A new report about Spotify's plans for a more expensive premium tier claims that Apple is considering doing the same, but Apple Music has no reason to ever copy its streaming rival.Apple MusicThe days of Spotify being the most popular streaming music service in the world and yet constantly losing money, may be over. The days of Spotify trying to get the EU to force Apple into giving it an unreasonably better deal, may never change.But a new report says that amongst Spotify's plans for the near future, there is the likelihood that it will introduce a new and more expensive subscription tier. Bloomberg says the new tier would, for a start, offer better quality audio which Apple Music already does as standard. Continue Reading on AppleInsider | Discuss on our Forums

On this episode of the HomeKit Insider Podcast, we break down the latest in smart home news such as more rumors on Apple's smart display, new devices, and smart robots.HomeKit Insider PodcastWe have even more rumors on Apple's smart home display this past week. One report reiterated a possible timeline, putting a launch at April at the earliest.Ming Chi Kuo also gave his analysis on the robot lamp that was demoed from Apple's machine learning group. Based on supply chain, the personal robot is still in the proof of concept stage and won't start production until 2028 at the earliest. Continue Reading on AppleInsider | Discuss on our Forums

Say what you will about the abundance of Souls-like titles over the years, but they continue to prove popular. The future looks especially promising, whether its high-profile efforts from Hexworks, FromSoftware, NEOWIZ, or relatively new titles from Nexon and Dark Star. Whether its an open world, a linear experience or a multiplayer-focused approach, theres something for everyone. Here are 15 new Souls-like games to watch out for in 2025 and beyond.Elden Ring: NightreignNew might sound like a misnomer for Nightreign, a spin-off of FromSoftwares massive open-world action RPG into a co-op experience. Nevertheless, its something truly unprecedented from the studio one or three players survive through multiple days, nights and bosses (including some from Dark Souls) to take on the Night Lord. Select a class, scope out areas with valuable loot, face the night, and try to survive. Its out later this year for Xbox Series X/S, Xbox One, PS4, PS5, and PC, and were intrigued to see how it fares.

Announced back in 2023, Exoborne is an upcoming extraction shooter by Sharkmob. Set in a parallel alternate timeline, Exoborne has players take part in a mysterious conflict. Its unique take on the extraction shooter genre, complete with dynamic nature and the constantly changing world made us ask the developers some questions. Narrative director and studio co-founder Martin Hultberg got back to us with quite a few interesting answers.With there being quite a few extraction shooters coming out these days, how does Exoborne seek to separate itself from the pack?We love shooters. Especially high-stakes variants, like extraction shooters. Some of us even worked on one of the very first games with an extraction mode, Tom Clancys the Division. What I think makes Exoborne stand out is a mix of things really not one single thing alone. The combination of dynamic forces of nature that affects game-play, fluid traversal and the Exo-Rigs create a unique experience within the high-risk / high-reward shooter segment. There is nothing quite like it out there as far as I know.Forces of nature, like storms and tornadoes, exist to a degree in other PvP games. None have integrated it to the same level we have however. They really help shape and evolve a session, forcing you to adapt and strategize on the fly. No plan survives contact with the enemy has never been as true as when a firenado barrels down on your extraction point!The traversal is something we played around with in Bloodhunt, Sharkmobs first game. We have honed that experience and tweaked it, taking many learnings with us to make sure it becomes a fluid traversal experience both in the vertical and horizontal space. That allows you to flank, take advantage of wind, fly long distances using the glider, grapple up cliff walls or rooftops etc. Since the exo-rigs prevent you from taking fall damage you can really play around with it and allow yourself to pull off crazy moves.And then the Exo-Rigs, the literal backbone of any Exoborne warrior. It is a mobility, weapon and abilities platform allowing you to create the playstyle you want. By modding it, combining it with special abilities and bringing the guns of your choosing you can either adapt it to excel under certain conditions, be a jack of all trades or fill a specific role in your team of three."It is a mobility, weapon and abilities platform allowing you to create the playstyle you want."How often can players expect to run into other players in a general outing in Exoborne?In Exoborne, player encounters are designed to be dynamic and unpredictable, influenced by the games unique features. The pacing of these encounters is crucial, and its something we continuously monitor in all our past and upcoming playtests. Its difficult to give an exact frequency, as this depends on many factors. For example, players might choose to stick to the outskirts of the map to scavenge or deliberately avoid high-risk areas during a particular run. Even the Forces of Nature can impact how often players run into each other. On a bright, sunny day, you might be able to spot enemies from a distance, but if a heavy fog rolls in, and youre not equipped for spotting, you may need to rethink your approach and retreat.The key takeaway is that the danger of PvP is always present. In addition to the dynamic world, the Forces of Nature keep you on your toes, ensuring that no two encounters are ever the same.What kinds of meaningful progression can we expect from Exoborne?Exoborne progression is primarily based around your Exo-Rig and your player base. Serving different purposes, they get advanced by performing assignments, finding valuables, doing successful extracts etc. A bunch of activities count towards your progression and even if you die you can expect to gain something. That long term progression is super crucial, because it allows us to keep high risk in every session. We noticed that a game like this requires the risk of losing everything when you go on a raid, at the same time players need to know they havent wasted their time and have nothing to show for it.Will there be things aside from scavenged loot and weapons that players can find in the world?Yes, in addition to scavenged loot and weapons, Exoborne features several other valuable items that players can find in the world, each contributing to the high-risk, high-reward gameplay. One of the most notable being Artifacts which may contain the highest value loot like mods that unlock new abilities for your Exo- Rig. These artifacts come in different tiers, with the highest-tier artifacts marking you on the map for all other players to see. This creates an intense risk-reward dynamicwhile these powerful artifacts can offer great rewards, they also make you a target, encouraging players to weigh the risks of carrying them versus the potential benefits.Another key element is Data Fragments, which are scattered throughout the world. These fragments are essential for unlocking important locations like Rebirth Facilities or special crates, adding another layer of strategy and progression. Players will have to decide whether to go after these data fragments and risk confrontation or focus on other