creators

Adobe presents itself as a champion of creators in the age of AI. Lawsuit alleges he used copyrighted books

by

Adobe has built part of its artificial intelligence strategy on a very recognizable banner: protecting creators in a time of profound change. While other technology companies accumulated criticism for the origin of their data, the company presented itself as a responsible alternative. That position is now facing a lawsuit which focuses on the training of one of its models and the use of copyrighted works. The case is not an anomaly, but rather a reflection of a question that the industry has not yet been able to clearly answer. The lawsuit was filed Tuesday in the U.S. Court for the Northern District of California and takes the form of a proposed class action. An author named Elizabeth Lyon accuses Adobe of using copyrighted books, including her own, to train the company’s AI models, with SlimLM at the center of the case, without permission. According to judicial documentation, these works would have been part of the training process of systems designed to respond to human instructions. Lyon claims to be acting on behalf of other rights holders who would find themselves in a similar situation. The great debate about data that trains AI To understand why this type of litigation is repeated with increasing frequency, it is worth taking a moment to look at how current artificial intelligence works. Beyond the visible applications, from chatbots to image generators, there are underlying models that act as the core of the system and learn from huge volumes of data. Generally speaking, more data can improve performance, although it is not the only factor. The problem appears when the key question arises about the origin of that information and the conditions under which it has been used. The model indicated in the lawsuit is not Firefly, Adobe’s best-known creative system, but SlimLMa family of smaller language models designed for specific tasks. These models are designed to assist users with document-related functions, especially on mobile devices. It is not an AI aimed at large-scale creative generation, but rather a system that operates in the background. That difference is relevant because it shows that the debate over training data is not limited to the most visible applications. According to the lawsuit, the conflict would not be in SlimLM as a final product, but in the data used during its training phase. Adobe has explained that these models were pre-trained with SlimPajama-627Ba open source data set published by Cerebras in June 2023. The court brief maintains that SlimPajama derives from RedPajama, another dataset widely used in the industry, and which in turn incorporates Books3, a massive collection of copyrighted books. That chain is the one that, according to the plaintiff, would have allowed the inclusion of works without authorization. Until now, Adobe’s public narrative on artificial intelligence has been primarily articulated around Fireflya product clearly identified with respect for creators and the use of licensed content. The company has defended that these models were trained with licensed content, such as Adobe Stock, and public domain material, and has accompanied that message with compensation programs for Adobe Stock contributors. The demand, however, is not directed at that visible front, but, as we say, at SlimLM, a more discreet model, integrated into assistance tasks and without a direct commercial presence. This separation is key to understanding the real scope of the case. The proceedings against Adobe are framed in a broader context of litigation in the United States related to the training of AI models. In recent years, authors and other rights holders have taken to court technology companies like OpenAIor Anthropicwith lawsuits alleging the use of protected works without authorization. Some of these processes are still open and others have ended in million-dollar agreements. This scenario explains why each new case is interpreted as another step in the legal delimitation of the use of data in artificial intelligence. For now, the case is in an initial phase and leaves many unknowns open. The plaintiff requests a unspecified financial compensation and raises the action on behalf of other potentially affected parties, while Adobe did not respond to Reuters’ request for comment. It will be the judicial process that determines whether the lawsuit is successful, is filed or results in an agreement. Beyond its specific outcome, the litigation once again puts the focus on an issue that remains unresolved: how to balance the advancement of AI with the rights of those who create the content from which it learns. Images | Rubaitul Azad | Adobe In Xataka | Gemini 3 Flash has surpassed GPT-5.2 Extra High in several benchmarks: Google has just changed the rules of the lightweight model

Udio closed fronts with Universal. The creators were then left unable to download their own AI songs

