also stepping away from DEI, with McDonalds and Walmart also scaling back programs.As noted by CNBC, Amazon also recently removed sections from its Our Positions page, which outlines the companys stance on various political and social issues. An archived version of the page from November 2024 shows two separate sections with the headings The inequitable treatment of Black people is unacceptable, and The rights of LGBTQ+ people must be protected.The current page removes both sections in favor of a single sentence reading We also believe that inequitable treatment of anyoneincluding Black people, LGBTQ+ people, Asians, women, and othersis unacceptable, and we advocate for policies designed to remove barriers to equity and create an inclusive environment for all employees.The company also replaced a section heading stating Diversity, equity, and inclusion are good for businessand more fundamentally, theyre simply right. with We are committed to creating a diverse and inclusive company that helps us build the best range of products and services for our broad customer base.We update this page from time to time to ensure that it reflects updates weve made to various programs and positions, Amazon spokesperson Kelly Nantel said in a statement to The Verge.Heres the December 16th memo in full:As we head toward the end of the year, I want to give another update on the work weve been doing around representation and inclusion.As a large, global company that operates in different countries and industries, we serve hundreds of millions of customers from a range of backgrounds and globally diverse communities.To serve them effectively, we need millions of employees and partners that reflect our customers and communities. We strive to be representative of those customers and build a culture thats inclusive for everyone.In the last few years we took a new approach, reviewing hundreds of programs across the company, using science to evaluate their effectiveness, impact, and ROI identifying the ones we believed should continue. Each one of these addresses a specific disparity, and is designed to end when that disparity is eliminated. In parallel, we worked to unify employee groups together under one umbrella, and build programs that are open to all. Rather than have individual groups build programs, we are focusing on programs with proven outcomes and we also aim to foster a more truly inclusive culture. You can read more about this on ourTogether at Amazon page on A to Z.This approach where we move away from programs that were separate from our existing processes, and instead integrating our work into existing processes so they become durable is the evolution to built in and born inclusive,instead of bolted on. As part of this evolution, weve been winding down outdated programs and materials, and were aiming to complete that by the end of 2024. We also know there will always be individuals or teams who continue to do well-intentioned things that dont align with our company-wide approach, and we might not always see those right away. But well keep at it.Well continue to share ongoing updates, and appreciate your hard work in driving this progress. We believe this is important work, so well keep investing in programs that help us reflect those audiences, help employees grow, thrive, and connect, and we remain dedicated to delivering inclusive experiences for customers, employees, and communities around the world.

Automattic, the company that runs WordPress.com, is scaling back its contributions to the WordPress open-source project, according to an announcement on Thursday. The company says its decreasing contributions to match the amount of time companies like WP Engine spend on the ecosystem, further escalating the tension between Automattic CEO Matt Mullenweg and the community.Now, instead of spending 3,988 hours per week developing the WordPress project, Automattic says it will now contribute around 45 hours as part of Five for the Future a program that encourages companies to give back five percent of their resources to WordPress.org. These hours will likely go towards security and critical updates, Automattic says. Mullenweg, who also co-founded WordPress, criticized the third-party host WP Engine for contributing 40 hours a week to the ecosystem and called it a cancer to the community. On the Five for the Future page that tracks contributions, the number of hours contributed by Automattic is already dwindling. RelatedAutomattic blames the cutback on the significant time and money related to the ongoing legal battle with WP Engine. It also points to the intense criticism it has faced from members of the community who want Matt and others to step away from the WordPress project:Weve made the decision to reallocate resources due to the lawsuits from WP Engine. This legal action diverts significant time and energy that could otherwise be directed toward supporting WordPresss growth and health. We remain hopeful that WP Engine will reconsider this legal attack, allowing us to refocus our efforts on contributions that benefit the broader WordPress ecosystem.WP Engine sued Automattic and Mullenweg last year after the co-founder waged a public campaign against the company and took over its ACF plugin. A judge later granted a preliminary injunction in favor of WP Engine, saying Mullenwegs conduct is designed to induce breach or disruption. A number of employees also left Automattic last year after the company offered to buy out staff who didnt agree with its fight against WP Engine. The company also shuttered its sustainability team this week, with a screenshotted Slack message from Mullenweg saying, its probably a good time to officially dissolve the team entirely, adding that it doesnt seem like creating a team around this was able to further any of its goals. The move has sparked criticism from the community, including journalist Kara Swisher.