objectives.Additionally, scattered around the map are story elements that help deepen the narrative and world-building. These pieces of lore give players a richer understanding of the Exoborne universe, enhancing immersion and offering context as they explore."In addition to scavenged loot and weapons, Exoborne features several other valuable items that players can find in the world"Are there any plans for a potential PvE-only mode in Exoborne?Currently no. The initial experience is focused around PvPvE they feed off each other and really require each other to provide a tense, thrilling experience. What happens in the future though, we will just have to see!What kinds of special abilities will players be able to use with the Exo-Rig?In Exoborne, everything you can do is tied to your loadoutyou are what you carry. By adding mods to your Exo-Rig, you can unlock a variety of abilities. These range from basic tools, like frag grenades, to more exciting and expansive skills that interact with the Forces of Nature.For example, there are passive abilities that can give you an edge in extreme weather conditions, like making you immune to lightning strikes during a storm, allowing you to move freely while others are forced to take cover. You can even harness lightning itself as an active skill, using it as a weapon to zap your enemies.Even less dangerous elements, like heavy fog, can be leveraged with the right Exo-Rig mod. In these conditions, you can mark enemies for yourself and your team, giving you the upper hand to sneak up on them or eliminate them from a distance before they even know youre there.The ability to adapt to these changing environments and leverage the Forces of Nature for your advantage adds a unique layer of strategy and excitement to combatHow has the response been for the Exoborne playtests?We have a community first mindset in the team. This means we test early and test often. Even though we are just gearing up for our first major online tests, we have been running several smaller ones for quite some time. Depending on where in the project you are, the feedback is of course different. We are eagerly looking forward to the feedback from this playtest, as it is the most advanced and comprehensive build to date."Even less dangerous elements, like heavy fog, can be leveraged with the right Exo-Rig mod."As a developer, what are your thoughts on the PS5 Pro? How does boost in GPU help in developing your game compared to the base PS5?As with any new technology, were always looking at how to best leverage it for our game. While I cant share specific details just yet, I know our teams are actively exploring its potential, and well have more to share in the near future.What are your thoughts on PSSR? What kind of opportunities will this open for the game?We dont have any details to share just yet. As we say above were open to how it can enhance the game, and well provide more updates in the future.Do you have plans to launch the game on Nintendo Switch 2?Given how hot off the press this announcement is, we think its too early to say.

Security leaders and CISOs are discovering that a growing swarm of shadow AI apps has been compromising their networks for over a year.Read More

Recent discussions on AI safety increasingly link it to existential risks posed by advanced AI, suggesting that addressing safety inherently involves considering catastrophic scenarios. However, this perspective has drawbacks: it may exclude researchers with different approaches, mislead the public into thinking AI safety is solely about existential threats, and create resistance among skeptics. As AI rapidly advances, policymakers must establish regulatory frameworks and safety standards. While existential risks dominate the current discourse, past technological safety fieldssuch as aviation, pharmaceuticals, and cybersecurityhave developed robust engineering and governance practices. These frameworks could inform AI safety, ensuring reliable and responsible system deployment.Researchers from the University of Edinburgh and Carnegie Mellon University highlight that AI safety discussions often focus on existential risks, which may exclude diverse perspectives and mislead public perception. Their systematic review of peer-reviewed research reveals a broad spectrum of safety concerns, including adversarial robustness and interpretability, aligning with traditional system safety practices. The study suggests integrating near-term and long-term risks rather than prioritizing existential threats. While AI safety research evolves rapidly, capturing relevant studies remains challenging. Expanding discourse to incorporate established engineering safety principles can help address immediate and future AI risks effectively.The researchers systematically reviewed AI safety literature using a structured methodology based on Kitchenham and Charters guidelines, complemented by snowball sampling to capture emerging research. They focused on two key research questions: identifying risks across the AI system lifecycle and evaluating proposed mitigation strategies. Their search process involved querying the Web of Science (WoS) and Scopus databases, refining results through hierarchical filters, and supplementing findings with influential seed papers. The review process included screening 2,666 database papers and 117 from snowball sampling, ultimately selecting 383 for analysis. Papers were annotated with metadata such as author affiliations, publication year, and citation count and were categorized based on methodological approaches, specific safety concerns addressed, and risk mitigation strategies.The studys bibliometric analysis revealed a steady increase in AI safety research since 2016, driven by advancements in deep learning. A word cloud analysis highlighted key themes such as safe reinforcement learning, adversarial robustness, and domain adaptation. A co-occurrence graph of abstract terms identified four major research clusters: (1) human and societal implications of AI, focusing on trust, accountability, and safety assurance; (2) safe reinforcement learning, emphasizing robust agent control in uncertain environments; (3) supervised learning, particularly in classification tasks, with a focus on robustness, generalization, and accuracy; and (4) adversarial attacks and defense strategies in deep learning models. The findings suggest that AI safety research aligns with traditional safety engineering principles, integrating aspects of reliability engineering, control theory, and cybersecurity to ensure AI systems are both effective and secure.AI safety research categorizes risks into eight types: noise, lack of monitoring, system misspecification, and adversarial attacks. Most studies address issues related to noise and outliers, affecting model robustness and generalization. A significant focus is also on monitoring failures, system misspecifications, and control enforcement gaps. Research methods include applied algorithms, simulated agents, analysis frameworks, and mechanistic interpretability. While theoretical works propose conceptual models, applied studies develop practical algorithms. Recent efforts emphasize reinforcement learning safety, adversarial robustness, and explainability. The field parallels traditional engineering safety, integrating verification techniques to enhance AI reliability and mitigate potential risks.In conclusion, the study systematically reviewed peer-reviewed literature to explore AI safety challenges. The findings highlight diverse motivations and research outcomes aimed at ensuring AI systems are reliable and