race

We are discovering how the brain “hacks” us to make us hungry. And it is a key step in the race towards losing weight.

by

Right now, treatments to lose weight are the order of the day, with a clear protagonist like Ozempic. The problem is that beyond the aesthetic effects that are achieved, there are many doubts about both the side effects as well as all the effects it has on the body. But little by little science you understand much better how they achieve their effectwhich seems like a real miracle for many. What we knew. In general, these treatments They are ‘copies’ of GLP-1 which is a hormone that we produce normally in our body and makes us have the feeling of satiety. The moment we increase it exogenously we have a greater feeling of satiety that allows patients to lose weight (although with a risk of bouncing when treatment is stopped). But beyond this effect, the action it could have directly on the brain was something that had only been explored in animals. Now, a new study published in Nature has crossed this frontier thanks to Casey Halpern’s team, which has taken advantage of a “unique opportunity” to observe, for the first time in humans, the impact of Mounjaro (tirzepatide) directly into the reward center of the brain. Why it is important. The discovery of how the brain can ‘hack’ our body to eat much less opens many doors for us in the field of pharmacology to be able to continue working on definitive treatment. against obesitybecause we are seeing that it is something in high demand by many people who find it necessary to have this help (although it is not a miracle) to be able to reduce their weight. And we even see how in the United States purchasing is becoming more and more accessible. And we say that it is a miracle, because Ozempic or Mounjaro does part of the work, but we must not leave aside the change in eating habits to adjust the diet and be able to maintain it after stopping the treatment. The problem is that there are people who after stopping the treatment continue eating normally, and logically they see that there was no miracle involved. How it was done. The study focused on a 60-year-old woman with treatment-resistant obesity and type 2 diabetes. This patient was already taking Mounjaro for diabetes, and coincidentally, she was participating in another trial to treat dysregulated eating. This coincidence allowed the researchers to do something unprecedented: use the electrodes, already implanted in its nucleus accumbens (NAc)for hear brain activity while the drug took effect. And this brain nucleus is really important as it is the center of pleasure in humans and reward, that is, it is the point that can be modulated to restrict food consumption. The sign of craving. Those cravings we have for eating a little chocolate, a greasy pizza or a hamburger are something we all have because it is what gives us pleasure. In this case it was seen that the signal changed over the months, specifically the delta-theta frequency band. In the first months of treatments with Mounjaro, the patient had no desire for food in that sense of craving. Something that corresponded to a null signal in this nucleus, so it could be said that the medication was silencing this ‘noise’ that is generated in the pleasure center. The problem is that in the fifth to seventh months, despite being on the maximum dose of medication, the patient again had severe concern about food. And here again the signal in the nucleus had spiked to match that of those people who had no treatment. An advantage for the future. The most important finding here is that the change in the brain preceded the behavior. That is, before having a relapse this signal was increasing as if it were a warning signal. That is, a future where a sensor can detect this brain signature and alert the patient or doctor that the effectiveness of the drug is decreasing, before that the person will feel the cravings again in an uncontrolled way. Much ahead. This is a study with a single person, and it has many limitations and its conclusions logically cannot condition the clinical activity of the use of these medications. What it is useful for (and a lot) is to understand that the brain has a lot to do with this weight loss as if it were a real button to control eating habits. Perhaps silencing this brain nucleus in a very specific and sustained way may be the ‘holy grail’ that weight loss science seeks to control these cravings that can ruin a diet imposed by specialists. Although there is still a lot to investigate and it is only a first door for other medications that can complement Ozempic or Mounjaro, which has given great results. Images | Shawn Day Victoria Shes In Xataka | This is the great hope of the competition to replace Ozempic. Your weapon: banish needles with a pill

