million

The filtration of 16,000 million passwords would be the largest in history. If it weren’t because it’s a gigantic refrito

by

A team of CYBERNEWS researchers He has discovered A large library with 16,000 million filtered passwords. That would be long the theft of credentials in history, but it has not been. That library is actually a compilation of 30 large sets of data already filtered in the past. Should I reassure us “just” that? Enough, but not quite. It is not a new massive theft of data. Cybernews researchers do not show their sources, but they point out that this is in essence a compilation of “supermassive data sets containing billions of credentials.” In those leaks – ancient, but it is not known how much – data appears that according to Cybernews come from Google, Facebook or Apple users. Again, without offering evidence of it. They explain that none of these data sets had been discovered before, except for one with 184 million credentials that They mentioned in Wired At the end of May. But it is still dangerous. The problem of a compilation like this is that it can serve as a great starting point for all types of cyber attacks. It is very likely that in that large database there will be a significant amount of duplicate credentials. No one has stolen data from Google, Facebook or Apple (recently). Although media such as cybernews or forbes talk that there are credentials of these platforms, again if any will be belonging to old data theft – which Of course Yeah There has been-. There have been no robberies (much less massive) of credentials in these companies recently, and if there were them, they should have made a communication responsible for the incidents – in the US the US is used FORM 8-K OF THE SEC– They can affect their businesses. Should I be worried? Relatively. It must be aware that these data theft give access to cybercriminals to millions of credentials (user and password) in all types of services, and they can gain access to our accounts in those services thanks to them and then attack them or use them as the basis of other attacks. But this is true every day, not only when a new security filtration appears. Have Ien Pwned is a useful service that allows us to enter an email address to check it with the databases of previous data theft. A curiosity here: the number of accounts hacked on this platform is almost 15,000 million, very similar to the one indicated in Cybernews. Check if you have hacked you. There has been a simple way to check if our email and password has leaked in any of these data theft. THE SERVICE HAVE I BEEN PWNED allows us Do it quicklyand here we have discovered something interesting: the database with which this platform works has almost 15,000 million accounts hacked in its registry. Almost the same number they are talking about in Cybernews, as if their great library was in essence the same one that has this service created by Troy Hunt cybersecurity expert. And acts accordingly. When introducing your email, services may appear in which your account may have been compromised. What you should do in case this happens is to change your password as soon as possible in those services, and, if possible, strengthen the security of that password. There are several methods to do so: you can create a strong passworduse a Password manageruse the two -step authentication “Specially recommended,” or even make the leap to the passkeys If the service allows it. Be careful out there. Those measures that we talk about should not only be applied if we realize that something has happened, but before it happens. It is better to prevent than healing, and it is worth reviewing the safety of our accounts on the Internet, especially the most sensitive. The mail (Gmail, for example) is especially delicate, since it is usually used as a password rescue method in many other services. It is a good idea here to go to the tools that allow verifying the security status of our accounts to reinforce it if possible. Google, for example, Maintain a control panel which we can access at all times to configure what is necessary and Avoid scares after. In Xataka | The old dream of the cybersecurity of the future has just come true: Microsoft corner finally the passwords