beneficial. AI safety research addresses various risks, including design flaws, robustness issues, inadequate monitoring, and embedded biases. The study advocates for framing AI safety within broader technological safety, expanding stakeholder engagement, and promoting inclusive research. While existential risks remain relevant, a wider perspective fosters productive discourse. Future research should explore sociotechnical AI safety and incorporate non-peer-reviewed sources for a comprehensive understanding, ensuring AI safety remains an evolving, inclusive, and multidisciplinary field.Check outthePaper.All credit for this research goes to the researchers of this project. Also,feel free to follow us onTwitterand dont forget to join our75k+ ML SubReddit. Sana HassanSana 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.Sana Hassanhttps://www.marktechpost.com/author/sana-hassan/Nous Research Released DeepHermes 3 Preview: A Llama-3-8B Based Model Combining Deep Reasoning, Advanced Function Calling, and Seamless Conversational IntelligenceSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Layer Parallelism: Enhancing LLM Inference Efficiency Through Parallel Execution of Transformer LayersSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Can 1B LLM Surpass 405B LLM? Optimizing Computation for Small LLMs to Outperform Larger ModelsSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Meet OpenThinker-32B: A State-of-the-Art Open-Data Reasoning Model Recommended Open-Source AI Platform: IntellAgent is a An Open-Source Multi-Agent Framework to Evaluate Complex Conversational AI System' (Promoted)

What Millions of Conversations Reveal About AIs Real Economic Impact 0 like February 17, 2025Share this postLast Updated on February 17, 2025 by Editorial TeamAuthor(s): Vita Haas Originally published on Towards AI. A new study by Anthropic offers a rare, data-driven glimpse into how artificial intelligence is actually being used in the workplace rather than how its supposed to be used. Based on over four million real-world conversations with Claude.ai, the research, titled Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations, maps these interactions to occupations and tasks using the U.S. Department of Labors O*NET database.The study set out to answer some fundamental questions: Where is AI proving most useful? What kinds of skills are people outsourcing to it? And is it helping humans work smarter or simply taking over their tasks altogether?From the Anthropic Economic IndexThe findings suggest that AI is not sweeping entire industries off the table, nor is it single-handedly ushering in an age of mass unemployment. Instead, it has cozied up in knowledge-based jobs, particularly in software development and writing-related work, where it plays the role of an indefatigable assistant. Together, these two domains account for nearly 50% of all recorded AI interactions. More broadly, 36% of occupations now use AI for at least a quarter of their tasks, though only 4% lean on it for over 75% of their workload.Rather than replacing workers outright, AI is proving itself to be a helpful, if sometimes overeager, co-worker. In 57% of cases, it assists by improving efficiency, refining ideas, or offering guidance a second set of (digital) eyes on a project. The remaining 43% of cases involve full automation, where AI is essentially left to its own devices, completing tasks with little or no human oversight.Where AI Thrives And Where Its Nowhere to Be FoundAI, it turns out, is a bit of a specialist. It excels in structured, text-based, analytical work things like coding, content creation, data analysis, and technical documentation. If a task involves generating reports, suggesting programming fixes, summarizing research, or transforming a dull paragraph into something vaguely readable, AI is right at home.But despite all the grand predictions about automation upending everything, AI remains conspicuously absent from roles that demand physical dexterity, improvisation, and real-world decision-making. Professions such as surgeons, electricians, construction workers, and restaurant staff remain largely unaffected possibly because no one in their right mind wants a chatbot performing open-heart surgery or tiling their bathroom (yet).Similarly, AI has made little headway in high-stakes leadership roles such as executives, doctors, and top-tier lawyers, where a misplaced decimal or a legally questionable clause could mean actual disaster. AI may be decent at suggesting a strongly worded email, but it still lacks the judgment needed to steer companies or make life-and-death decisions.At least, for now.From the Anthropic Economic IndexCo-Pilot or a Replacement? The 57/43 SplitThe study highlights two primary ways AI is being used: as a collaborator or as an automator.In 57% of cases, AI acts as an augmentation tool, making workers more efficient without actually replacing them. A writer refines a draft with AI-generated suggestions. A programmer uses AI for debugging. A financial analyst double-checks a models assumptions. In these cases, AI is less of a replacement and more of a productivity booster like a calculator, if calculators also occasionally got things hilariously wrong.But in 43% of cases, AI takes full control of tasks, completing them with minimal human input. An executive generates a report and barely skims it before sending it off. A marketing team lets AI produce ad copy with only a token edit. A financial model is built entirely by AI, with no human adjustments. Here, AI has shifted from assistant to invisible worker, quietly absorbing responsibilities that once belonged to humans.For now, augmentation remains more common than automation but as AI improves, that balance could change.Who Uses AI the Most?AI adoption follows a surprisingly clear pattern across different wage levels.Mid-to-high-wage professionals especially in software development, finance, marketing, and content creation use AI the most. These jobs involve structured, repetitive, and digital work, making them a natural fit for AI assistance.But at both extremes of the income spectrum, AI adoption drops. Low-wage jobs, like those in food service, warehouse work, and construction, dont use AI much because their work is physical, unpredictable, and requires hands-on adaptability. At the other end, top executives, doctors, and high-end lawyers have also shown low AI adoption, likely due to regulatory complexity, human judgment, or the sheer consequences of getting something wrong.So while AI is reshaping middle-class knowledge work, it has yet to make serious inroads into either physical labor or elite professions.The Strange Case of Outsourced ThinkingOne of the most eyebrow-raising findings in the study is the type of skills people are handing over to AI.At the top of the list is critical thinking.Closely followed by active listening and reading comprehension all of which, you know, seem fairly important if humans want to continue being good at thinking for themselves. The fact that AI is already being used to support reasoning and decision-making suggests that people arent just outsourcing what to think they may be outsourcing how to think.This isnt necessarily a bad thing. AI can help structure arguments, point out gaps in logic, and offer alternative perspectives. But if people start leaning on AI for comprehension, judgment, and decision-making, we may find ourselves in an era where independent thinking is more of an optional skill.If AI continues to take over tasks that require analysis, problem-solving, and reasoning, will humans get better at