difficult

If the problem is too difficult, they give up immediately

by

Machines do not think, that’s an illusion. We do not say it, a group of Apple researchers who have just published a revealing study entitled precisely (‘The illusion of thinking‘). In it these experts have analyzed the performance of several AI models With the ability to “reason”and their conclusions are striking … and worrying. Puzzles for the “reason”. The normal thing when evaluating the ability of an AI model is to use benchmarks with programming or mathematics tests, for example. Instead, Apple created several Tests based on logical puzzles that were totally new and that therefore could not be part of the training of these models. Claude Thinking, Deepseek-R1 and O3-mini participated in the evaluation. Models that crash. In their tests They checked Like all these reasoning models, they ended up starring Bruces against a wall when they faced complex problems. In those cases, the accuracy of these models fell resorted to 0%. It was also not matched that you granted more resources to these models when trying to solve those problems. If they were of some difficulty, they could not with them. They get tired of thinking. In fact, something curious happened. As the problems became more complicated, these models began to think no more, but less. They used less tokens to solve them and riddled before they could use unlimited resources. Not with help. Apple researchers even tried to give the models An exact algorithm that guided the models to find the solution step by step. And here, another capital surprise: none of the models managed to solve problems despite having those guided solutions. They could not follow instructions consistently. These graphs show the differences between models that do not reason (Deepseek-V3) with those who do (deepseek-r1) in low complexity (yellow), medium (blue) and high (red) problems. There are only advantages for “reasoning” in medium difficulty problems. In the high models they simply collapse. Source: Apple. Three types of problems. In their evaluation they divided the problems to be solved in three classes and verified if the reasoning models really contributed something to the traditional models that do not “reason.” Low complexity problems: reasoning models effectively surpassed those who did not have that reasoning capacity. Of course, they often think too much to solve these simple problems. Average complexity problems: there was still some advantage over conventional models, but not too much. High complexity problems: All models ended up starring these problems. Thinking, nothing. According to these researchers, the reason for this failure when reasoning in complex problems is simple. These models do not “reason” at alland all they do is use advanced patterns recognition techniques to solve problems. That does not work with complex problems, and there the foundations of these models are completely falling apart. Given these problems, if a model is given clear instructions and more resources should improve and be able to try to solve them, but this study demonstrates otherwise. Far from AGI. What these results suggest is that the expectation that these models have generated is undeserved: the current reasoning models simply fail to move from a certain barrier by adding data or computing. Some pointed to how reasoning models could be a possible way Towards the search for the AGIbut the conclusions of this study reveal that in fact we are not closer to achieving models that can be considered general artificial intelligence. They do not find solutions, they memorize and copy them. In fact, the study corroborated something that others defended in the past: These models simply have knowledge, and reproduce the solution they already had memorized when they find corresponding patterns that lead to that solution. Thus, these models could solve the famous problem of the Hanoi towers From many movements because they once know the solution can be applied systematically. However, in other puzzles they failed to the few movements. Stochastic parrots. Many of the critics of the AI ​​always They have defended That the generative models, reason or not, are basically parrots that repeat what has been taught. In the case of AI they detect patterns and are able to find/predict the following word/pixel when generating text or images. The result is usually convincing, but just because they have become extremely good when detecting these patterns and responding properly and coherently. But it is not new knowledge: it is to repeat the queya. They don’t think. Other critical experts of these expectations have been alerting us to alert us for the dangers of anthropomorphism of the IAS. I explained it Subbarao Kambhampti, from the University of Arizona, which, for example, analyzed the “reasoning” process of these models and their “chain of thought”. We use verbs like “think”, when they don’t think. They do not understand what they do, and that contaminates all the assumptions we do about their capacity (or lack of it). Do not trust what the AI ​​tells you. The behavior of these models confirms what is known since Chatgpt appeared on the scene. As convincing that these models may seem – “reason” or not – the reality is that they can make serious mistakes and make mistakes, although others certainly right. In fact there are cases in which these models do surprise by their ability to solve problems: In Scientific American A group of mathematicians were overcome by an AI model that managed to solve some of the most complex mathematical problems that they failed to solve, or that took longer to solve. Image | Puzzle Guy In Xataka | Copilot, Chatgpt and GPT-4 have changed the world of programming forever. This is thought of programmers