They had stolen 1.2 million euros

by

The most spectacular robbery in June has had an unusual victim: Nintendo. A truck that transported almost 3,000 units of its last console It has been devaluedwhich can be a problem for an apparatus that is already having Supply problems. Although perhaps the biggest problem is going to have thieves to give them out. More than one million euros. That is what the almost 3,000 consoles cost Switch 2 stolen on June 9 of a truck in Colorado. The truck driver did not know what goods he transported, only that it was games and toys. He was on the way from the headquarters of Nintendo of America in Redmond, Washington, to a Gamestop branch in Grapevine, Texas. Caught Possibly, thieves have difficulty getting all the desirable game to consoles: Nintendo will have perfectly located the serial numbers of the consoles. If someone connects to Internet or uses an account with them, they can disable them at a distance without any difficulty, so the consoles will have to be cracked using a Jailbreak not authorized to convert them into functional consoles. Of course, this last procedure involves its risks, and the user could meet a luxurious pisapapeles if something fails. To the Dark Web. As usual in these cases, consoles (unless thieves want to resort to direct scam and sell consoles that know that unusable as soon as they connect to the Internet) possibly move For the Dark Webselling them at much lower prices than the originals. There is another solution: selling the consoles outside the United States, although that does not guarantee that they will not be blocked by Nintendo and, in any case, it could imply an size in the robbery that perhaps it was not so planned. It is not found. The desired Nintendo console, which despite the precautions that the Japanese company is taking are experiencing supply problems, has become the most desired technological product of the moment, and has quickly become the launch of a console most successful in history. This robbery adds a peculiar mythical desired product to the launch. More and more steal is complicated. When a technological device is stolen, there are several systems that can lead to useless. For example, in the case of mobile phones, the IMEI (International Mobile Equipment Identity) is a unique code that identifies each device, which allows the company to Distance block. This prevents the phone from connecting to mobile networks and disables it for calls, messages and data even if the SIM card is changed. Something could make Nintendo with these consoles. With the church we have encountered. Nintendo takes the Protection of your devices and, in general, of their legal rights. It is one of the companies that has most exhaustively fought against unauthorized copy of your gamessuch as cartridges, USB memories and modchips, which allow playing roms. Nintendo has obtained numerous judicial sentences in their favor worldwide in that sense. Nintendo also supports these efforts with unique certificates and tickets linked to both the device and the Nintendo account to identify unauthorized users, to which regular software updates of their consoles are added to close vulnerability. It has reached to change hardware components (such as chip Mariko in Switch) to hinder the Homebrew and piracy. Header | Arapahoe County Sheriff Office In Xataka | I have tried the Nintendo Switch 2 and I already have an answer to the question “Is it worth buying it if I have the original switch”?

It’s no longer about doing better chips, but connecting half a million

by

The new battle of the AI ​​is no longer earned only with integrated circuits, but orchestrating complete systems that connect half a million chips working at the same time. Why is it important. For decades, the technology industry has been governed by the Moore law: The number of transistors in a chip doubled every two years. Today, calculation power advances come from the ability to connect endless amounts of chips between them, an effort that does not follow any predictable “law”. The turn. AMD has presented Your mi400 chip Not as an isolated piece, but as a complete system called helios Rack of the pre -installed data center. It is the implicit recognition that the era of individual chips is over. The context. Nvidia dominates the AI ​​because Jensen Huang had the vision of using GPUS for these calculations, which gave him the advantage of those who hit first. But AMD is shortearing fast distances (Bet on smooth his usually goes wrong), and the inherent change of the industry has also changed competition. Between the lines. Companies no longer hire chips engineers, but also more concrete profiles: Specialists in Design of racks. Systems architecture specialists. Specialists in data centers optimization. AMD is now using More than 1,000 specialized engineers in design of racks. Architecture chiefs who previously focused on concrete domains should now be somewhat more general to create more integrated systems. The threat. The biggest competitive pit (‘Moat‘) Nvidia is no longer hardware, but software. His suite CUDA It is the standard in the AI ​​industry to train models. AMD competes with ROCM. And although changing system can be too expensive, especially for small companies or independent developers, large companies can justify spending if the greatest efficiency gives them the right return. Yes, but. Companies are packing so much hardware in the racks that High soils and electrical networks They begin to be too tension. The question here is when they can sustain this race without finding physical limits. Deepen. The CEO of COPEAidan Gomez, said In the event that was “very easy” to migrate one of your company’s models to the AMD platform. Openai, Oracle and other great technological ones are looking for alternatives to Nvidia since he boom of chatgpt began to make your chips scarce. Supplier diversification is also strategic for projects like Stargatethe initiative of half a billion Openai dollars next to Oracle. Outstanding image | İsmail Enes Ayhan in Unspash In Xataka | Taiwan has decided to prevent by all means that TSMC technology arrives in China. Your future is at stake