by

Generative music applications have achieved something that seemed unthinkable a few years ago: allowing anyone, with just two prompts, to can produce complete songs with vocals, arrangements and structures that can sound surprisingly real to most who hear them. This experience, which is presented as magical and accessible, has a much less visible side, linked to how these models have been trained and their legal implications. Many of these platforms have turned to large volumes of content available on the weboften copyrighted, to build their systems. The user enjoys the result, creates and shares, until a legal change, an agreement or a lawsuit transforms the tool and the experience is no longer the same. Until just a few weeks ago, udio It was one of the services that best represented that promise of instant creativity. It had managed to attract both curious people and experienced musicians thanks to its simple system, the tools to extend, mix or remake songs and, above all, the possibility of downloading songs for use outside the platform. There was nothing to suggest that this model was about to change. The first indication came when the company began to talk about a “transition phase” linked to new agreements with record companies. It did not yet detail what was going to happen, but it made it clear that the platform was entering a different stage. The day the download button disappeared. Confirmation came when Udio announced thatas part of its transition, audio, video and stem downloads would be disabled for several months. It was a feature that many considered essential, but now they could only play their creations in Udio and share them using links from udio.com. In exchange, the company reported an increase in credits and more generation capacity, although that did not compensate for the feeling of loss. The message was clear: the songs still existed, but they no longer left the walled garden. Warner and Universal chose a different path than the judicial confrontation: turning Udio and Suno into partners rather than adversaries. Universal signed agreements for the next version of Udio to be based on licensed music and offer artists new avenues of income, while Warner did the same with Suno and also sold the Songkick platform to incorporate it into that new ecosystem. Record companies went from denouncing to collaborating, with a clear condition: at least in the case of Warner and Udio, artists and composers would have the possibility of deciding whether their voice, their image or their style could be part of the creations generated by AI. From defendants to partners. Once the content is within the legal space, what is relevant is not only that agreements have been signed, but how the industry’s priorities have changed. A year ago the goal was to put AI platforms on the bench for using protected music to train their models. Today, a growing part of the sector has understood that it may be more profitable to integrate them than to stop them. The move does not eliminate legal conflicts, but it opens the door to a model in which record labels oversee, license and participate in revenue, rather than reacting only through lawsuits. It is a change of focus that signals where the music business is moving. What nobody sees: scraping as the foundation of musical AI. For years, the actual functioning of many generative music models was far from transparent. Some startups, like Suno, admitted to having trained their systems with “virtually every quality music file available on the web,” trusting that such use would be protected by the fair use. However, when record companies began to examine that process, the conflict ceased to be technical and became legal. Images | Universal Music | udio | Unsplash In Xataka | AI has become the best example that if you don’t pay for the product, you are the product

Fed up with excessive luxury, social media users turn to normality: creators with everyday lives

by

A recent television controversy with the content creator @supaa97 has put on the table a series of issues that are perhaps at the opposite end of the topics we always talk about in reference to the influencers (fortunesluxuries, excesses): can content be created from absolute normality? Is that close to normalizing precariousness? And if it does, is it a problem? The Suyapa case. The controversy started, just as Suyapa says (which is his real name), when he agreed to do an interview for ‘Public Mirror’ to comment a video of your profile in which she told how she lived in a single room with her husband and son, and was classified as a “Poverty Influencer”, along with users who make videos with unboxings of government aid. Suyapa has stated that she is far from that type of content, and although it is true that she lives in very modest conditions in a single room, she earns her living by working as a cleaner and without resorting to aid, so she could not be included in a category of poverty. The appeal of normality. Suyapa makes a type of content closer to normcore (which is still a label created from top to bottom): these types of profiles share ordinary activities (from choosing simple and functional clothing to routines such as making a coffee, taking care of a pet or sharing morning tasks) moving away from the cult of luxury or drama that predominates in other digital spheres. They embrace simplicity and naturalness in both fashion and lifestyle: basic garments, discreet brands, homey environments and a staging that is not aspirational but friendly and accessible. He normcore as a label. This type of content is sometimes, as we say, a reaction to more luxurious and frivolous creators. If it arises spontaneously, because the creator does not ascend the social scale even if he wants to (as happens with Suyapa), or as a voluntary limitation, it is another question where you can talk about posture. That is to say, sometimes normcore is a false normality that arises as a reaction to luxury saturation. A more relaxed visual narrative is artificially sought, where the emotional connection is based on trust, identification and everyday honesty, but sometimes it is also a pose that seeks, paradoxically, to convey an image of coherence and credibility. What did they think it was? What ‘Espejo Público’ alluded to and where it mistakenly included @suyapaa97 was in a different type of phenomenon that we know as “pornomiseria” or “poverty porn”, which has two aspects: on the one hand, influencers on social networks that viralize acts of charity towards people in poverty to monetize these contents through likes, views and donations. One of the best known cases is that of Jimmy Dartswho with more than 12 million followers on TikTok, makes videos with homeless people, testing their honesty or proposing challenges. It is a controversial format that has a large number of ethical implications, even though influencers reward the people they portray with a large amount of money, as detailed this article. Something similar happens with amateur journalists who, under the pretext of portraying poverty and misery, create sensationalist content, a format whose origins date back to the seventies and that again has very complex moral connotations. Yonfluencers: from normality to luxury, and back again. Recentlythe rejection of social media consumers to the exaggerated and elitist display of luxury into which many have fallen influencers has made me think in how the perception we have of this type of content creators has changed. Many of them began as a daily reflection of our lives and as they earned money and followers, they distanced themselves from reality, generating a certain aversion from those who followed them for being a close and identifiable replica. That’s why content creators like Suyapa work, who have to overcome obstacles that are easy to identify with: tightening their belts to make ends meet, juggling time off from work or looking for affordable forms of leisure are some of the problems that the vast majority of people face. In Xataka | The influencer María Pombo defends her right not to read. And by the way, it raises an interesting controversy about habits