China is quietly winning the AI ​​race thanks to something very simple: cheap energy

by

“China is going to win the artificial intelligence race,” warned Jensen Huang, CEO of Nvidia. Many thought he was exaggerating, interested in fueling demand for his chips. But, as analyst June Yoon explained in her column for the Financial TimesHuang’s argument contains an uncomfortable truth: the availability of electricity—not chips—is becoming the critical factor for the development of AI. A model like GPT-4 can consume more than 460,000 megawatt-hours per year, the equivalent of the energy consumption of 35,000 American homes, according to a study. The world’s data centers—already colossal—could double their electricity consumption before 2030. And that changes the rules of the game. When there are plenty of chips, but there are no plugs. The race for AI It started with a GPU fever. Big tech companies rushed to buy every Nvidia chip available, but they soon discovered something more worrying: there weren’t enough sockets to connect them. Satya Nadella himself, CEO of Microsoft, he said it bluntly: “The biggest problem we have now is not excess chips, but energy.” Electricity demand has skyrocketed so much that Google, Microsoft and Amazon are already contemplating build nuclear reactors to keep your servers on. The paradox sums up the moment well: the digital leadership of the West encounters a physical limit, that of cheap energy. Energy as a new geopolitics. Analyst June Yoon throw a question that reorders the technological map: what if the AI ​​race had nothing to do with chips, but with electricity? If the last century was defined by oil, this one will be defined by the current China no longer lives off oil: generates it. It has gone from being a petrostate dependent on crude oil to becoming the first electrostate on the planet. More than one quarter of your electricity It comes from renewable sources and its network is growing at a speed that no other country can match. Now that energy sovereignty fuels a new front: artificial intelligence. How did you find the formula? Since September, the Chinese Government Subsidizes up to 50% of energy costs of data centers that use national chips. The inland provinces—Guizhou, Gansu, Inner Mongolia—have become “electric hearts” of Chinese AI: there energy is abundant and cheapand local governments offer historically low rates of just 0.4 yuan per kilowatt-hour. The measure has a dual purpose: Compensate for the lower efficiency of domestic chips compared to Nvidia’s. Promote technological independence in the midst of a trade war. As Bloomberg has detailedthese regions are connected by ultra-high voltage (UHV) lines that transport renewable energy from the interior to the coastal areas where big technology companies, such as Alibaba, Tencent and ByteDance, are concentrated. The goal is clear: ensure abundant, low-cost energy for AI training clusters. According to Rystad Energythe electricity consumption of data centers could more than double before 2030, reaching 1,800 terawatt-hours in 2040. Beijing is preparing to absorb it. The result is a planned, centralized energy ecosystem designed to scale AI. An example is the Talatan Solar Parkwhich extends like a sea of ​​metal mirrors: more than 600 square kilometers of panels that are combined with wind and hydroelectric parks. From there, the power travels along high-voltage lines to data centers on the coast. It is a postcard of the new Chinese power: sun, wind and silicon. China’s electrical advantage. The strategy is also working in the markets. According to Bloombergshares of Chinese power companies have risen up to 40% in a week, driven by demand for AI data centers. UBS forecasts that electricity demand in China will grow 8% annually until 2028. Meanwhile, in Washington, the Trump administration has launched an AI Action Plan to accelerate the construction of data centers and remove obstacles to energy projects. But, as FT analysts point outchip improvements are stuck in single digits, while Chinese renewable energy grows by double digits every year. The power is in the socket. In the race for artificial intelligence, chips are the brain. But the heart beats with electricity. The United States retains leadership and has the best semiconductors (for now); China, the network that keeps them on. As June Yoon wroteall the technological superpowers in history—from coal England to oil America—were built on a source of cheap energy. Today, artificial intelligence needs electricity as it once needed steam. And on that new board, China seems to have found the key: plug in the future before anyone else. Image | Pixabay and Hanwha Xataka | SoftBank abandons the king of chips in its prime. And he bets everything on OpenAI