Israel is not only attacking Iran with missiles. He also just steal 90 million dollars in cryptocurrencies

by

It is not much less cryptocurrency theft more large in historybut those 90 million dollars are especially significant for whom they are involved and why. Those involved They are Israel and Iran. And why, unfortunately, The war that they maintain both nations. Israel hacking to Iran. The war between Israel and Iran is intensified, and does not only New and striking tactics On the battlefield. It also does it through cyber warmen. A group called Gonjeshke Darande, (predatory sparrow, in a translation from the Persian) Hackeo has been attributed of the Iranian market of Cryptodivisas Nobitex. Possible link to Israel. There is no definitive evidence of a direct link between the Government of Israel and this grip hacker, Sophos experts indicate. Rafe Pilling, Director of Intelligence of Threats in that firm, explained that the cyber attack had all the characteristics of an attack supported by a government. 90? Millions of dollars. The Hacker Group has achieved according to The Guardian steal 90 million dollars, although the page of the Wikipedia Persian edition It indicates that the robbery amounted to 3.76 billion rials, about 47 million dollars, although that money may be stolen from one of the two troonscan wallets destination From Gonjeshke Darende they have also threatened to publish both the company’s internea information and the source code of its cryptodivsis purchase platform. The final amount is not entirely clear, and According to Cointegraph It is exceeding 81.7 million dollars. Cold purses, safe. The attack, They support Nobitexhas allowed to steal the cryptocurrencies of the “hot” purses of the platform, used to facilitate daily transactions. He has not affected cold, safer purses. Nubitex blocked access to the platform as soon as they and those responsible say that “all damages will be compensated through the insurance fund.” In a later message They have revealed that the impact of the attack is “more complex than it was initially estimated.” And above, Internet cuts. From Nobitex they explain that their capacity to respond to cyber attack has also been affected by the cuts in the country’s internet infrastructure, “together with limited access to the facilities due to the current national crisis.” They hope to recover and restore their services in the next 4 or 5 days, but in the meantime the platform is still unable to be accessed. A hacking with political motivation.Yehor Rudytsia, security researcher at the Hacken firm, said in Cintelegraph how this cyber attack is more “a political statement than a robbery with economic motivation.” In fact, according to The Guardian the hackers have “burned” those funds storing them in custom addresses (“Vanity Addresses“) that they do not have a known private password or possibility of recovering. For example, a purse such as” 0x0000000000000000000000000000000000000000000000 Transferring cryptocurrencies to this type of addresses is actually destroying them voluntarily by leaving them blocked forever. The hackers have used directions with variations of the term “Jo *** Osterrorists”. Image | Wikipedia | Art Rachen In Xataka | Iran and Israel are starting another war in the background: that of the false images created with AI

YouTube risked a lot trying to be more like Tiktok. 200,000 million visualizations have proved him right