The new edition of ‘Operación Triunfo’ is being criticized for elitism: it is exactly what its creators were looking for

by

“I study musical theater in London, I have been living for two years in Sweden and four years in India, where I also attended an American school” or “I lived for three years in Indonesia and now I am studying a double degree in audiovisual communication and journalism” are the cover letters of some of the contestants of Operation Triumph 2025. As expected, the networks have not taken long to uncover your memes and make fun of these introductions, accusing the edition of a certain elitism. The truth is, some satire does fit in an edition in which the average contestant is 19 years old and has barely traveled around the world and learned five languages. @javierpv_26 Average OT2025 Contestant #operaciontriumph #OT2025 #otgala0 #ot ♬ Cross The Catwalk – Operación Triunfo 2023 What is quite ironic is that it is mostly Generation Z that caricatures this model of contestant, because this edition, more than ever, has put the post-millennials in front of the mirror, and not precisely because of the challenge that the participants had to read a wall clock. If we combine Prime Video’s interest in increase subscribers and the eagle eye of Noemí Galera, who has been dedicating herself to casting processes for years, we find a format that considers its greatest asset to attract this generation and turn it into its majority audience. The members of the generation millennial We are still part of the fans, but we are no longer the focus of interest of the program. Gone are those contestants from the 2000s and that “I was wrong, olé!” of Rosa de España in the middle of a performance. Now the majority of contestants They have already been through several musical competitions and they know what they undergo in the academy. Visits from famous artists, nor a possible tour, are no longer surprising, and they are more aware than ever of the place where they are. The spontaneous, more natural and the more “working class” profiles have disappeared. In Xataka The return of ‘Operación Triunfo’ starts exclusively on TikTok: it demonstrates everything that the program owes to the social network As almost all of Spain knows, Bustamante, before pursuing his musical dream,worked in construction with his father in his beloved San Vicente de la Barquera; Not to mention the case of Vero, a hairdressing assistant by training and aerobics instructor by profession. In 2025, a profile like this is almost unthinkablebecause those trajectories are no longer those sought by casting directors and become the discordant note of the edition. The clearest example this year is what happens with Guille Toledano, head chef who defines himself as an absolute fan of Atlético de Madrid; His presentation during Gala 0 and in the middle of the rest of his companions was the closest that human beings have been to time travel. Goodbye bricklayer, hello drama student That “two-thousand” essence, where profiles like Chenoa stood out for their mettle and experience on stage, no longer exists. The contestants have more training, but not necessarily many kilometers traveled with their orchestra van or nights worked in casino rooms as was the case with Chenoa herself. The everyday in this new edition, total crystallization of the process that has been subordinating the identification of the public (“I could be that one”) to the search for hyperformed virtuosoare contestants who study double degrees, musical theater, are tap dancers or graduates in Spanish dance and have traveled to countless countries. In Xataka Nobody knows what the audience is for ‘Operacion Triunfo’, but Prime Video has gotten what it wanted with the program We are not saying that these characteristics are what guarantee the success or popularity of each contestant. In fact, the latest OT winner is Naiara, an orchestra singer; and in this new edition Guille himself seems to stand out among the favorites every week. Popularity or success aside, what is striking is that the group of contestants has changed, a reflection of a large part of the generation that OT wants to target in 2025. We are talking about that young audience, mostly university students; a generation responsible for a record number of registrations in 2024, who also studies abroad, knows languages ​​and among his usual leisure is travel more or less recurrently. In this context, how could Tinet Rubira and the entire OT network not design a selection of contestants made tailored for generation Z? Gone are those first editions of Operación Triunfo where the weekly challenge fell to the colleagues who had to defend a song in English and suffered it as if it were Aramaic. And, yes, back in 2001 the public empathized or felt represented listening to Rosa’s tacky English in “It’s Raining Men”but in the middle of 2025 that has nothing to do with reality and, elitist or not, OT knows it perfectly. They want the representation of a modern and prepared generation, with standard English and twice the qualifications that they then ask for on Linkedin. Carlos Lozano giving way to a version of Bad Bunny? This turn and fine-tuning of the contest is not something new. Yes ok This year it has become more evidentthan ever, it was in 2017 when the true transformation of the program occurred, adapting to the codes of the new generations. That OT 2017 with its 24-hour channel and fuming YouTube chat revolutionized a generation that had not experienced the OT phenomenon before. It was nothing more than the beginning, because OT 2023 arrived from the hand of a streaming platform like Prime Video, sensing the facelift to which they wanted to continue subjecting the format. On a personal level, I miss Capde, an OT professor who was there until two editions ago, and who already seemed uncomfortable helping the contestants with certain urban rhythmsbut it is true that Operación Triunfo itself has reinvented itself not only from the casting. Pablo Lluch and Vic Mirallas, young, talented and on the same wavelength as the current program, now … Read more