Samsung had it difficult with its exynos processors. Google and Xiaomi have only aggravated their situation

by

The future of Samsung processors is an unknown. The company It crosses one of the most difficult stages in its historyunable to follow competitors such as TSMC and SK Hynix. One of the few shelters who stayed beyond exynos in Galaxy were the Google Pixel. Google’s mobiles incoporab tensor chips manufactured by the South Korean giant. But the agreement between both companies comes to an end, after years of criticism for the low performance of the Google Pixel with respect to its direct competence. The turn of Google TSMC, accompanied by giants such as Xiaomi demonstrating what they are capable with the help of TSMC They put Samsung in an even more delicate situation: they need to demonstrate that they are competitive in chips, and need customers who trust them. Goodbye to Samsung. Google, According to Taiwanese sourceshas closed an interannual agreement with TSMC for the manufacture of its tensioner processors. It is a collaboration that, in case Google wants it (they are still those who have the power to design the chip), you can completely change the Google Pixel as we know them today. According to Digitimes, Google executives recently visited the TSMC headquarters in Taiwan to close a cooperation agreement between three and five years. Own design … or mediatek. Although it is an open secret that Google will have processors with TSMC technology, we are not sure if they will design them or if they will leave this work in the hands of a third manufacturer. Recent leaks talk about An alliance between Google and Mediatekone that would make enough sense. MediaTek is known for offering a better quality-price compared to Qualcomm, and solutions such as dimensity 9400 have proven to live up to the best chips on the market. The Xiaomi case. Betting on TSMC is betting on the winning horse, Xiaomi knows it well. The Chinese company has designed one of the best processors in the market, at the height of the most powerful and efficient of the moment: the Apple A18 Pro, Qualcomm Snapdragon 8 Elite and the Mediatek 9400 Dimensity. What Samsung has not achieved in all his career with Exynos has achieved by Xiaomi with the first chip that is taken seriously. The second generation process of TSMC three nanometers is fantastic, and they have not hesitated to get on this ship. The exynos 2500. We have been listening to the Exynos 2500the processor that will allegedly give life to the Galaxy Z Flip7. The problem? The same as always. According to Geekbench data, your scores are far from the current high -end chips. Specifically, yield as a Snapdragon 8 Gen 3. Being a new horn processor, to perform as a platform almost two years ago is not the best news for the company. At the moment, this is filtered information, and it will remain to verify how this processor pays in the final units. The race for the two nanometers. With its main rivals betting on TSMC (including Samsung herself, which currently uses Qualcomm Chips in her best mobiles), and without customers like Google, Samsung needs to rival TSMC and attract new customers to make chips. The exynos division is losing moneyand the manufacture of their processors It has been falling eight percentage points In the last five years. Samsung wants its second production plant in Taylor (USA) Reutersdue to lack of customers who make it viable. This factory will assume the manufacture of semiconductors with two nanometers lithographic nodes, an essential role for Samsung to fight to recover its competitiveness against giants such as TSMC. Image | Xataka In Xataka | Google Pixel 9A, Analysis: The mobile that reminded me why I like so much Android