by

The future of the video is brief. You just have to look at the disproportionate growth of short video platforms such as The Almighty Tiktokwith 1.6 billion users on a global scale and a sustained growth that will lead to 113.2 million in the US in 2023 to 121.1 in 2027. So it has nothing strange that the rest of video apps go behind, behind, exactly the same with Netflix and its competitors in streaming. Many of Tiktok’s rivals are going well: YouTube, for example, not to grow thanks to its open tiktokization. Shorts better. YouTube Shortsthe section of short videos of the platform (no more than 60 seconds – then extended to three minutes – and with vertical format) has reached a new maximum when averaging 200,000 million daily viewsas announced by the CEO of YouTube, Neal Mohan, during his speech at Cannes Lions International Festival of Creativity 2025. It represents a growth of no less than 186% compared to the previous year: in March 2024 the platform spoke of 70,000 million daily visualizations from the launch of Shorts in 2021. Now, with Ia. In this growth context, Mohan made another announcement that can make an important difference between Tiktok and YouTube: the arrival of I see 3 To YouTube Shorts this summer. The Google video generation model will allow users to create funds and clips, something that already did the previous version, Dream Screen, but I see 3 exceeds it in many aspects, from the technical quality to the audio implementation also generated by AI. The usual youtube is eaten. The rise of YouTube Shorts has been generating a debate about the possible “cannibalization” of the traditional long YouTube content. As in Tiktok, Shorts favors content discovery and rapid consumptionwhich has captured the attention of younger audiences and with mobile devices as the main platform. The expansion of the maximum short duration of 3 minutes in 2025 reduced the gap between both formats and has allowed content creators to develop more complex ideas, tending a bridge between the traditional YouTube and shorts and between Shorts and Tiktok, a platform that has been experiencing with the Creation of videos of up to 30 minutes. He lacks a boil. Although growth is unstoppable, the truth is that Tiktok has an unmatched capacity to discover new videos to the user. All thanks to a much more refined algorithm, as anyone who spends some time on both platforms and demands not an avalanche of videos for anyone, but content thought a little more specifically for him. At the moment the trends, musical discoveries or memes continue to be born in Tiktok, and shorts are following the wake, still without algorithm. There are multiple reasons for it to be like this: The Tiktok algorithm Analyze thousands of user behavior signals (likes, comments, visualization time, repetitions, all to customize the content that appears in you). The possibility that any creator, however unknown, is viralize is what favors experimentation and creativity. In that same way, applied to music, Tiktok has revolutionized how songs are discovered and popularized: The platform is now the main way for new themes and artists to be known, thanks to how the platform reuses music, favoring the creation of Challengesmemes or choreographies. There is a game. YouTube, however, has some advantages over Tiktok, all structural and business. On the one hand is the eternal fear that Tiktok is prohibited In territories such as the United States, which would allow Shorts (and Instagram) to occupy that space. Besides, Monetization is much more worked On YouTube, where money reaches the creators more robust and predictable. It is a negligible detail, and has made Influencers and Streamers They end up returning to the platform after passing through others, thanks to the advantages offered by YouTube (and despite the inevitable ups and downs). In Xataka | The PAU is over, so the true students’ festival has begun: react to their notes on Tiktok

The Nvidia IA supercomputer costs three million dollars. And to function wears a switch with three km cable

by

When Nvidia presented her new AI chips, The B200 with Blackwell architecturetook the opportunity to present an AI accelerator called GB200. And by joining 36 of those accelerators created its AI server, the monstrous DGX GB200 NVL72, which also keeps some spectacular surprises. Each node is bestial. Each of those GB200 accelerators has a CPU Nvidia Grace with 72 ARM Neaven V2 nuclei and two B200 GPUS. By combining its power we end up having a kind of bestial GPU combined with a power of 1.44 Exaflops in precision FP4. A closet that weighs a quintal. The appearance of the GB200 NVL72 DGX is that of a small and narrow closet that is above all very dense: this rack weighs 1.36 tons. Inside there are 18 Bianca computing nodes in 1u format, and each of them has two GB200, or what is the same, with four B200 GPUS (hence 18 x 4 = 72). He estimated cost of this AI server is about three million. Liquid cooling is key. The heat dissipated by these components is remarkable, which makes in this case the best option to cool those elements is the liquid cooling. This system not only applies to the CPU Grace or in the B200 GPUS, but in the NVLink chips of the switches, which can also be heated a lot due to the massive transfer of data between the accelerators. Interconnections everywhere. For all these GPUS to work together, each of the 36 GB200 has specialized network cards with NVLINK support of fifth generation that allow each of the computer nodes to be connected to others. For this there are nine switches that provide that huge amount of interconnections. 3 km cable. The system allows you to enjoy a bidirectional bandwidth of 1.8 TB/s between the 72 server GPUS. But as they point out In The Registerthe really surprising thing is that in total inside that “closet” there are 3.2 kilometers of copper cable. Only the module with the switches weighs more than 30 kilograms due to both these components and the more than 5,000 cables that are used so that all Nvidia GPUS work together and in perfect synchrony. Why copper? It may be able to opt for copper cable seems strange, especially taking into account the needs in terms of bandwidth imposed by this machine. However, the solution with fiber optic cables imposed clear problems: we would have to use electronic components necessary to stabilize and convert optical signals. That would have increased not only the cost, but the consumption of the final system. Can Crysis run? The performance of each B200 chip It is already brutal on its own: Its power is the triple than that of the GeForce RTX 5090, and the entire server includes 72 of these specialized GPUSs for AI, which demonstrates the computing capacity that said machine possesses. It also has RT (Ray-Training) nuclei of the fourth generation, which would theoretically allow you to use these AI chips to play video games, although of course that is not even its purpose. In fact your performance in this area will probably be almost as poor as the Nvidia H100. Cloud consumption. Although new chips are much more efficient than H100 –25 times less, says Nvidia – this AI server has an estimated TDP of 140 kW. Since the average consumption of an average home in Spain round The 3,000 kWh per year, in an hour of use of the Nvidia server we consume the same as an average Spanish home in 17 days. Have it on and running all year raises a consumption similar to 415 middle homes throughout the year in Spain. In Xataka | AMD has a splendid roadmap for its AI chips. The problem is still in your software