Grok is one of the best AI in the market. The only problem is that its creators insist that it is not

by

Elon Musk wants to resume the reins of his businesses. In a recent meeting with the workers of XAI, the tycoon communicated to his staff his intention to create “maximalist truth” systems, but also raised two new projects related to artificial intelligence. The truth above all. That old Musk ambition to create an AI that always tells the truth even if that implies not being politically correct now returns strongly. Editors From The New York Times They were able to attend that meeting in which Musk stressed that Xai is “the only company in which the mission is the truth. If the AI ​​forces to lie or believe things that are not true, you are at risk of creating a dystopian future.” Grok and uncertainty. Recent negotiations For new investment rounds in XAI they have made the assessment of this company and its AI model, Grok, rise to 120,000 million dollars. Musk predicted that the product of Xai will quintuplicate the advertising of X to bring them to 10,000 million dollars, but it is not clear what these predictions pass. Meanwhile, the company follows the wake of other Big Tech and is Burning money in ia as if there were no tomorrow. Musk highlighted the XAI connection with the Tesla optimus, which according to the tycoon “will be more productive than the entire global economy.” XAI acquired x With a strange operation that among other things justifies that Grok does not seem to use X data to train. 64 against 700. According to managers who participated in that event Grok currently has 64 million active users per month, while the latest data suggest that Chatgpt has 700 million active users per month, almost 11 times more. Chaotic reorganization. This summer there have been drastic changes in the XAI structure. Musk has totally reorganized the hierarchy of it and has led an effort to recruit new talents and engineers. But at the same time he has thrown several researchers or has seen how others have left for a serious problem. Igor Babuschkin, XAI co -founder with Musk, He left the ranks from Xai in August, like other engineers and that the Excfo de XAI, Mike Liberatore, which has ended signing for OpenAi. Grok does not stop creating controversy. According to sources close to the company, several of the IA researchers who have abandoned Xai have done it because the startup has set aside scientific orientation to fully focus on the development of products that manage to capture user attention. The examples are clear: Source: ARC Prize. But it is a spectacular model in the benchmarks. The strange thing about that obsession with attracting attention is that Grok probably does not need it. Grok 4, the latest version of the model, has proven to be especially strong according to independent platform tests such as LM Arena, LLM STATS, Artificial Analysis either Vellum. Moreover: right now it is the model that best behaves in the very complex (for machines) Benchmark ARC-AGI 2and also surpassed his rivals in July in the mathematical tests From the benchmark of Epoch Ai. That, however, does not always mean that the model is successful, and in Reddit a user He stood out That idea by saying that “disappointment with Grok 4 is proof that benchmarks mean nothing.” New products in sight. At that meeting Musk spoke of creating a Microsoft competitor who nicknamed “Macrohard” (combining the two antonyms of his rival) and that would use Grok to create software. In addition to that, he raised the development of a AI product for children between 2 and 12 years old, which he called “Baby Grok.” This last product is launched or not, it would be a clear response to controversial that lately surround the AI ​​when it is used by minors. Colossus 2. Another of Musk’s megaprojects consists of start Colossus 2, which would be more than three times more powerful than the Current Colossus: Of the 300 MW of this would go to More than one GW Colossus power 2. Musk has been working for months achieve financing This project with 12,000 million dollars, Elon has returned. “Papi is very home.” Thus he declared Elon Musk on Tuesday his apparent definitive return to the direction of his companies, Tesla and Xai. The scarves in the world of politics – which They came out very expensive– They seem to have been left behind, and the tycoon seems to have activated those febrile work cycles that not only (theoretically) He is self -imposedbut imposes everyone who wants to work For your companies. Image | DVIDS | XAI In Xataka | Tesla has a good reason to offer a bill bonus to Elon Musk: it is no longer his golden egg chicken