Sateliot is the great Spanish hope to have its own voice in the new satellite space race

by

There is a new space race and no one wants to miss it. Rivaling with Starlink seems like a utopia, but a Spanish company has managed to get ahead to the American giant on a specific point: 5G. While Elon Musk’s satellite company remains anchored in 4G, Sateliot boasts of being a pioneer in offering 5G connectivity from space, not only to IoT devices, but also to conventional mobile phones. This milestone has not gone unnoticed by the governments of Spain and Europe. Sateliot brings together all the ingredients to become an option for technological sovereignty in the satellite race. A race where Starlink dominates with more than 90% of global launches, but where any advance of its own is seen as a great victory. Now Sateliot inaugurates the Europe’s first 5G satellite development center. A pioneering center located in Barcelona that has more than 100 employees, two laboratories, a control room and a clean room of more than 100 square meters. From Xataka we have visited the center of the Catalan satellite company and learned about its ambitious plans. Triton, the new generation of satellites moves to full 5G Since 2018, Sateliot has launched six satellites, the last four in orbit since August 2024. They plan to launch five more next year. However, beyond getting ahead with 5Git will be with their second generation of satellites when they will begin to have a more competitive service. Triton, in homage to the Montseny amphibian, is the name chosen for its new satellites, about four meters long and 150 kilograms in weight. These new satellites represent a radical advance compared to those already sent by Sateliot, because in addition to having a capacity up to 16 times greater, they also change their concept. Tritón not only offers connectivity to IoT devices, but will offer 5G connectivity for data, voice and video to conventional 5G mobiles. Without the need to add any antenna or modifications to these phones and compatible with all operators (3GPP). The satellite, with a cost 10 times higher than the first generation, will allow Sateliot to offer a service that will range from critical security applications to civil protection and defense. The company explains that its satellite connection service will not focus on providing specific coverage to specific consumersbut serve for industrial, maritime, energy or location applications. Jaume Sanpera, CEO of Sateliot, together with the monitoring of its four satellites in orbit The first Triton satellite is scheduled to launch during the first quarter of 2027from Vandenberg (California), one of SpaceX’s two launch bases. The future goal is to be able to use European launchers, such as the Vega and Ariane of the European Space Agency. In this space race, the dates given are no coincidence. 2027 is the date on which it is also planned that Starlink begins upgrading its satellites to 5G. Barcelona bets on aerospace technology Jaume SanperaCEO of Sateliot, is proud that his satellites are “100% manufactured in Barcelona.” Now they have inaugurated the development center, but in the future they plan for the industrial phase to also have a factory in Barcelona. A phase that is still far away. “Next year we will exceed 200 employees. Being more than 80% engineers and having doubled the staff in the last year,” Sanpera explains to Xataka. “We have agreed to expand to the ground floor,” he points out in reference to the recently inaugurated offices. An inauguration that was also attended by multiple public authorities, including the president of the Generalitat of Catalonia, Salvador Illa. “You have to lose your shyness. Everything outside is better and seems to come from the US or China. Well no: Here we also do very powerful things that no one else has“Illa defended. Salvador Illa, president of the Generalitat of Catalonia, visits the clean room of the new 5G satellite development center | Satellite Sateliot is a startup that currently brings together much of what Europe is looking for: cutting-edge technology companies and local development. The new development center wants to become the base of a cluster of aerospace companies in Barcelona. And investors are taking note. Sanpera assures that at this time Sateliot is not looking for a new round, although defines it as a company “that requires a lot of capital”. Last March, the The Spanish government announced an investment of around 14 million euros in Sateliotfor a total of a round of about 70 million euros. In addition to the Spanish Society for Technological Transformation (SETT), Global Portfolio Investments, Indra, Cellnex and SEPIDES have also invested and 30 million euros have been loaned from the European Investment Bank (EIB). For the moment, since his birth They have invested about 50 million euros in R&D. According to Sateliot, they already have signed contracts worth 285 million euros annually and offer coverage in 58 different countries. In total 734 different contracts to connect a total of 10 million devices that cannot have good coverage and where the satellite service opens a whole field of possibilities. The new development center in Barcelona employs 110 employees (80% engineers), with plans to exceed 200 in 2026. “We have 30 different patent applications“, they explain to us. During the explanation of how satellite monitoring works, the CEO of Sateliot hints that not all of its advances have been patented, in order to “not give clues to the competition”, pointing out that there is a high level of industrial espionage in the sector. “The difficulty is in the radio, in the antenna,” says Sanpera. Sateliot cannot compete against Starlink in quantity, but unlike the American company, they are betting on satellites whose connectivity is more modern and, above all, widely compatible. The Triton satellites have a 7 year shelf lifecompared to four or five years for the first generation. The main limiting factor is the radio and software. The company points out that this information is important, because “space debris is a problem for everyone and can prevent us from launching more … Read more