How to manufacture five million tires every year

by

Between the Plaza Mayor of Valladolid and the Bibendum roundabout there are 4.1 kilometers and thousands of distance lives. Exactly those who work, have worked or work at the Michelin factory in the city of Pisuerga who, Like Renaultis one of the great engines of the city. The union with the city is so great that “in fact, the idea of ​​calling Bibendum Glorieta at the entrance of the factory left the City Council. We had a generic name of industrial polygon. We wanted to change it and they themselves proposed the name,” they point to us from the company whose doors proudly look its famous Pet in a gigantic size. And it is that 2,200 workers pass every day (almost 1,700 of them are part of the workforce and another 500 work in maintenance or cleaning, among other functions) that keep the plant to full performance, with three eight -hour shifts that keep alive an annual production of five million tires. The objective is clear: fill the 50 trucks that come out every day. Valladolid is not one more factory for the company. It is one of the so -called “digital factories” for the high investment that has been made in all types of systems to carry out a controlled production to the extreme. This investment is manifested in the machinery that gives life to gums for cars, trucks and tractors that are manufactured there but also for the control of the entire assembly line. It is there, exactly, where our visit begins. A control tower We could say that the Valladolid machine room is the opposite of what we could imagine. The Castellanoleonese plant moves to the rhythm of two operators that supervise everything that happens locked in an office. By hand they have two computers with multiple monitorson the wall the screens are multiplied with cameras distributed throughout the facilities, colored notices of all kinds and a volume of numbers, names and information for profane eyes. Our guide points to a side wall in which we had not repaired given the communicative epilepsy we had in front of us. Simply, lines and bulbs. This worked before and thus still controls the management of the inventory or the situation of the machines in some plants. Valladolid, they explain to us, have some of the most advanced control and supervision systems and, in fact, have worked to implement some of the solutions in other plants of the company. The intention is clear: all unforeseen must be controlled at the time for the production chain to stop the minimum and essential. From a breakdown to the shortage of a product. We leave that luck of control tower to put the feet on the mainland. In front we have the compounds that give life to a rubber. They are so simple that they look like a children’s game. Rubber, a metallic ring, the fabric that structures the wheel and the rolling band. So simple that it seems a lie that we talk about a product capable of resisting the force exerted by a car that rolls more than 100 km/h, a truck that transports tons of merchandise or a tractor that, with the agricultural tires More advanced, it can circulate at 80 km/h. “In the United States it is legal for a large tractor to circulate at that speed.” What is unthinkable here “creates a problem when thinking about how to face wear. So we have a system so that, from the tractor itself, the tire pressure is increased so that a smaller surface touch the ground when it moves through the asphalt. And, on the contrary, you can lower the pressure to gain trace when it gets into the field,” they explain. This way of working, without leaving from the tractor cabin, is a solution like the one that They use the pilots in the Dakar Rally That, from inside the car, they put pressure if they want to gain speed on a fast track or remove it if they need to cross a dune since it requires greater contact with the ground to have greater traction capacity. From the rubber to the road This is the key to everything. With the hand, our guide holds the tire rubber. “Each rubber has its own mixture and is secret,” they point to us. It is about making the most balanced rubber for each vehicle and moment. Not only is it a matter of offering the best grip, you also have to get the slightest rolling resistance (lower consumption) or less noise. And those concepts are enemies of braking, traction and, of course, performance. Join those Three pillars In a single tire is the biggest challenge, prioritizing with each type of product one or another game. For example, a Michelin Primacy 5the tire that we have come to know and that is designed for great general performance and good behavior in summer, that a Michelin CrossClimatedesigned to replace the chains in case of Nevada, or a Michelin Pilot Sportfor sports driving. The heart of each has its own mixture of ingredients (they add up to 200,000 tons of rubber per year). And that mixture enters through a machine that is in charge of flattening and shaping the compound. The mixture is heated, flattened and forms the heart of the tire. That already ironed rubber is painted with brands to facilitate the subsequent process management. As if it were a very long plasticine without principle or end, the machines are cutting into strips and barrels serve to generate the rounded shape. This is where the metallic ring and textile meshes are inserted in charge of structuring and supporting the vehicle pressures. The heavier, the greater the resistance the mesh should exert, so in the specific models for SUV or electric cars, two meshes are usually included. Finally, everything is covered with a last layer that will form the rolling band. This is drawn in the process of Vulcanizedat which time the tire … Read more