What is and how the app works to find aid, scholarships and subsidies of the creators of my citizen folder

by

Let’s explain What is and how the app works Help linesa public administration tool created by those responsible for My citizen folder. It is a new tool to centralize procedures, in this case related to aid you can get. We are going to start the article explaining what this application is and what you can do with it. Then we will tell you how it works and how you can use it to request various types of help, either to train or to undertake, independent or other things. What is the Aid lines app Help lines is an official application of public administration. It seeks to centralize information on all kinds of aid that the government gives, as well as scholarships or subsidies. Therefore, it is another application as interesting as My citizen folder either Midgt. In this application you can find subsidies, aid, scholarships and awards from all administrations. Therefore, it is basically an index to find this type of aid. The application has a search engine With various types of filters and options. You can search only in some provinces or specific communities, choose the type of aid, the convening body, or the subject with which it is related. Even if you have already heard of some help you will be able to search for your reference number. For the rest, it is a fairly simple application. Your interface is clear, in each type of help you have A complete description including the deadline to request it, and little else. It serves what it serves, it does well, and that’s it. How to use help lines The first thing you have to do is download application Help linesavailable In Google Play For Android and In the App Store From your iPhone. It is a totally free application, and that you do not need to log in. When opening the app for the first time you will have to accept the privacy agreements and their terms and conditions of use. The application itself is very simple. You only have three sections, one start with information about the latest aids issued, the Search and Favorites section. In the main one, you will see all subsidies, awards, aid, scholarships in general nationwide, as well as its location, responsible administration and the term date. If you enter the search engineyou can make simple searches or use the advanced ones. In the advanced searches section you will have many fields that you can specify, including subjects, type of aid, the convening organ or the geographical scope, in addition to the publication or limit dates. In the geographical field you can choose national, international or those of concrete autonomous communities. Everything to try to refine the type of help to the maximum and why you need it or where. When you press one of the aid you will enter your file. In it you will have all the information, including description of the aid and the beneficiaries to know if you can opt for it. Also the convening organ, the geographical scope or the deadlines. In addition, you will have external links to expand the information entering the aid website or to request it directly. Finally, at the top of the file you have options to share it or to mark help as a favorite and thus be able to access from the corresponding section without having to look for it again. In Xataka Basics | New public transport aid: free fertilizers and with discounts that return to 2025

The True Crime are such success that the creators of the least scrupulous have been thrown over the genre, opening a moral debate