using these tools or simply forget how to do it themselves?From the Anthropic Economic IndexWhat It Means for the Future of WorkAI isnt taking over the workforce overnight, but it is fundamentally changing how work gets done. Instead of eliminating jobs outright, its making workers faster, more efficient, and increasingly reliant on AI for both routine and complex tasks.For now, AI is a tool, not a replacement. But this study suggests that its role is expanding. Writing assistance has evolved into fully automated content generation. Code suggestions have transformed into self-writing software. Data analysis has moved beyond interpretation to prediction and action.At what point does AI shift from enhancing human work to becoming an independent force in economic decision-making? The study suggests we may already be testing that boundary. AI is increasingly involved in critical thinking, decision-making, and comprehension skills once considered uniquely human. If this reliance continues, industries may need to redefine what human expertise actually means.This raises urgent questions for policymakers, businesses, and workers. If AI keeps absorbing more cognitive tasks, will professionals shift from deep expertise to AI oversight roles? How will companies measure competence and accountability when AI is doing more of the heavy lifting?The future of work wont be defined by whether AI replaces humans, but by how much responsibility we hand over to it. Whether this transition leads to greater productivity and innovation or a slow erosion of human autonomy and expertise depends entirely on how we choose to integrate AI into our workflows.For now, AI remains a silent partner, amplifying productivity while still requiring human oversight. But as businesses adopt AI at scale, one question looms larger than ever:not just what AI can do but how much of the decision-making we are comfortable handing over.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

LatestMachine LearningRAG vs Fine-Tuning for Enterprise LLMs 0 like February 17, 2025Share this postLast Updated on February 17, 2025 by Editorial TeamAuthor(s): Paul Ferguson, Ph.D. Originally published on Towards AI. RAFT vs Fine-Tuning Image created by authorAs the use of large language models (LLMs) grows within businesses, to automate tasks, analyse data, and engage with customers; adapting these models to specific needs (e.g., responding to support queries with current policies or analyzing legal documents with custom terms) is increasingly important, and with it, the need to plan accordingly.Two approaches prevail in this space:Fine-tuning, which adjusts the models core knowledgeRetrieval-Augmented Generation (RAG), which incorporates data in the responseEach method has its advantages, disadvantages, and tradeoffs, but the choice is not always obviousThis guide provides a step-by-step framework for technical leaders and their teams to:Understand how RAG and fine-tuning work in plain termsChoose the approach that best fits their data, budget, and goalsAvoid common implementation pitfalls: poor chunking strategies, data drift, and othersCombine both methods for complex use casesUnderstanding the Core TechniquesFine-TuningFine-tuning is a technique of adjusting the parameters of a pre-trained LLM to specific tasks using domain-specific datasetsThis ensures that the model is well-suited for that specific task (e.g., legal document review)It excels in tasks that require specialised terminologies or brand-specific responses but needs a lot of computational resources and may become obsolete with new data.For instance, a medical LLM fine-tuned on clinical notes can make more accurate recommendations because it understands niche medical terminology.Fine-tuning Architecture Image created by authorWithin the fine-tuning architecture weve included both:In green, the steps to generate the fine-tuned LLMIn red, the steps to query the modelNote: within the query section, weve labelled the system responsible for controlling/co-ordinating the query and response an intelligent system: this is just a general name for illustration purposes. Within enterprise systems there are many different variations that can exist within this intelligent system, which themselves may include AI agents, or other LLMs to provide more sophisticated functionality.Retrieval-Augmented Generation (RAG)RAG enhances LLMs by fetching additional information from external sources during inference to improve the response. It combines the users query with other relevant information to ensure the accuracy of the response (potentially incorporating live data).Some of its key advantages include:Less hallucinations since the model is forced to rely on actual data;Transparent (it cites sources);Easy to adapt to changing data environment without modifying the model.Example: A customer support chatbot using RAG can fetch the real time policy from internal databases to answer the queries accurately.RAG Architecture Image created by authorAgain, weve coloured-coded the architecture:Green, denotes the pre-query aspects of the system: associated with indexing the documentsRed, identifies the steps that are executed at query timeFrom examining the two architectures of fine-tuning vs RAG, we can see a number of key differences, one of the most striking is the overall complexity of the RAG system: but we should be careful not to lose sight of the complexity associated with the fine-tuning step in fine-tuningAlthough it is only represented by one step in the architecture, it is still a complex and potentially costly processIt also requires careful preparation of the custom data, as well as correct monitoring of the fine-tuning to ensure that the learns the desired informationHowever, one key aspect related to the RAG complexity is that at query time there is a lot more work being done by the RAG system: this will naturally result in longer query times.Key Decision FactorsWhen selecting between RAG and fine-tuning, consider these factors:RAG vs Fine-Tuning Decision Factors Image created by authorNote: in certain circumstances, a Hybrid Approach is needed (which we discuss below)Common Challenges and SolutionsRAG Challenges1. Chunking IssuesProblem: The poor chunk size leads to incomplete context or irrelevant document retrieval.Solution: Use overlapping chunks (e.g., 25% token overlap) or semantic splitting at logical segments (sentences or paragraphs).2. Retrieval QualityProblem: Overreliance on vector similarity missing keyword critical matches.Solution: Combine vector embeddings with keyword based BM25 scoring for hybrid search.3. Response ConsistencyProblem: Noisy retrieved contexts lead to varying output.Solution: Create structured prompt templates that enforce source citation and output format.Fine-Tuning ChallengesCatastrophic ForgettingProblem: Models lose general knowledge in the process of domain adaptation.Solution: Use parameter efficient methods like Low Rank Adaptation (LORA) with Bayesian regularisation to preserve the overall capability.2. Data QualityProblem: Biased or outdated training data affects the output.Solution: Build a validation pipeline with domain experts and automate checks for the dataset (e.g., balance, outliers).3. Version ControlProblem: Managing model iterations is prone to error.Solution: Keep a model lineage registry (with tools like Hugging Face Model Hub) and document hyperparameters and training data.Implementation Best PracticesRAG ImplementationData Pipeline Design: Use semantic search in vector databases like Pinecone and chunk documents to achieve relevance efficiency.Evaluation: Set