The European motorcycle has it more difficult than ever

by

KTM has dodged the bullet. Or, rather, it has made them extract before bleak. Of being One of the main manufacturers in Europe went on to face a critical financial situation and insurmountable without external help. Factory closures, stop in the production and commitment of a complete restructuring plan to get out of the hole. KTM has been rescued from bankruptcy, although its future is still in the air. Why is it important. KTM is One of the historical brands in the world of motorcycle. An Austrian with almost a hundred years of history, and a wide presence in the main best -selling segments in the world: trail and naked. It is one of the few companies with a presence in Moto GPand one of the few European survivors with BMW and Triumph. In a nutshell, losing KTM was losing a European giant. Where bankruptcy comes from. The KTM crisis It is mainly due to a simple reason: they sold a lot, but They did not earn enough money. It is a company that did not hesitate to make strong investments to grow in emerging markets, development for new models and participation in competitions such as GP. This ended up translating in motorcycles notably more expensive about its competition. A snowball that resulted in a global reduction in the demand for their motorcycles, overproduction problems (they failed to sell everything they made) and a debt of more than 2,000 million euros. Safe. As they collect in Motorpasionthe Austrian company has confirmed obtaining the necessary financing to carry out its restructuring plan. “Piere Mobility Ag and KTM AG have received financing commitments, subject to the execution of the corresponding agreements, which They will guarantee that the fees of the quotas can be fulfilled on time. “ An oxygen ball in the form of 600 million euros that KTM had to deposit to avoid bankruptcy. It is news that It was ahead in February 2025after This plan is approved and be waiting for the deposit. The person in charge. Although KTM has not directly expressed who is responsible for this economic injection, all eyes are in Bajaj Auto. This Indian giant has 49.9% Piere Bajaj Ag (KTM matrix), and has been injecting millions for more than two years to keep its production line afloat. Bajaj is also interested in keeping KTM alive. The group is in full expansion in Europe with their own bass motorcycles, and rescuing KTM is a good plan to stay in a privilege position within the motorcycle world. What happens now. KTM is obliged to meet a restructuring plan. What we still do not know is how they plan to reverse this situation to be a competitive brand again. Adjusting the production to the sales forecast, reviewing the price strategy and focusing on low -cost models aim to be three of the pillars that, or not, will have to start cementing. Although KTM will continue producing flagship models In Austria, India will be the country of choice for its motorcycles of 125 and 390, at the hands of Bajaj Auto. The European motorcycle, more and more exception. There are currently only three European brands in the best -selling top 10 motorcycles in April 2025. The first is BMW, which occupies the seventh position. In positions ten -nine, Rieju (Spanish) and Piaggio (Italian). At the head, the historical Japanese Honda and Yamaha (have been unquestionable leaders in each category for years) … followed by The Chinese, Zontes and Voge. The paradigm shift is clear: China will inevitably conquer the motorcycle market, where the adjusted price has not yet been lost. India keep pressing. Although the looks are put in China, India has a lot to do in the future of the motorcycle industry in Europe. Bajaj Auto is not only behind KTM, it also produces motorcycles for one of the most prestigious brands in Europe: Triumph. Speed ​​400 and Scrambler 400x are produced in India, in Bajaj plants. Royal Enfield, who despite his name, has been India for years, has tripled sales in 2024with a very aggressive strategy of prices and balanced models in the middle displacement segments. Two names and two examples of how the future of the European motorcycle happens, less and less, through Europe. Image | KTM In Xataka | In his obsession to end the noisy motorcycles, the EU has just knew the coup of grace

China is about to have the ability to make 5 Nm chips, although it faces a difficult solution problem

by

SMIC (Semiconductor manufacturing international corp), the largest Chinese semiconductor manufacturer has been working on the development of Your own 5 nm photolithography. In early February 2024 the newspaper Financial Times He said he had access to two experts in the integrated circuit industry who defended that this company was finalizing the refinement of their semiconductor manufacturing processes in their machines deep ultraviolet lithography (UVP). Its purpose was to have the necessary technology to make 5 Nm chips massively before the end of 2024, although it did not succeed. If its 5 Nm chips had already been successful in this project, the first Huawei devices or any other SMIC client equipped with this type of integrated circuits would have even seen the market. Be that as it seems, now, this technology is ready. The challenge facing SMIC is the performance by wafer According to Dr. Kiman expert in the manufacture of integrated circuits who has worked in Samsung and who currently investigates for TSMC in the US, SMIC is about to start the production of 5 Nm chips. It is perfectly credible because, as we have just seen, we know with certainty that this company has been working on this technology for several years. And, in addition, Dr. Kim is a reliable source. However, this expert has pointed out something crucial that we should not overlook: the performance per wafer that SMIC has currently achieved in its 5 Nm nodes is less than 30%. An incipient integration technology usually moves in the orbit of 50% performance per wafer When semiconductor manufacturers produce a chip wafer, some of those nuclei do not work properly. It is normal. When they launch a new lithographic node, their performance by wafer usually has a margin of broad improvement, but little by little, as engineers refine their integration processes, This parameter improves. A mature lithography can deliver to integrated circuit manufacturers a very high performance, but an incipient technology usually moves in the orbit of 50% performance, so only half of the chips produced work correctly. The problem is that for an integration technology to be profitable from an economic point of view, its performance by wafer has to be At least 70%. And, as we have just seen, Dr. Kim argues that the SMIC 5 NM node is below 30%. It is objectively a very poor performance, but we know what this low figure explains: the technique used by this manufacturer to produce these semiconductors. It is known as Multiple patterningand SMIC has used it for more than a year and a half to make 7 NM chips for Huawei and other customers. This strategy consists in transferring the pattern to the wafer in several passes with the purpose of increasing the resolution of the lithographic process. It works, but is responsible for wafer performance is clearly improvable. SMIC engineers have been forced to resort to Multiple patterning because The US and Netherlands sanctions They prevent Asml from selling their extreme ultraviolet lithography equipment to their Chinese customers, which are the ideal to make chips of 7 nm or less. With the UVP machines that SMIC has, it will be very difficult for wafer performance to be optimal, so in all 5 Nm integrated circuits they will be scarce and expensive. The definitive solution to this problem for SMIC, Huawei and the other Chinese companies that are dedicated to semiconductors inevitably goes through developing their own UVE lithography teams. They are in it. Image | SMIC More information | Dr. Kim In Xataka | The US has declared the total war on Huawei: he does not want him to sell his chips for the most advanced outside of China