The San José Galeon sank in 1708 with a treasure of 20,000 million. A handful of coins has revealed its destiny

by

Throughout the planet it is estimated that there are close to three million Of pecios, vessels that caught in their day and now rest at the bottom of the oceans, turned into sentences of marine life. This list is included from transatlantic as The Titanic to destroyers of World War II, abandoned boats or colonial caravels. Few arouse the interest of San Joséthe galleon sunk in 1708 off the coast of Colombia with the wineries full of gold, silver and gems, a treasure that some value in almost 20,000 million of dollars. His exact whereabouts was a mystery for centuries. Now Colombian archaeologists believe Having located it without a doubt thanks to a track that confirms the opulence of his treasure: the gold coins that he transported on board. A Milmillonario Treasury. In his day the San José was an imposing galeon, a large ship of 40 meters of length12 of manga, three masts and 64 cannons built in Guipúzcoa commissioned by King Carlos II. Your mission: work on the Indian fleet. Neither its power nor its dimension avoided however that the history of San José was brief. The ship ended up sinking from the coast of Colombia in June 1708, only a few years after its launch, beaten by the cannons of a squad of British privateers during the Battle of Barúin full war of Spanish succession. The most curious thing is that the legend of San José began just then, after its shipwreck. And the reason is simple: in addition to a crew formed by hundreds of sailors, the galleon loaded with gold, silver and gems, among other treasures. There is talk of a loot of 200 tons that today would reach a value of billions of dollars. Some estimates place it in 17,000 millionothers in more than 20,000. There are also those that reduce that calculation, but without leaving the land of the thousands of millionsan imposing treasure. And where is the San José? For centuries that was the great unknown. It was known that the wreck was In the Colombian Caribbeanbut … where exactly? What were your coordinates? In 1981 A company announced to hype and saucer having found the Galeon and allegedly delivered the information to the government in exchange for keeping a part of the treasure, but the story soon complicated. In 2015 The country’s authorities claimed to have located the remains of the Spanish ship in a different place, which tightened the disputes about who has the merit of the finding and (more importantly) the rights over the treasure. The big clue: the coins. Historical discoveries often depend on small details. And underwater archeology is no exception. Although experts have been convinced that the vestiges resting in front of the Colombian coast are the remains of San José, a New research published in Antiquity He has just reaffirmed the identity of the wreck. And in a fairly peculiar way: analyzing the gold coins located in the underwater deposit, at approximately 600 meters deep To get it a Colombian team undertook Between 2022 and 2024 Several expeditions focused on the remains of the wreck. He did it With the help of a rova non -manned submarine vehicle that allowed experts to obtain high resolution photos of the coins that rest on the site. Thanks to techniques such as The photogrammetry They were also able to draw a three -dimensional reconstruction of the wreck and several models and digital replicas of the environment. And what did they discover? They found Macuquinashand coined coins and that were used for more than two centuries in the trade of the Americas. With the help of high resolution photos taken in situ – the Colombian Institute of Anthropology and History assures That no object object was extracted – experts discovered that the pieces have an average diameter of 32.5 mm, weigh 27 grams and are coined with revealing designs. On the obverse, the coins show a variant of The Jerusalem Cross (A large cross with four smaller) next to a shield with castles and lions, symbol of Castilla y León. On the back they carry the Hercules crowned columns On waves, which relates the pieces to the Lima Mint. Squeezing clues. They are not the only clues identified by archaeologists, who have identified a “L” that seems to refer to Lima, an “8” that shows the value of the currency and an “H” related to the brand of Francisco de Hurtadohe Major trial of Indies in 1707. In the central part of the coins the legend “PVA”, “Plus Ultra” is also appreciated, a wink to the maritime expansion of the Spanish monarchy. The information is completed with three digits (707) located at the bottom of the pillars, a reference to the year in which the coins were coined: 1707, just before the San José sink into the Caribbean. Why is it important? Because all these data are clues that help to better understand the site and the circumstances in which the galleon that rests there. “Unraveling the characteristics of the currencies helps to determine the age and origin of the wreck”, collect the article. “First, it establishes a temporal frame: the sinking had to occur after 1707. This, along with other tests, such as the presence of Chinese porcelain Kangxi And inscriptions in the cannons dating from 1665, suggests that the ship sank at the beginning of the XVIII. “ “Corroborate identification”. The coins also give clues on the route that must have followed the sunken galleon. “In the Viceroyalty of Peru, several gold mines were registered, mainly in Puno and Huamanga. It is likely that the material to coined the coins would be obtained from these mines and processed in the house of La Moneda de Lima, which resumed operations in 1683 and began to coined gold in 1696,” Experts reveal In your study. All those evidence, added to what historians already know about colonial history, leads them to A resounding conclusion: “The set of tests corroborates … Read more