up an automated testing framework like Ragas to assess the accuracy of responses and how well they are grounded in the data.Security: Secure sensitive data with access control (role-based) and metadata.Fine-Tuning ImplementationData Preparation: Use large, properly labelled datasets (>10,000 examples) to build the model and reduce the risk of overfitting.Parameter Efficiency: Use LoRA to reduce computational costs while retaining the general capabilities of the model.Validation: Check the output against domain experts to ensure that it meets the requirements of the task.Hybrid Approach: RAFTRAFT (Retrieval-Augmented Fine-Tuning) combines the best of both worlds by integrating RAG with fine-tuning to create models that excel at knowledge-based tasks (e.g., legal and healthcare domains are some of the most common industries for hybrid approaches, due to their need for domain specialisation as well as requiring highly accurate, and traceable results).In terms of the RAFT architecture, this involves a straightforward combination of the two architectures already illustrated:Firslty creating a fine-tuned LLMThen integrating the fine-tuned LLM (instead of a pre-trained LLM) into the RAG architectureKey ComponentsTraining Data Design: Select a set of oracle documents that contain correct answers and a set of distractor documents that contain irrelevant information to teach the model to focus on credible sources.Training Process: Fine-tune the model to explicitly refer to the retrieved passages and to reduce the chance of hallucinations using techniques like chain of thought prompting.Implementation Steps:Index domain documents (e.g., policy updates).Create synthetic QA pairs with oracle and distractor contexts marked.Fine-tune using LoRA to preserve the general ability to generate text while adapting to new data.BenefitsReduced Hallucination: Basing responses on verified sources.Domain Adaptation: Outperforms other methods in dynamic, specialised environments.Key Takeaways & ConclusionRAG, Fine-Tuning, RAFT decision matrix Image created by authorThe key to successful deployment of enterprise LLMs depends on aligning the strategy with operational realities:RAG vs. Fine-Tuning: Use RAG for transparent solutions with dynamic data (e.g., customer-facing chatbots). Fine-tune when deep domain customisation is needed (e.g., healthcare).Hybrid Strategies: Some examples of hybrid approaches include RAFT or RoG (Reasoning on Graphs) for combining real-time retrieval with domain expertise for tasks like building a legal compliance tool.Continuous Evaluation: Periodically check the retrieval accuracy (using tools like Ragas) and model outputs to prevent drift or hallucinations.No single approach fits all, but understanding these principles ensures your LLM investments deliver scalable, accurate results.If youd like to find out more about me, please check out www.paulferguson.me, or connect with me on LinkedIn.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

Even if you dont play Magic: The Gathering, youve probably caught wind of its numerous video game crossovers in recent years, including Fallout, Tomb Raider, and Assassins Creed. But today weve got an exclusive first look at one of the most exciting yet: Final Fantasy. And not just one Final Fantasy from Terra to Yshtola, four different mainline games are represented across the preconstructed Commander decks headlining the set.Flip through the image gallery below for a first look at the lead card and packaging for each deck, and read on for a conversation with Wizards of the Coast about what to expect inside them, why these four games were picked, and plenty more. Final Fantasy x Magic: The Gathering - Commander Decks RevealSet to launch this June, Magics Final Fantasy crossover will be a fully draftable, Standard-legal set accompanied by the four preconstructed decks in the gallery above. Each of those includes 100 cards made up of both reprints of existing cards with new Final Fantasy art and brand new cards designed with Magics most popular format, Commander, in mind. And while Commander precons are always themed around certain characters, colors, or strategies, the twist with these is that each one is built around a single Final Fantasy game specifically 6, 7, 10, and 14. Final Fantasy games are so rich with flavor, beloved characters, and unique settings that choosing a single game as the theme provided more than enough material to design a full deck of cards around, says Senior Game Designer Daniel Holt, who is the Commander Lead for the set. Approaching each deck as a single game also provided the opportunity to dive deep into the lore of each, capturing even more beloved moments from across the games storyline that we may not have been able to achieve otherwise."...we have so many passionate Final Fantasy fans in the building.The team decided on these four Final Fantasy games based on a balance of what they wanted for the gameplay of each deck and how well known each story was overall. Holt says that while Final Fantasy 7 and 14 were clearer picks, 6 and 10 took more discussion to arrive at, but were chosen partially because they were favorites among the team. This was one of those products where everyone in the building was invested in each stage of development, as we have so many passionate Final Fantasy fans in the building.Of course, even within those picks, there are some decisions you have to make about direction. For instance, Final Fantasy 7s remake trilogy has been running alongside the development of this Magic set, so is its Commander precon telling the story of the original or its reality-bending reimagining? Dillon Deveney, Principal Narrative Game Designer at Wizards of the Coast and Narrative Lead for the set, says youll see a bit of both throughout that deck's art, but the story itself will follow the 1997 classic.Our core approach to Final Fantasy VII was to capture the original PS1 games narrative, while using Final Fantasy VII Remake and Final Fantasy VII Rebirths modern aesthetics to push the fidelity of each character design, story moment, and memorable location to the next level, Deveney explains. With that in mind, if a scene existed in both games, we had our choice of showing it as the original game did, how the modern version interpreted it, or a unique fusion of both. Hopefully, this deck feels familiar and nostalgic to players of both the original game and the modern series!Speaking of nostalgia, Final Fantasy 6 is a trickier nut to crack in some ways as it lacks the more modern art references the other three games can provide. Deveney says they wanted to stick close to what fans would expect from its pixel art sprites and limited concept art while also expanding upon them, and hopes that the character designs should feel like how you remember them, despite being a hybridization of many different references and new ideas. To help accomplish that, WOTC even consulted directly with the Final Fantasy 6 team as they updated these character to the standards needed for Magic art.We developed a workflow that asked concept artists and card artists to pull hallmark character designs from Yoshitaka Amanos original (and legendary) concept art, the original FFVI sprites, and the FFVI Pixel Remaster character portraits, to synthesize the most consistent elements of each characters design into something new. We also encouraged our artists to punch up finer details and bsexplore elements where they thought they could push (fabric, patterns, textures, etc.) Then, we would meet with the Final Fantasy VI review team, and they would give us feedback on