how difficult it is to protect copper in a 15,000 km network

by

“You cannot monitor 24 hours 15,000 kilometers of network, but you will have to put more means.” The phrase left her yesterday during An interview In Antena3 the president of Renfe, Álvaro Fernández de Heredia, and is interesting for several reasons. First for what he says. Second, for how he says it. And third (and fundamental), when he says. The rail operator complaint comes after the bird line between Madrid and Sevilla lived chaotic hours on Sunday by the Cable theft of copper in several points of the layout. What happened is serious, but it reveals something more worrying: how difficult it is to shield a network thousands of kilometers. Collapse on the Madrid-Seville line. The bird line between Madrid and Seville (The dean of the Spanish high -speed network) did not go through its best moment on Sunday night and Monday morning. Delays Stakes arrested for hours. Collapsed stations. AND More than 16,000 passengers affected. Although the collapse coincided with An incident starring an Iryo train, both Adif as the Minister of Transportation, Óscar Puentethey did not take long to relate what happened to the theft of cable on the line. “A serious sabotage act”. Sunday night, with the bird line between Madrid and Seville still knocked out, Oscar Puente spoke already of a “serious sabotage act” and pointed specifically to the theft of cable at various points distributed within about 10 kilometers. The thefts were recorded in five different locations distributed among the PK 102+200 and 92+800in the province of Toledo. In total the thieves took 150 m of copper cable. Click on the image to go to Tweet. A booty of 300 euros. The big question that was bouncing on Sunday and has continued to do so yesterday and today is what is the reason behind what bridge closets “sabotage.” Copper It has been revalued coinciding With the tariff war unleashed by Trump, but a priori the metal stolen in the Toledo line is rather scarce. The Government delegation has made accounts and calculates that its value Barely reaches 300 eurosso it has suggested that the real objective was to “block the road.” “That cable, which has very little value, is optimal for primar service to the line,” ditch. Bridge insists on talking about “A coordinated action” perpetrated by someone who “knew what was going” while the issue of the political debate. The PP has even related to the “obvious deterioration” of public services and He has demanded an “audit of the entire network”. Beyond the political or research sand that It has already opened A court of Toledo, what happened in the bird line between Madrid and Seville raises a fundamental question: Is it so easy to steal on the network? It is not the first time that the rail network suffers a robbery or sabotage. In 2022 The Civil Guard stopped to a band that was dedicated to stealing copper in the bird line in Valladolid, Palencia and Burgos. In total it had been made with a boot of 185,000 euros. And years before, In 2015he had already arrested 28 people from a Madrid organization related to the theft of more than 30,000 m of rail cable in several communities. In that case it was estimated that their “blows” had cost about 840,000 euros. The Rodalies case. The above are only two examples that can be easily found in the newspaper library. There are moredistributed by different latitudes of the Spanish geography, and that not only affect the high speed network. Does Just a year Without going any further Catalonia suffered the theft of 40 meters of cable on the Rodalies network at 300 m from the Montcada-Bifurcació station. The incident in turn caused an over -teaching that ended up affecting the service. Yesterday Europa Press It echoed From a balance of the security forces that show that only in 2024 they registered 4,433 copper wiring and conductive materials, 87% more than a five years. In total, 987 people were arrested and investigated, double that in 2019. The balance is general and does not only relate the robberies that affect the rail network, but still gives an approximate idea of ​​how frequent these types of crimes and also how it has been able to affect them the price increase of copper. A great network, a great challenge. The key was given yesterday by Fernández de Heredia during Your interview In Antena3: Spain has a wide (very wide) railway network and that is at the same time a chance and a huge security challenge. The Adif and Adif Av enclos 15,519 kilometers of network, of which 9,984 are electrified and just over 3,700 high -speed connections of different types. And that is just the Railway Network owned by ADIF. To control them in 2021 the organism He tendered a contract of surveillance and security services for three years (from April 2022 to March 2025) that amounted to 210.8 million euros. But still the challenge of monitoring the entire network is considerable. The bridge itself He explained That the 150 meters of stolen cable over the weekend were taken out of difficult access areas, between forest and olive groves. “We will have to put more means”. “You cannot monitor 24 hours 15,000 kilometers of network, but more means will have to be put to avoid it because the disorder caused by these robberies to travelers is very high,” He insisted yesterday The president of Renfe. In the past and before terrorist threats the government came to use the army to monitor the lines, as happened In July 2005after the attacks suffered in London. Images | Nelso Silva (Flickr) and TRANSPORT MINISTRY (X) In Xataka | The US has been dreaming of its first high -speed train decades: the California project is being a real nightmare