A Basque startup of AI has just lifted 189 million euros with a great idea: compress the AI

by

Before We compressed files with Zip. Now what we begin to need is to compress the AI ​​to make it more small and efficient. That is just the idea that the founders of Multivrse Computing had, a Spanish startup which is becoming the new jewel of the crown of our AI industry. Its founders, (in the image, from left to right, Román Orús, Enrique Lizaso Olmos and Samuel Mugel) and Alfonso Rubio have much to celebrate. Investment Round. Multivize Computing He has just closed an investment round of 189 million euros (215 million dollars). The round (series B) has been led by Bullhound Capital, but it has also participated HP Tech Ventures, Sett, ForgePoint Capital International, CDP Venture Capital, Santander Climate VC, Quantonation, Toshiba and Euskadi Risk Capital of Euskadi – Spri Group. Last March the company received An investment of 67 million euros by the Government of Spain. The inference AI by flag. Although the current prominence usually takes it the great technological ones that invest billions of dollars in data centers to train Great language models (LLM)there is more and more focus on the other part: the one we use users when asking things to Chatgpt, for example. It is the so -called AI inference, and the estimate is that in 2025 the value of that industry reaches 106,000 million dollars. In Multivrse Computing they want a good piece of that cake, and to achieve this, its great trick is a unique technology. Compactifai. This is the name of The compression technology of AI models developed by multivance computing. What this allows is to convert very large models – which costs a lot to “execute” – in much smaller and efficient models, which allows them to make them more manageable and save many resources (and time) during inference. How to compress an AI model. Román Orús, scientific director of the company, led A study May 2024 in which they precisely explained the concept of “tensioning networks” of quantum inspiration and that allow compressing these models. Its operation is based on decomposing the matrices of pesos from the neural networks “truncating them” and retaining only the largest and most relevant values. In essence the concept focuses on discarding the less relevant information of the model to be left alone with the most relevant. But that does not make the model less accurate? In fact, but the degree of truncation can be controlled so that there is a good balance and commitment between compression and loss of precision. Even by compressing these models, in Multivars Computing they say The fall of the models It is only 2 to 3%. Same yield in a size 95% lower. To mitigate that precision fall, this system includes a rapid resentment phase called “healing” that can be repeated several times to achieve even closer accuracy to the original version. In the end, they affirm in the company, they can compress up to 95% a model of the performance. It lowers the use of AI. According to Your dataa model as it calls 3.1 405b has an operational cost of about $ 390,000 if we want to run it at home (13 GPUS H100, 9100 W of consumption), but thanks to Compactifai it is possible to reduce that cost to 60,000 dollars (2 GPUS H100, 1,400 W). One more “thin”. The “Slim” models provided by the company – Derivatives of Llama 3.3 70b or Call 4 scout– They are compressed versions that theoretically do not lose precision. They can be executed through the AWS platform or by licenses that also allow us to use it on-premisethat is, in local/own infrastructure. According to their metrics, these models are between 4 and 12 times faster than their non -compressed versions, which translates into an inference cost that is between 50% and 80% lower. Image | Multivize Computing In Xataka | Spain is finally