what core elements they wanted to preserve and what new elements they were excited about."It's our hope that the [FF6] character designs should feel like how you remember them..."So they had the four games decided, but when it comes to Commander, thats only half the equation WOTC also needed to figure out which characters would lead each deck. Cloud may be the obvious pick for 7, but Holt tells me the others involved some brainstorming. Celes was considered for 6 since the deck is focused around the World of Ruin and the second half of that game, while Yuna was considered for 10 as a fan-favorite. They ultimately decided to stick with the lead characters on those, but Final Fantasy 14 is an MMO, which presented its own problem.For Yshtola, it came down to a mix of popularity of the character as well as being a spellcaster, her story and character development offered a wide range of abilities and moments to pull from to lead the deck, Holt explains, going on to say this deck represents her specifically during her Shadowbringers arc. They did explore the concept of making a Warrior of Light commander players could make their own in some fashion, but Holt says those ideas had some pretty in-depth executions, and that your own personal hero still has plenty of moments in this deck even without being at the helm. PlayNow comes the question of figuring out how to fit an entire video games story, characters, and themes into a single deck and the carefully balanced restrictions of Magics five colors. Building these decks called for us to decide color identity for the chosen game itself, but also the gameplay we desired from them, says Holt, going on to point out that its no mistake all four include White, for both theming and it helped cover the wide range of Heroes we wanted to include in each deck.As mentioned, the deck for 6 is centered on the latter half of that game as such, its essentially all about rebuilding your party by bringing other creatures back from your graveyard. For 7, Cloud wielding a big sword lines up nicely with the equipment strategies of a white-red Magic deck, but Holt says that adding green to the mix let us tie in the power matters cards and cards that reference the Planet and lifestream.While Tidus ability to pass counters around might lead you to think the deck for 10 is all about Blitzball, this classic white-blue-green strategy actually primarily came from the idea of that games Sphere Grid leveling system as a way to empower your creatures. Holt also admits that the theme for 14 was tricky to land on, but its white-blue-black color identity allowed them to lean into a noncreature spell casting theme while also including the characters they wanted to."...fans can expect to see some of their favorite characters in the 99 of each deck..."Last but not least, Commander might be all about your leader, but RPGs are often about your party as a whole, and WOTC wont leave the supporting casts of these games hanging. "Final Fantasy games are full of loveable AND villainous characters and getting them in these decks was very important, Holt says. While I cant talk specifics on who is included in each deck or their abilities just yet, fans can expect to see some of their favorite characters in the 99 of each deck as new legendary creatures and captured in action on other exciting spells. Magics Final Fantasy set will release on June 13, and if you dont see your personal favorite Final Fantasy game or character here (or even if you do), dont worry Holt assures me that all sixteen of the mainline games will have their moments to shine in the accompanying products. Returning to the same territory as the Warhammer 40,000 Commander decks from 2022, all four of these decks will be available in both a regular version (MSRP $69.99) and a Collectors Edition (MSRP $149.99), the latter of which will feature all 100 of the cards in each deck in a special Surge foil treatment. Read on for the full, unedited interview with Wizards of the Coasts Daniel Holt and Dillon Deveney:IGN: Why did you decide to build each commander deck around a whole game this time? Daniel Holt, Senior Game Designer: Final Fantasy games are so rich with flavor, beloved characters, and unique settings that choosing a single game as the theme provided more than enough material to design a full deck of cards around. Approaching each deck as a single game also provided the opportunity to dive deep into the lore of each, capturing even more beloved moments from across the games storyline that we may not have been able to achieve otherwise.How did you land on these four games specifically? Were there any close runner-ups you considered? DH: In exploratory design, before getting too deep into deckbuilding and card design, we brainstormed what games offered the most for what gameplay we wanted with the decks as well as being the more commonly known storylines among the audience. Final Fantasy VII being one of the more referenced games in media made it a strong first contender. Final Fantasy XIV offered a unique gamestyle, that of an MMO, and has SO MANY active players that we knew we could delight this group by selecting it as one of the four. After that, it became a bit more of a team discussion. Final Fantasy X is a favorite of mine and Zakeel, the FIN Product Architect, so we were a little biased in choosing it. Final Fantasy VI touched the pixel era of games and was a favorite of several other designers in the studio. So with that, we ended up with our four game selects! That said, all sixteen of the mainline games will have their moments to shine in the accompanying products.Should we expect reprint art/lands/etc. within each deck to also be tied to each game?DH: Absolutely! We want players to feel like they are entering the world of that game as they summon creatures and cast spells from it. This means that the nonland reprints are placed from within that decks game, while lands can appear from any of the four games included in the product.Can you talk a bit about your approach to Magics color identities across these decks as it relates to Final Fantasy why these colors for these characters/games? DH: Building these decks called for us to decide color identity for the chosen game itself, but also the gameplay we desired from them. Being three colors each did help me have a wider range of options to apply to different characters and moments within the game. For example, youll notice all four decks include White in their color identity for both theming and it helped cover the wide range of Heroes we wanted to include in each deck. Final Fantasy VI This is a graveyard reanimator deck, so white-black-red was a great fit gaining access to discard and mill effects in BR and return from graveyard effects from WB. Thematically this deck rebuilds its large ensemble of characters in the World of Ruin, so the deck theme really plays into that section of the story.Final Fantasy VII As an equipment matters deck, white-red is well-known for that archetype in Magic and showcases this groups iconic weaponry and battle prowess. Adding green to the mix let us tie in the power matters cards and cards that reference the Planet and lifestream. Final Fantasy X White-blue-green counters are a classic staple in Magic deckbuilding and having access to all three here really let us build out a whole deck about moving the counters on the board and leveling up your team. Final Fantasy XIV This deck was tricky to land on, but by being white-blue-black, we had access to the colors we needed for the characters we wanted to include, but also be a noncreature matters