bananas prices are in the clouds and the explanation is as simple as it is difficult to solve

by

At the beginning of the year, The Nightmare of the Canary Islands banana lasted 24 months. From January 2023 to October 2024, only in three months of the 22 the banana has had a “remunerative” price. The situation was terrible, almost unsustainable. And then, the penultimate week of February arrived. Since then, continuous increases in the price of the Canarian bananas have created a situation that we have not seen for a long time: up to four euros per kilo in the supermarket and 1.5, in origin. The banana is in the clouds. What is happening here? That is, how is it possible that the situation has changed so much in such a short time? And the answer, although it is the product of two different situations, is surprisingly simple. The first is the week cuttingl. As analysts recognize, on the islands, the weekly bananas cutting has been reduced. At the beginning of March (the latest estimates available), The figures were “Below eight million a week and even seven, which means much less embedded fruit.” That, by pure offer-demand, tends to raise prices. The second is banana scarcity. Because yes, the banana (the “banana dollar”) is the main competitor in the peninsular Spain of the Canarian banana. In fact, As Román Delgado explained in the Canary Islands now“Last year, half of the market share of this fruit in Spain has already been clear.” Well, the shortage of Canarian banana has coincided with banana shortage. The result is that, well, prices have shot. Above all, because the demand has remained. So Sergio Cáceres has recognized itManager and director of Marketing and Communication of the Association of Organizations of bananas producers of the Canary Islands (ASPROCAN), the organization that brings together 100% of the Canarian producers. And it is not uncommon: as we have learned with olive oil, The inelasticity of demand It is the main trick of national producers in times of crisis. So … the problem is already solved? The answer is also simple: no. Of none menra. To start because There are not two good months that can ‘cure’ the wounds caused for 24 bad months. Current prices are a good news, but teaching producers accounts requires some commercial stability. And not four euros per kilo. It doesn’t take so much. Just have remunerative prices for a prudential time. Secondly, for something that is closely related to this: the market is terribly volatile. The commercial chaos of recent weeks makes anyone really know what will happen to product flows worldwide. Who can assure that all bananas will not reach Europe in Europe that cannot be sold in the US (for tariffs)? Carpen Diem. Anyway, it would be fool not to celebrate that the islands have left their peculiar silver nightmare. It only remains to expect prices in the supermarket to begin to normalize and the increases do not erode the demand. In the coming years, we will have to make many decisions around the agricultural sector and the better we get to them, the better. Image | Kamila Maciejewska | Doğan alpaslan demi̇r In Xataka | If the question is what to do with the millions of bananas that Canary Islands throw every year, there are already those who are clear: wine