by

He True Crime It is a genre that, Despite its immense popularityis ethically more in question than ever: controversial like The recent book by José Breton They make the public propose the scope and consequences of the documentaries on real events. That now look at a new border, with the creation of content by the based on real events. Or even beyond: completely inventing crimes. The tremendist. “The husband’s secret gay romance with his stepson ends up in a spooky murder” or “the secret romance of a wife with a neighbor’s teenage daughter ends up in a spooky murder” are some of the titles of True Crime Case Files’ videos. It was a channel missing today in which they were related, normally before static images and generated by artificial intelligence, crimes that could go through real. But intentionally, the author concealed the artificial origin not only of the images, but also of the stories, which he generated through Chatgpt. A subgenre. They soon caught the attention, How has 404 averageof journalists from the real locations mentioned in the videos, who were surprised not to have heard of such striking cases. When it was made public that the content was generated by AI, but this was not mentioned anywhere, YouTube canceled this and other channels of its person responsible for breaking the conditions of the platform (among others, “Children’s Security policies, which prohibit the sexualization of minors”). It is still possible to access the content in audio format, however, Through platforms like Spotify. The cartoon is not enough. The head of True Crime Case Files was an AI programmer who had already dedicated himself to generating small parodies of romantic comedies Hallmark type on YouTube o Fill content on social networks such as Facebook. According to 404 Media, he disproportionately exaggerated the details of the cases so that it was evident that it was invented cases but, as with Facebook and Instagram videos obviously generated by thethere are dozens of people who believe them. Other True Crime with Ia. Of course, the case of True Crime Case Files is not isolated, although perhaps it is the only one who has tried to go through real crimes his stories: Cen Stories It has the same visual style, in the images and in the holders, and a disclaimer which warns of the content generated by AI, although the notice appears only in the description, buried between summaries and hashtags, and not within the video itself. In Tiktok they also abound, always specifying its artificial origins, and in innumerable variants. In Detective Challenge The cases are invented, and in Thruecrimeaimedia real facts are mixed with invented. There are channels in Spanish, such as Based cases And others who not only tell crimes, such as my history, but also introduce morbid cases of historical cases, such as Nero’s castrated lover or make popular cases of cinema and television, such as Freddy Krueger’s cases. A catalog of channels, some closer to moral limits than others, but all taking advantage of AI to generate three or four videos a week that accumulate thousands of visits. More garbage. Although all these cases are covered from a legal point of view, they undoubtedly have traits in common with what is known as’Slop‘, Low quality content generated by AI and that floods social networks. The least thing is that it can be distinguished whether it is artificial material or not but how the platforms themselves favor these contents with the algorithm, since The amount interests more than the quality. What is the use of True Crime. To this are added their own limits of True Crimethat since their origin in novels such as ‘in cold blood’ played with the confusion between reality and fantasy. Crime and suspense films have always been inspiring their stories in real cases, which manipulate according to dramatic interests but without accrediting their origins in reality, which perhaps also has moral gray areas that should explore. Often Documentaries are criticized True Crime for exploiting the trauma of the victims and their families and transform your suffering into a showbut the use of AI adds a new layer. Artificial intelligence feeds on documentation of real and fictitious cases, and proposes an indistinguishable amalgam of reality that can become more harmful: sensationalism twists the facts, and the AI ​​does, but in an even more perverse way. Crime ethics. The ethics that are supposed to a media traditional is blurred in these cases, submerged in a misty area where reality and fantasy are confused, but without attending ethical limits, what is worse? A story manufactured but whose characteristics are identical to a real case, or a True Crime that proves its similarities with reality, but that the hurry and lack of means lead to videos that feed the morbidity of the spectators without the need to respond to legal limits? He True Crime It continues to evolve, and not necessarily towards more reassuring areas. Header | Cen Stories In Xataka | The ‘True Crime’ pending black Spain: the cases that television fiction have not yet dared to play

The human voice loses ground and automatic dubbing with AI makes its way between creators

by

Artificial intelligence (AI) continues to gain space in unthinkable sectors just a few years ago. What began revolutionizing software development is now leaving a mark on very different fields. The phenomenon of ‘Vibe Coding‘, a new way of programming based on Prompts In natural language, without writing code as such, the rules of development are changing. At the same time, tools such as Dall · e of OpenAi or Adobe Firefly They are redefining visual creation, allowing generating images and videos from textual descriptions. And now, AI also reaches voice professionals. For a few months, YouTube allows some creators to use automatic dubbing through AI. This function, which is being deployed progressively, is now available for all members of the YouTube Partner Programa system that allows creators to monetize their contents on the platform. Thanks to this tool, videos can reach global audiences more easily, as an alternative to traditional dubbing and postproduction processes. An activated function by default. One of the most striking changes is that this function is activated by default: if you upload an English video, YouTube will automatically double it in several languages, including Spanish, French, German, Hindi, Indonesian, Italian, Japanese and Portuguese. If the video is in any of those languages, an English version will be generated. The objective, according to the company, is to “break the barriers of the language.” But if the creator prefers not to use this function, it can deactivate it following concrete steps: enter YouTube Studio from the computer and follow this route: Configuration > Subsequent adjustments > Advanced configuration. There you have to uncheck the ‘allow automatic dubbing’ box and click on store. The management of this function, in addition, can only be done from YouTube Studio on a computer, and allows you to even manually review dubbing before publishing them if you want greater control of the result. YouTube admits that technology still has improvement margin. Translations may not be exact and the generated voice may not faithfully represent the author. In fact, the company itself recognizes that intonation, emotional tone and certain cultural nuances do not always move correctly. The basis of this function is in the technology of Google Deepmind And Google Translate, but even with that background infrastructure, not all videos can be folded successfully. Factors such as the original accent, the background noise, the use of jargon or even proper names can affect the quality of dubbing. A video with automatic YouTube dubbing To know if a video is bent with AI, it is enough to look for the “auto-dubbed” label. It is also possible to change audio track from the player’s menu, through the gear icon in the lower right corner. The language adapts to that the user has configured as preferential, which makes the experience, in many cases, more transparent. We can see the active function in this example video. The prelude: real dubbing with professional actors. YouTube had already taken previous steps in this direction. With the function “Add audio track“, Some creators like Mrbeast They began to offer versions of their videos in several languages using professional voice actors. He even counted With the dubbing actress that plays Naruto in his videos in Japanese. This strategy, focused on expanding the scope of the content without losing interpretive quality, allowed many of these videos to reach new audiences. Mrbeast has not been the only. According to YouTube datathe videos that offered versions in several languages ​​made more than 15% of the visualization time come from audiences that saw the content in a language other than the original. In January 2023, users came to consume more than 2 million hours of folded content every day. And that only with the initial tests. The extension of this function to all members of the Partner Program is contributing to the growth of the use of voices generated by artificial intelligence on the platform. This raises questions about how this technology will be integrated into the professional dubbing ecosystem. Some creators could opt for this automatic alternative in certain types of content, especially in informative or structured pieces. However, in other formats where vocal interpretation provides key nuances, the role of dubbing professionals remains a difficult element to replace. Development continues, and the balance between technology and creativity is still evolving. Images | Lorenzi | YouTube screen capture In Xataka | Two students created an AI to “cheat” in work interviews: they have rewarded them with 5.3 million dollars