theme, representing spell casting and ability usage you perform as the player while playing the Final Fantasy XIV game! Terra, Cloud, and Tidus are the clear picks for their games, but how did you land on Yshtola for Final Fantasy 14?DH: I wont deny that Cloud was a pretty obvious choice for the face commander of the Final Fantasy VII deck. The others went through a few rounds of brainstorming though! In Final Fantasy VI, since we were focusing on the themes of the second half of the game, Celes was actually a big contender for the lead. And with Final Fantasy X, we know Yuna is a fan favorite and could have led the deck to success. We ultimately landed on the main protagonist character for each of these titles, resulting in Terra and Tidus filling those slots respectively; though Final Fantasy VI is an ensemble cast, Terra is often portrayed as the face of the series. For Yshtola, it came down to a mix of popularity of the character as well as being a spellcaster, her story and character development offered a wide range of abilities and moments to pull from to lead the deck.Was using the Warrior of Light as the lead ever on the table, perhaps with some sort of background/class system to let players customize their precon?DH: You caught me! We did explore card designs and themes around the Warrior of Light with somepretty in-depth executions, before ultimately landing on Yshtola as the commander. That said, there are still plenty of Warrior of Light moments in the deck!7, 10, and 14 have plenty of modern references when it comes to Magic-style art, but youre pulling from a comparatively smaller pool for 6. Did you feel like there was more freedom to define your own look for those characters, or did you try to stick as closely to what was out there as possible? Dillon Deveney, Principal Narrative Game Designer: I love this question! We wanted to stick closely to what fans remembered about these characters, while looking to merge the two major aesthetics of FFVI: the concept art and the in-game sprites. There was a lot of freedom in this approach, while trying to stick closely to the original designs. We worked directly with the Final Fantasy VI team to determine the aesthetics, designs, and costumes of the FFVI characters and world. To do that, we developed a workflow that asked concept artists and card artists to pull hallmark character designs from Yoshitaka Amanos original (and legendary) concept art, the original FFVI sprites, and the FFVI Pixel Remaster character portraits, to synthesize the most consistent elements of each characters design into something new. We also encouraged our artists to punch up finer details and explore elements where they thought they could push (fabric, patterns, textures, etc.) Then, we would meet with the Final Fantasy VI review team, and they would give us feedback on what core elements they wanted to preserve and what new elements they were excited about. Ultimately, it's our hope that the character designs should feel like how you remember them, despite being a hybridization of many different references and new ideas.Conversely, was there more pressure to get a character like Yshtola exactly right when fans know so precisely what she looks like? DH: From the game design side for each of these popular characters, this was one of those products where everyone in the building was invested in each stage of development, as we have so many passionate Final Fantasy fans in the building. At every playtest of the decks, I would get feedback about coworkers favorite characters, story moments, etc. that they wanted to make sure got included and implemented properly.Some characters, like Yshtola, have such deep character development throughout portions of their game. Each card that depicts one of those types of characters needed to capture them at a specific point in time of their story. For the Final Fantasy XIV commander deck, we have Yshtola in the Shadowbringers arc where she has become a talented spell caster.For the Final Fantasy 7 deck, did the recent remake series influence or impact your thinking/designs much, or were you mostly looking toward the original here? DH: We had the fortunate overlap of some of the remake game series releasing WHILE we worked on this set. So it was a really cool experience getting to leave work, go home, and play games featuring the characters of cards I was designing in a new video game.DD: Our core approach to Final Fantasy VII was to capture the original PS1 games narrative, while using Final Fantasy VII Remake and Final Fantasy VII Rebirths modern aesthetics to push the fidelity of each character design, story moment, and memorable location to the next level. You will see a mix of classic FFVII moments and modern FFVII visuals in the deck, but generally, the product is intended to follow the original games story. With that in mind, if a scene existed in both games, we had our choice of showing it as the original game did, how the modern version interpreted it, or a unique fusion of both. Hopefully, this deck feels familiar and nostalgic to players of both the original game and the modern series! Its super cute that Tidus is all about passing counters around like a ball. Did you know this was going to be a counter deck already or did counters come from exploring Blitzball as a concept? DH: Everyone loves Blitzball! People in the office got a kick out of every time we mentioned it or referenced it somewhere in the product. But when starting the Final Fantasy X deck, the actual theming of counters came from the Sphere Grid experience system used in the game. This system is visually portrayed as a board where you add colored markers on the tracks to gain new abilities. In commander, I wanted to capture that idea by filling the deck with +1/+1 and other counters and then moving them around your team for added abilities!Commander is all about your leader, but what is an RPG without a party was it important to yall to also highlight the folk supporting these leads within each deck? DH: Of course! Final Fantasy games are full of loveable AND villainous characters and getting them in these decks was very important. While I cant talk specifics on who is included in each deck or their abilities just yet, fans can expect to see some of their favorite characters in the 99 of each deck as new legendary creatures and captured in action on other exciting spells.I assume a popular character like Cloud is going to be on more than just this card in the same way Frodo or Galadriel were in the Lord of the Rings set I know you cant get into main set spoilers here, but is there anything about designing around these characters for a commander-only product that you enjoy getting to do differently from a product that has to be in Standard? DH: I think its safe to say youll see some characters making multiple appearances in the product and each is appealing to different kinds of players. With the ones we see here, they are all commander-sized, meaning they are aimed to be exciting in the commander format, both in and out of the precon. All the legends in these decks have a unified goal of working together within their decks theme and its fun building cards like this to have that creative team synergy, whether thats sayinteracting with counters in Final Fantasy X or playing with the graveyard in Final Fantasy VI. On the other hand, legends in the main set can be designs built around the character independently of other cards to spark new archetypes of decks.Tom Marks is IGN's Executive Reviews Editor. He loves card games, puzzles, platformers, puzzle-platformers, and lots more.