In 1995, a reading club began reading James Joyce’s most difficult book. 28 years later it is finally finished

by

More than a quarter of a century has taken Gerry Fialka, a Californian experimental filmmaker, in bringing a very ambitious purpose: a reading club of ‘Finnegans Wake‘, James Joyce’s book that is famous not only for his extraordinary literary quality, but for the difficulty involved in his pages. Literary nightmare. ‘Finnegans Wake ‘was published by deliveries from 1924, and was only edited as a book fifteen years later, when its title was also revealed. Since its first edition, the hostility of critics and readers was won by Your difficultywhich sometimes seems to be written in An invented language (in fact, mix words of seventy languages), and with which Joyce seeks to reproduce the way in which memories are ordered and reproducedwith words of multiple meanings and that try to challenge literary conventions at all times. 28 years. From this monumental fuck (‘Finnegans Wake’ is the closest that literature has been to generate a completely new means of expression), Fialka congregated every month in a local library to a group of between ten and thirty people. Your mission: comment in each session two pages of the book. The purpose was so ambitious that they ended up having to reduce it to a single page a month. They began in 1995 and 28 years later, in November 2023, they managed to finish reading full ‘Finnegans Wake’. Why get into this authentic scrub? The Guardian He spoke with Fialka when the reading came to an end, and some of his usual people commented on the appeal they had found in the monumental task. Bruce Woodsis, a 74 -year -old Disney retired animator, says that although “there are 628 pages of things that look like typographic errors,” he has not stopped rereading the novel since his adolescence, and that he finds in it “something of visionary.” Woodsis allowed himself to leave the club for two decades to return to him when he found no other to analyze it so intelligently. At that time, the club had only advanced fifteen chapters. A special club … With such a special purpose and novel, it is clear that we do not talk about a club to use. Fialka himself defines him more as “a Performance Artistic that a reading club “, and also speaks of the club as” a living organism. “The group ended up finding a purpose despite a few initial months of chaos and gallimaties comparable to the sensations that the book itself awakened. The curious thing is that the interpretations of the work themselves are all valid, because Joyce died not long after publishing it: he could not explain it. … for a special book. Sam Slote, One of the greatest experts In Joyce of the world, he affirms that “we must accept that no one will understand it, and that is where the idea of ​​community reading enters.” After all, Joyce himself affirmed that “the demand I make to my reader is that I dedicate all his life to read my works.” Fialka and his people seem to follow their indications, although they are not the only ones: Slote states that there are more than fifty reading groups of ‘Finnegans Wake’ throughout the world. Other clubs. Some of them seem to be trapped in an eternal literary return: the ‘Finnegans Wake’ club of Zurich has read it three times in forty years. One of them lasted eleven. And when they end, they start again, something that the book itself helps: the last sentence is interrupted in the middle and recover on the first page. Of course, Fialka himself, who is already seventy years old, has had no choice but to start again: in November last year they began their second reading of ‘Finnegans Wake’. Header | Unspash In Xataka | I thought I should always read new books, until the rereading showed me what I was losing me

We do not know what the Benchmarks of Ia measure. So we have talked to the Spanish who created one of the most difficult