Adopting advanced AI technologies, including Multi-Agent Systems (MAS) powered by LLMs, presents significant challenges for organizations due to high technical complexity and implementation costs. No-Code platforms have emerged as a promising solution, enabling the development of AI systems without requiring programming expertise. These platforms lower barriers to AI adoption, allowing even non-technical users to leverage AI tools efficiently. By 2025, nearly 70% of applications are projected to utilize Low-Code or No-Code platforms, showcasing their growing role in democratizing AI technologies. Additionally, LLMs have proven transformative in various applications, including generative AI, which creates new content like text, images, and videos, and multimodal AI, which integrates diverse data forms for tasks such as image recognition and cross-modal retrieval.The development of LLM-based MAS has further advanced AIs capabilities by enabling multiple autonomous agents to collaborate on complex tasks through natural language interactions. These systems integrate specialized agents that process data from different modalities, manage temporal and spatial relationships, and coordinate task allocation. Adopting multimodal learning techniques, such as embedding spaces and cross-attention mechanisms, enhances understanding of diverse data types, enabling tasks like image-to-text transformation and cross-modal search. These advancements make AI systems more flexible, efficient, and accessible, driving innovation in enterprise environments while addressing implementation challenges.Researchers from SAMSUNG SDS, Seoul, developed a multimodal LLM-based MAS using No-Code platforms to simplify AI integration into business processes without requiring professional developers. The system, built using tools like Flowise, integrates Multimodal LLMs, image generation with Stable Diffusion, and RAG-based MAS. Evaluated through use cases like image-based code generation and Q&A systems, it highlights collaborative agent synergies. The study emphasizes technical implementation, business applicability, and performance evaluation, showcasing improved efficiency and accessibility for non-experts and SMEs. The research offers a scalable methodology for AI adoption, reducing manual tasks and advancing the practical use of MAS across industries.Implementing a multimodal LLM-based MAS using the Flowise platform involves setting it up in the cloud, securely managing API keys, and integrating external services like OpenAI and Stable Diffusion. A hybrid relational and NoSQL database system efficiently handles structured and unstructured data. Agents for Image Analysis, RAG Search, Image Generation, and Video Generation process input types, such as text, images, and audio, to produce corresponding outputs like text, photos, and videos. These agents are integrated into a unified workflow with a web-based user interface for seamless functionality and real-time input processing.The study discusses the implementation and results of a multimodal MAS, focusing on various use cases like image analysis, code generation, RAG-based search, image generation, and video generation. The system processes incomplete code images, generates code through agent collaboration, and reviews it for quality. RAG search agents retrieve answers from RAG knowledge and external sources when needed. The image-generation agents create visuals from text descriptions or sketches, while the video-generation agents produce videos based on textual or image inputs. Integrating these agents into a unified system enables seamless user interaction and execution of tasks.In conclusion, The study presents a multimodal LLM-based MAS built using a No-Code platform, Flowise, to simplify AI adoption in enterprises. It demonstrates the systems effectiveness in automating tasks like code generation, image and video creation, and RAG-based query responses, reducing the need for specialized development teams. The research highlights the practical benefits of AI in business, such as improving efficiency and content generation. It also offers a novel methodology for integrating multimodal data with No-Code platforms, though it acknowledges limitations in customization, data handling, and agent communication that require further refinement.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. [Recommended Read] Nebius AI Studio expands with vision models, new language models, embeddings and LoRA (Promoted)

Paramount is offering an extended one-month free trial to its Paramount+ with Showtime subscription by clicking through this link (no coupon required). This is the premium subscription with no ads and normally costs $12.99 per month. Best of all, this promotion is available to both new and previous subscribers. Now is a good time to sign up because you can watch the NFL playoff games this weekend or catch up on the recently released Transformers One movie.Paramount+ with Showtime (30-Day Trial)Extended 1 Month Free Trial to Paramount+ with Showtime Annual PlanParamount+ (formerly CBS All Access) is one of the better populated streaming services available, especially since this plan is ad-free. Its extensive library of media has been bolstered with the addition of Showtime in 2023, making it a great service for those who enjoy movies, original shows, live sports, news, and more.Watch Two NFL Wild Card Games This WeekendWhat a weekend for football fans. There are a total of six wild card games playing this weekend, and CBS and Paramount+ have rights to stream two of them: the Los Angeles Chargers vs. Houston Texans on January 11 at 4:30pm EST and the Denver Broncos vs. Buffalo Bills on January 12 at 1pm EST. Watch both games for free with this trial.Watch Transformers OneThe recently released animated Transformers One movie is now available to stream on Paramount+. Transformers One is an origin story that focuses on the relationship between Optimus and Megatron and the events leading up to their rift. It follows a separate canon from the Michael Bay films and is an excellent movie for kids and adults alike (although Tom Jorgensen doesn't necessarily agree in his IGN Transformers One review). It's definitely worth a watch.Looking to try out other streaming services before you decide to pay? We've got you covered. Check out these steaming services that are currently offering free trials.Why Should You Trust IGN's Deals Team?IGN's deals team has a combined 30+ years of experience finding the best discounts in gaming, tech, and just about every other category. We don't try to trick our readers into buying things they don't need at prices that aren't worth buying something at. Our ultimate goal is to surface the best ossible deals from brands we trust and our editorial team has personal experience with. You can check out our deals standards here for more information on our process, or keep up with the latest deals we find on IGN's Deals account on Twitter.Eric Song is the IGN commerce manager in charge of finding the best gaming and tech deals every day. When Eric isn't hunting for deals for other people at work, he's hunting for deals for himself during his free time.