With every new WWE 2K release comes the promise of the largest roster ever and WWE 2K25 breaks the record again, powerbombing through the 300+ barrier this year when including variants. But who has made the cut in WWE 2K25? Lets find out.The following is a full list of playable superstars in WWE 2K25, pulling from the Raw, Smackdown, and NXT rosters, and of course, a mouthwatering amount of WWE legends. Many superstars such as Shawn Michaels and The Undertaker also have variants, letting you play as superstars from different eras. However, because these are sometimes unlocked in the Showcase and MyRise modes, we wouldnt want to spoil them for you. This list is variant-free. But enough talk. In alphabetical order, here are all the superstars on the roster for WWE 2K25.WWE 2K25 Roster:AJ StylesAkamAkira TozawaAlba FyreAlexa BlissAlundra BlayzeAndradeAndre ChaseAndre The GiantAngelAngelo DawkinsApollo CrewsAshante "Thee" AdonisAsukaAustin TheoryAxiomB-FabBaron CorbinBatistaBayleyBecky LynchBertoBianca BelairBig EBlair DavenportBoogeymanBooker TBraun StrowmanBray WyattBret HartWWE 2K25 Roster - RawBritish BulldogBron BreakkerBronson ReedBrooks JensenBruno SammartinoBrutus CreedBubba Ray DudleyCactus JackCandice LeRaeCarlitoCarmellaCarmelo HayesCedric AlexanderChad GableChanning Stacks LorenzoCharlie DempseyCharlotte Flair Chelsea GreenChynaCM PunkCody RhodesCora JadeCruz Del ToroD-Von DudleyDakota KaiDamian PriestDexter LumisDiamond Dallas PageDieselDoink The ClownPlayDominik MysterioDragon LeeDrew McIntyreDude LoveDuke HudsonDusty RhodesEddie GuerreroEddy ThorpeElektra LopezElton PrinceEric BischoffErik Ethan PageEve TorresFaarooqFallon HenleyFinn BalorGeorge "The Animal" SteeleGigi DolinGiovanni VinciGrayson WallerGuntherHarley RaceHeadshrinker FatuHeadshrinker SamuHonky Tonk ManHollywood Hulk HoganHulk HoganHunter Hearst HelmsleyWWE 2K25 Roster - Smackdown!Ilja DragunovIndi HartwellIsla DawnIslander Haku Islander TamaIvarIvy NileIYO SKYJacob FatuJacy JayneJade CargillJaida ParkerJakara JacksonJake "The Snake" RobertsJBLJD McDonaghJe'Von EvansJey UsoJim "The Anvil" NeidhartJimmy UsoJoaquin WildeJoe CoffeyJoe GacyJohn CenaJohnny GarganoJosh BriggsJulius CreedKairi SaneKaneKarl AndersonPlayKarrion KrossKatana ChanceKayden CarterKelani JordanKen ShamrockKevin NashKevin OwensKiana JamesKit WilsonKofi KingstonKurt AngleLA KnightLash LegendLex LugerLexis KingLitaLiv MorganLogan PaulLola ViceLudwig KaiserLuke GallowsLyra ValkyriaMark CoffeyMaryseMankindMaxxine DupriMichelle McCoolMichinMick FoleyMighty MollyWWE 2K25 Roster - NXTMolly HollyMontez FordMoshMr. FujiMr. PerfectNaomiNatalyaNathan FrazerNia JaxNikki CrossNikkita LyonsNoam DarOba FemiOmosOro MensahOtisPat McAfeePete DunnePeter MaiviaPiper NivenR-TruthRandy OrtonRandy SavageRazor RamonRaquel RodriguezRey Mysterio RezarRhea RipleyRick RudeRick SteinerPlayRicky "The Dragon" SteamboatRidge HollandRikishiRob Van DamRoddy PiperRocky MaiviaRoman ReignsRoseyRoxanne PerezSami ZaynSantos EscobarScarlettScott Hall Scott SteinerSensational SherriSeth RollinsShawn MichaelsShawn SpearsShayna BaszlerSheamusShinsuke NakamuraShotziSol RucaSolo Sikoa Sonya DevilleStacy KeiblerStardustStephanie McMahonStone Cold Steve AustinSuperstar Billy GrahamWWE 2K25 Roster - LegendsSyxxTama TongaTaminaTatum PaxleyTegan NoxTerry FunkThe FiendThe Great MutaThe HurricaneThe Iron SheikThe MizThe RockThe SandmanThea HailThrasherTiffany StrattonTommaso CiampaTonga LoaTony D'AngeloTrick WilliamsTriple HTrish StratusTyler BateTyler BreezeUltimate WarriorUmagaUndertakerVaderValhallaWade BarrettWendy ChooWes LeeWild Samoan AfaWild Samoan SikaWilliam RegalWolfgangX-PacXavier WoodsYokozunaZelina VegaZoey StarkNon-Playable Managers:AvaAdam PearceAfaArmando Alejandro EstradaBobby "The Brain" HeenanBrother LoveCaptain Lou AlbanoCathy KelleyJimmy HartMick FoleyMiss ElizabethNick AldisPaul BearerPaul ElleringPaul HeymanStephanie McMahonTheodore LongPre-Order Bonus Characters:Wyatt Sicks - Dexter LumisWyatt Sicks - Erick RowanWyatt Sicks - Joe GacyWyatt Sicks - Nikki Cross Wyatt Sicks - Uncle HowdyWhat do you think? Did your favourite WWE Superstar make the cut? Whos missing and who do you hope to see in future DLC packs? Let us know in the comments. Be sure to stick with IGN the entire month of February where were exclusively taking a deeper look into WWE 2K25. We got to share our hands-on thoughts of WWE 2K25, the first look at the brand new Underground match type, and the cover star Roman Reigns react to gameplay from the Bloodline showcase mode.Dont miss more WWE 2K25 plus Monster Hunter Wilds, ID@Xbox, Daredevil: Born Again, The Monkey, Mickey 17 and much, much more at IGN Fan Fest streaming across IGN platforms February 24-28.