Now its creators want to carry the innovative concept much further

by

When someone steps on the Vegas for the first time, it is wrapped by a dazzling show: sources that dance to the rhythm of music, imposing structures and a frenzy that it never seems to go out. But beyond the casinos and bets, there are corners that offer a difficult experience: viewpoints that challenge the vertigo, museums that unravel their history and shows that lead entertainment to another level. A giant screen in the middle of the desert. The Sphere is not just an auditorium, but a monumental structure Designed to offer shows in a completely new way. Its most striking feature is its double LED screens system, which covers both the interior and outside of the building. Inside, the enclosure is equipped with a 15,000 square meter LED screen, an enveloping surface that allows the space to be transformed into any imaginable scenario, from a starry sky to a futuristic city. Outside, the building’s exosphere has an even larger LED screen, 54,000 square meters. Both were designed by Saco Technologies, a Canadian company specialized in LED video screens and lighting. U2 at the inauguration of The Sphere Since its inauguration in September 2023 with a U2 concertThe Sphere has attracted attention both for its size and its technological deployment. During the day, the structure shows great distance images, and at night, its interior houses shows designed to take advantage of its enveloping screen. With capacity for 20,000 people, its format is rare in the entertainment industry. The cost of costs. Developing such a project was not cheap. The Sphere’s initial budget was 1.2 billion dollars, But it ended up costing 2.3 billionalmost twice as planned. To that we must add another problem: not any artist can play there. Unlike a stadium or conventional theater, a show in The Sphere requires audiovisual content specifically designed for its screens. This implies an additional investment in production, which limits the number of events. The Sphere de Las Vegas in full construction With these costs in mind, replicating the model in other cities seems complicated. Few companies or governments would be willing to assume such a high investment without guaranteeing money. However, Sphere Entertainment believes having found a solution. The commitment to smaller versions. To make the most viable model, the company has announced the development of smaller versions of The Sphere. According to Av MagazineSphere Entertainment revealed in a call with investors that works in ‘Mini-Spheres’ with capacity for 5,000 people. These versions will reduce both construction costs and show production, maintaining the immersive experience of the original enclosure. It has not yet been confirmed in which cities these new enclosures will be built, but the possibility that some are located outside the United States is real. Abu Dhabi, in fact, It will be the first In having a large -scale version, although it still remains to be seen when the project will materialize. In any case, The Sphere makes it clear that immersive entertainment has potential, but its expansion will depend on whether the company manages to make its model sustainable without the costs making it unfeasible. Images | SOUNDERBRUCE | LISETTE ( CC by-SA 4.0) | Sphere At The Venetian Resort (Instagram) In Xataka | A man has climbed the top of The Sphere in Las Vegas. No one knows how he got it