by

Gemini 2.5 Pro is the best model in history. The smartest. At least, right now. I don’t say it, he says The Chatbot Arena classificationa platform in which they run various tests or benchmarks to try to measure the global capacity of modern AI models. According to these evidence, at this time Gemini 2.5 pro experimental, launched On March 25, it has a score of 1,440 points, well above GPT-4O (1,406), Grok 3 (1,404), GPT-4.5 (1,398) and of course an Depseek R1 that despite its fame is in seventh place with a score of 1,359 points. In current Ranking of Chatbot Arena, it places Gemini Pro 2.5 experimental as the most capable model of AI at the moment. That (probably) does not last long. Google herself presumed the capacity of Gemini 2.5 Pro experimental in the official announcement. As usually happens in these ads, companies show a table in which they compare their performance with that of other comparable models in different tests. In almost all of them Google crushed their rivals in well -known tests in this segment. Is for example the Humanity’s last exam (general knowledge and reasoning), GPQA Diamond (science), Aime 2025 (math), Livecodebench V5 and Swe-Bench Verified (programming) or Mmmu (visual reasoning). All these benchmarks try to measure the ability of these models in more or less specific fields, and all help to demonstrate that models, indeed, are improving. And yet none of them answer the fundamental question: Is the AI so intelligent Like the human being? There is the really complicated, because the definition of intelligence is not entirely clear either. There are different types of intelligence, in fact, and measuring them in humans is not simple or even possible either. And comparing the ability of an AI with the ability of human intelligence is usually not easy. Some experts wonder if IA laboratories will not be cheating with the benchmarks There are in fact who argues that the progress of AI models is misleading. It recently Dean Valentine, from the Startup Zeroopath. He and his team created an AI system that analyzes large code projects in search of security problems. With Claude 3.5 Sonnet They noticed a great leap, but from there the subsequent versions have seemed much less striking. In fact, this expert pointed out that today many of the companies that launch these models focus too much on going well on the photo of the existing and most popular benchmarks and “sound intelligent” in conversations with human beings. Wonders if the laboratories of AIs are cheating and lying: For him the evolution shown by Benchmarks does not correspond to the real benefits when using them. Frontiermath and the challenge of solving problems that (almost) nobody has solved But there are attempts to answer that question. One of them comes from the team that develops THE ARC-AGI 2 PROJECTa set of evidence derived from the Moravec paradox: They are relatively easy for human being, but very difficult for AI models. Jaime Sevilla, CEO of Epoch Ai. These tests measure the ability to generalize and abstract reasoning with visual puzzles, and are undoubtedly an interesting part of that effort to value how far we have arrived at every moment with the AI ​​models. Another of the most striking tests of recent times is Frontiermath. This benchmark created by the company COPHAI It consists of about 300 mathematical problems of different level. They have been designed by a team of more than 60 mathematicians among which Terence Tao, winner of the Fields Medal. Although there are some more affordable problems, 25% of them are qualified as especially complex. In fact, only the best experts could solve them, and It would take even days In doing so. This set of tests is also special for another aspect: these are unpublished problems and therefore have not been part of the training sets of any AI model. To solve them the machines need to be able to show a special “mathematical intelligence.” One that It helps precisely to something increasingly difficult: Assess the evolution of these models. In Xataka we have been able to talk to Jaime Sevilla (@Jsevillamol), which is precisely the CEO of COPHAI and has a very clear and personal vision on how the tests should be to measure the ability of an AI model. To begin with, he points out, “you need to have a way of measuring how the AI ​​is advancing. Interacting with it can give you perspective, but you do not have a rigorous impression of where it will arrive and in what domains it is most expert.” That, he explains, makes it necessary to have standardized test batteries that allow us to form an idea of ​​their skills. For this expert the Benchmark Arc-AGI is more representative of that other vision, making an easy benchmark for humans but difficult for AI. The models are improving in Arc-Agi, but for him that was obvious and that had to happen. With yours the tests are difficult for each other, and that the models advance and are increasingly better when solving these problems is not so obvious. Thus, with FrontierMath they wanted to “try to measure if AI can solve genuinely difficult problems.” Until now the mathematical problems that were subjected to the AI ​​models were relatively easy, so the models “saturated the benchmarks”, that is, they soon managed to overcome all these tests and achieve a 100% score. “It will be a challenge to saturate this benchmark“He stressed. Here I set an example with OPENAI’s O3-mini model, which already solves 10% of FrontierMath. It is not much, but it is brutal, he says, and has already surpassed expert mathematicians like himself. However, he says, “That the AI ​​overcomes certain benchmarks does not mean that it can operate as a human expert. You have to adjust them because they are adjusted to very specific scenarios. We are measuring those limits of that AI, and that will be a continuous process.” For Seville … Read more

The difficult thing has not been to build a yacht of 80 meters and 200 million dollars. It has been to take it to the sea without destroying it