“The real winners of the AI ​​race will be the electricians or plumbers”

by

The conversation about the future of work is taking an unexpected turn conditioned by the progressive incursion of AI into administrative positions, especially among the youngest who see the old promise of a prosperous job future after college. Influential voices in the development of AI, such as Mark Zuckerberg or Mark Chenthey are showing the value devaluation of university degrees. Jensen Huang has gone a step further by pointing out that most of the future economic opportunities could arise from traditional professions more than pure university careers. “The millionaires of the future will be electricians or plumbers,” said the millionaire in an interview for the british Channel 4 News. The winners of the race for AI. Jensen Huang, CEO of the main producer of AI chips, pointed out the significant imbalance in the current labor market, in which there is enormous offer of university graduatesbut a great shortage of qualified labor in traditional trades such as carpenters, electricians, plumbers or bricklayers. According to this shortage, it has made their jobs much more difficult. best rated and salaries could double in a short time. “The big winners in the AI ​​race will be electricians and plumbers,” said the NVIDIA CEO. Huang highlighted that, although these professions were not directly related to the development of AI models, they were related to the construction and maintenance of the data centers that support these technologies. “Whether you are an electrician, plumber or carpenter, we will need hundreds of thousands of them to build all these factories and the specialized crafts sector of all economies is going to experience a boom,” the manager stressed. Traditional jobs versus AI. One of the great advantages of traditional trades over many so-called “white collar” jobs is their limited exposure to AI automation. Furthermore, in other interventions, Huang has pointed out: “As we speak, AI has no chance of doing what we do,” stated the CEO of NVIDIA during a conference in Mumbai. To a greater or lesser degree, all positions with administrative burden or repetitive routines are susceptible to automation of all or part of their work. It is something that is already beginning to be observed with a reduction of junior positions in the main consulting firms, where AI is replacing recent graduates in basic tasks such as reporting or programming assistance. On the other hand, jobs in the manufacturing industry or traditional trades are not feeling the same pressure because, at the moment, an AI cannot fix faults in the electrical installation of a house. AI can’t build your house. Even though you may currently have a waiting list of more than six months To request a quote for a bathroom or kitchen renovation, NVIDIA’s concern about the lack of qualified labor is more related to the deadlines in the construction of its data centers. According to a report McKinseythe industry is projected that global capital spending on data centers will reach $7 trillion by 2030. That’s a lot of processors waiting to be sold. The problem is that, just like renovating your bathroom, without enough manpower Construction deadlines are longer than expected. As and how I collected FortuneLarry Flink, CEO of BlackRock, was also concerned in that regard, noting that the workforce to build the data centers they needed had been dangerously reduced following Trump’s anti-immigration policies. “I’ve even told members of Trump’s team that we’re going to run out of electricians, which we need to build AI data centers. We just don’t have enough,” declared the manager at a recent conference. In Spain they also bet on traditional professions. In Spain, the media millionaire José Elías, CEO of the La Sirena supermarket chain and president of Audax Renovables, has been giving a clear and forceful message: “Artificial intelligence is going to take over 80% of office work,” and that is going to change everything. But not for manual trades. On the contrary, those practical and physical jobs, which many today despise, will be the ones with the most opportunities and the best salary.” “They will end up charging 50, 60 or 200 euros per hour”highlighted the Catalan millionaire in an interview. In addition, he points out that these jobs offer greater growth prospects and professional stability. “A lawyer from a large firm who works long hours does not earn more than an electrician. Or someone from a Big Four audit who works 17 hours a day does not live better than an electrician either,” he explains. “An electrician works nine hours and one is having breakfast. I don’t know any electrician who doesn’t have an apartment and if you’re good and don’t get paid well, raise your hand and tell the boss ‘Hey, I’m thinking about starting to send resumes,’” explains José Elías in one of your videos. In Xataka | Overqualification in Spain becomes chronic: 34% of workers perform tasks below their educational level Image | Jose Elias, NVIDIA