The US Navy wants to modernize its F/A-18 with sensors that cost 16 million each. They do not resist 40 hours without failing

by

There are more modern, more expensive, newer fighters. He F-35for example, with his futuristic cabin and his advanced stealth. Or the F-22 Raptorless young, but so well known that you barely need a presentation. Even the future F-47. But while that happens, a good part of the United States aerial muscle continues to rest on the shoulders of a veteran: the F/A-18E/F Super Hornet. A plane that began to fly in 1995 and that, with constant updates, it is still the versatile hunting par excellence of the Navy. The challenge is to keep it up. And for that, added added. One of them is the IrST Block IIan infrared search and monitoring system designed by Lockheed Martin. It is integrated into a modified central deposit and Cuesta, According to the latest GAO report16.6 million dollars per unit. Its function is to detect threats from long distances without the need to turn on the radar, which allows the pilot to “see without being seen.” A key tool against poachers, long -range drones and environments with intense electronic warfare. In theory, a tactical jump. In practice, for the moment, a problem While IrST Block II has already been tested in real operations, and its capabilities are well documented, it has a serious obstacle: reliability. According to the same GAO report, the system fails, on average, Every 14 hours of flight. The minimum required by the Navy is 40. That is, the sensor does not endure even half of the time that should suffer a critical failure. And this is making its large -scale deployment difficult. During the operational tests carried out between April and September 2024, the IRST Block II showed unstable behavior. According to a Dot & e reportthe system suffered unexpected flights in full flight, software blockages and hardware failures that, in many cases, required direct assistance of engineers from Lockheed Martin. The marine maintenance crew could not solve them alone. The failures are not limited to software. In 2023, a previous GAO report warned that between 20 % and 30 % of the manufactured components did not comply with the technical specifications. They identified themselves Microelectronics problemsthe cooling system and the general assembly of the Pod. Although some of these deficiencies have been corrected, but many others persist, as we have just seen above. The schedule of the program has been deteriorating year after year. The decision to go to production in full cadence was scheduled for early 2025, but was postponed. And that has consequences. The Irst Block II is not just a punctual improvement: it is an essential piece within the effort to keep the Super Hornet competitive against more modern rivals such as China and Russia. The ironic thing is that while the navy still hopes to trust its star sensor, the American Air Force has already integrated similar systems in its F-15 and F-16. In Western Europe, Eurofight Typhoon also incorporates a similar solution. Apparently, operating from an aircraft carrier implies other conditions, and that is complicating things for the US Navy. United States | DOWRY | Lockheed Martin In Xataka | We prepare to say goodbye to Windows 10, but part of the US Air Control still works with disks and Windows 95

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.