by

Imagine living in a quiet town near Rotterdam, and when you look at your window you see a colossus for you, 14 meters wide and with the height of a three -storey building. It is what has happened (once again) to the inhabitants of the quiet town of Alphan (Netherlands). According to published the local media AD Those who have approached one of the channels that are going through the population have seen how A 200 million superyate of dollars Lawrence Strollowner of the Aston Martin Formula 1 team 1 and head of Fernando Alonso, juggled to reach the sea sailing through narrow channels and raffling bridges and all kinds of obstacles in his odyssey. The megayate odyssey Feadship is one of the world’s main manufacturers in the world. From their shipyards in Aalsmeer they have left colossi like the Launchpad by Mark Zuckerbergwith 118 meters of length. All of them have had to go through that intricate journey of narrow channels, curves that test the expertise of the engineers involved in the transfer and several traffic cuts in the populations through which they pass. This type of operations are not simple. They require millimeter planning and the perfect execution of each step. Any calculation or maneuver error could have ended up damaging the helmet of a vessel valued at more than 200 million dollars or, worse, putting people’s safety at risk. The epic odyssey of PROJECT 714production name that the Stroll Yate has received, was recorded on video for the Dutch Yachting Channel. Some sections of the transfer were especially tense, such as the pass There were hardly a few centimeters of margin so that the superyate helmet scratches its pillars. The tight turns in the channels, as a chicane in an F1 circuit as in which their owner competes, also contributed their touch of tension during the hypnotic transfer to high seas. Project 714 Leaving the shipyards When even money can pave the way PROJECT 714 is neither the first nor the last supereyate that makes this journey, but its complexity reaches the extreme when it comes to large vessels such as the commission of the owner of the Aston Martin Formula 1 team 1, with an estimated heritage in 3.8 billion dollars, according to Forbes. The journey of Koru de Jeff Bezos From the Oceanco shipyards in Alblasserdam, it was a challenge the exit to the high seas through some channels similar to those that has had to travel the Spery of Stroll. On that occasion, the Dutch builder had to face A serious problem: A historical drawbridge built in 1927 was not high enough for the 70 -meter masts of the Koru to pass under its structure, so the construction company proposed to temporarily dismantle the bridge so that the sailboat could cross it. According to The published by the local medium Truuwthe refusal of the neighbors forced the refusal of the City Council, so the builder had no choice but make the journey without masts to complete the construction of the Koru in the shipyards that the company has in Greenport. A floating mansion Ad picked up some of the reactions of Alphan’s neighbors who came to see how Project 714 slid over the waters of their narrow channels. “This is quite impressive,” “is not normal”, “the closer, more impressive” or “we are in first class” are some of the comments that raised in its path the impressive mole of steel and aluminum. The luxurious yacht is designed to be A floating mansion destined for leisure and It has five covers, A beach club with pool in the stern, an sharp bow cover that could well host a helipad and wooden floors in all covers. When being in manufacturing phase, Feadship does not give many details about his Interior equipment and finishessomething that will be addressed after overcoming the navigation tests that it now faces. Which He has shared It is that the yacht has a more efficient diesel-electric hybrid propulsion system that reduces vibration and improve comfort on board. In Xataka | Ultrararicos change the ground for a superyte during the summer: so are some of these floating mansions In Xataka | Keeping a mooring supereate comes out very expensive. Some researchers have a proposal for the rich: donate them to science Image | Feadship, Aston Martin F1

France puts Apple a fine for making advertisers difficult. The problem of the fine is that it is symbolic

by

Apple It has been fined with 150 million euros By the Fancesa regulatory authority, I authorized her in the concurrence. The sanction is due to its dominant position between 2021 and 2023 in the advertising segment in mobile applications. It is the first fine that an antimonopoly regulator issues Apple by the call Tracking transparency app (ATT). This technology is supposed to prevent apps from tracking us more than the account. In iPhone and iPad ATT it allows users to decide which apps can monitor their activity. However the system has been criticized by advertisers And for Apple’s rivals –With Facebook as a great example-, to which it harms by depending on that online advertising. In fact, the investigation that has ended up causing this fine comes from the complaints of Several online advertisers associations and also of Internet suppliers who accused Apple of abusing their privileged position. The French regulatory entity indicated in a statement that “although the objective pursued by ATT is not critical in itself, the way in which it applies It is not necessary or provided to the declared objective of Apple to protect personal data. “ In fact, the statement also stood out as ATT “particularly penalizes small advertisers”, who depend largely on third -party data for their business. The fine, of course, is much smaller than the European Union imposed on Apple last year for Spotify demand for “limiting options and drowning innovation.” Then The fine was 1.8 billion euros. Apple has indicated that it was disappointed with the fine, and that the French regulator has not specified what changes should make for its privacy control tool. The ATT system is also being investigated by the regulatory entities of Germany. The fine, as we say, is almost symbolic, especially if we compare it with what the EU imposed last year. Even so, this could return to Increase existing tensions with Donald Trump’s governmentwhich in recent weeks has begun to launch tariffs that raise a global commercial war and that of course They significantly affect Europe. Image | Anthony Choren | La Moncloa In Xataka | The Spanish car will be unscathed from US tariffs for a very simple reason: we manufacture cheap models

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.