We spoke with the creators of ALIA, the 100% Spanish AI, to understand its future

by

This Monday it was announced release of ALIA language models. The initiative has been in development for years and it is now that the first fruits are beginning to be seen, still modest, but promising. To learn more details about ALIA, at Xataka we have spoken with Martha Villegas (@MartaVillegasM), head of the Language Technologies Unit of the Barcelona Supercomputing Center (BSC). This has allowed us to clarify the status of the project, its objectives and its next challenges. To compete with ChatGPT, nothing The first thing we wanted to know is how ALIA had been created, and here Marta Villegas clarified that the model is based on the Llama architecture – Meta’s Open Source model –, “but the model has been trained from scratch and with zero initial weights“. This is important because ALIA is not a Llama-based model that has undergone a refinement or “fine-tuning” process. In those cases, this expert explained, “you start from a model trained with other data and with initialized weights, and you do it to adapt that model to your needs, either because you have more data and you want it to be better or because perhaps you want to adapt it to a particular domain. But here, he told us, “the vocabulary (set of tokens) is completely different.” In other models the corpus or training data set may be mostly in English, which causes the set of admissible tokens to be calculated through English. That, Villegas indicates, would make it adapt less efficiently to other languages. That is precisely what has been sought with ALIA: reduce the relevance of English to increase the number of 35 languages ​​of the European Union and, especially, Spanish, Catalan, Basque and Galician. How ALIA has been trained The ALIA training process began with some experiments in April 2024. It is necessary because as Villegas explained, “training is not pressing the button after feeding the data and that’s it.” It had to be taken into account that MareNostrum 5the supercomputer located at and managed by the BSC, had just come into operation at full power and there was high demand to use it. MareNostrum 5 In this training process, the ALIA project has had gradual availability of the computing capacity of MareNostrum 5. Although for a short period of time they had access to 512 of the 1,120 specialized nodes of the supercomputer, 256 nodes were used for many months and since September They are using 128 nodes, “which is a lot,” Villegas highlights. During the training process, he told us, there are so-called “checkpoints”, in which it is possible to evaluate how the training process is going. These “pauses” also allow certain training data to be updated, as in fact happened in that process in which at a given moment they introduced a new corpus with high quality that allowed them to replace some data they had. This is just the beginning: it’s time to “instruct” and “align” ALIA Villegas explained to us that ALIA is a foundational model: it is not prepared to be an alternative to ChatGPT. The latter is based on GPT-4, a much more ambitious foundational model that involved much more investment. Here we must differentiate the foundational model from the “educated” and “aligned” models with which we usually interact. As this expert told us, “ALIA-40b is a foundational model that is not instructed or aligned. For a model to be a ChatGPT and understand the conversation and have a certain memory and be “politically correct,” the foundational model (which only learns to say the next token) is “instructed” by passing a bunch of texts.” Even so, the goal is to gradually consider these options. “In March, the instructed version of ALIA-40b is expected to be launched, with a first set of open instructions,” Villegas told us. These instructions are going to be subcontracted – the ones that allow these models to be instructed – and a million euros are going to be invested in that set of instructions from scratch. This data will also be published so that it is available to institutions and developers: if it has been paid with public money, explains Villegas, it is logical that this data will also be public, something that does not usually happen with other AI models from private companies. While training AI models provides guidance on how to respond and defines the context and purpose of those responses, alignment solves problems such as avoid discriminatory biasprevent misinformation or protect privacy. Precisely this lack of alignment means that using these models in this initial phase can produce responses with errors and biases that are precisely mitigated to a great extent with this alignment phase. ALIA and the competition: it is neither a rival of ChatGPT nor does it intend to be In fact, Villegas highlights, “the objective is not to compete with ChatGPT, for that we would need 5 billion dollars.” ALIA-40b “is a good model, and a chatbot can be made in the future because the intention is to instruct and align it, but that will take time.” Within the ALIA family we have the Salamandra models (2b and 7b), smaller and more modest but which already have first instructed versions. Its performance and capacity still have room for improvement, but they are good starting points for the future. It was inevitable to ask how ALIA then intends to compete with other models, both closed and developed by private companies and Open Source models. For her “There is a demand for intermediate models that each person can then adapt to their specific use case, not everyone can use ChatGPT for reasons such as privacy or use case.” Villegas also wanted to highlight how these smaller models can have exceptional performance in specific tasks, and can work at levels of security and not sharing important data. The objective is not to compete with ChatGPT, for that we would need 5,000 million dollars Not only that, he reveals: “we also took out the … Read more

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.