The new arms race is being fought at more than 6,000 km/h. And America is late

by

At more than 6,000 km/h there is no room to think twice. The new generation of hypersonic missiles operates in that speed range, a terrain in which the global military balance begins to shift. Russia and China they have already shown systems capable of flying above Mach 5. The United States, accustomed to setting the technological pace, moves forward with more doubts than it would like. The term “hypersonic” is not military marketing, but a clear category: devices that travel faster than five times the speed of sound. The real complexity comes with the trajectory. Unlike ballistic missiles, which ascend and descend in an arc, these systems can stay relatively low and change course in flight. This ability to maneuver, added to the thermal loads and ionization they suffer when passing through the atmosphere at such speed, explains why their development is so challenging. Hypersonic weapons enter the scene Russia was the first to proclaim operational capabilities. Its Avangard system, an intercontinental missile-launched glider vehicle, was announced for service in 2019 and Moscow claims it can carry a nuclear warhead. Experts in kyiv also claim that Russia used the zircon against the ukrainian capital in February 2024. China, for its part, demonstrated the DF-17 and tested the DF-27, which according to reports from 2023 flew about 2,100 kilometers in 12 minutes. In addition, it has shown the YJ-21, integrated into destroyers and bombers, consolidating a more visible deployment. The United States has focused on the Long-Range Hypersonic Weapon. Dark Eagle has a range greater than about 1,725 ​​miles, that is, about 2,780 kilometers, and a first system valued at about 2.7 billion dollars, according to the Government Accountability Office (GAO). The official plan aims to deploy it at the end of 2025, after a sequence of tests with failures in 2023 and 2024 that the GAO collected in June 2025. In August 2024, the CRS reported of the first satisfactory end-to-end flight. In parallel, the Navy is leading a common glider vehicle and the Air Force is working on an air-launched glider and a cruise ship with DARPA. The hypersonic threat tests the most fragile link in modern defense: time. The radar has less useful horizon at low altitude and Trajectory changes break prediction patterns. Furthermore, the dynamics of flight itself generate phenomena that can complicate detection. The forces trying to stop these systems are working on layers of sensors, more advanced tracking algorithms and more agile data links, but it is a challenge that is not yet solved. What sets hypersonic weapons apart is not just their performance, but the effect they have on the logic of deterrence. The impossibility of knowing what type of cargo they are carrying until impact creates fertile ground for misunderstandings. The United States assures that its development focuses on conventional ammunition, but rivals such as Russia and China have shown systems directly linked to their nuclear arsenal, which fuels distrust. Faced with this scenario, the allies are rearming their surveillance and defense architecture. In 2022, the United States, the United Kingdom and Australia expanded their cooperation within the framework of AUKUS to include “hypersonics and counter-hypersonics“, with emphasis on distributed sensors, shared intelligence and new interceptors. The objective is not only to have equivalent missiles, but to build a system capable of detecting threats in early phases and coordinating the response between different military nodes. The focus is on the next deployment milestones and on validating that this cooperation translates into real capabilities. Today, the initial advantage is not on the American side, and that realization has already had an effect on its military planning. Russia and China have moved first and have forced Washington to accelerate decisions and prioritize resources in the middle of a year of technological validation. It remains to be seen whether the deployment planned for this year consolidates a balance or confirms the gap. Images | People’s Liberation Army | Russian Aerospace Forces In Xataka | China promised them very happy with the catapult system of its new aircraft carrier. Until the US took a look

Real Betis Balompié has joined the space race to solve a pressing problem: collisions between satellites

by

It sounds unlikely, but it is a fact. Real Betis Balompié has entered the space sector. And without leaving Seville. GMV’s new partner. The historic football club and the aerospace company GMV have installed in the Rafael Gordillo sports city a satellite surveillance and tracking antenna. The agreement makes Betis the first football club in the world to host a facility dedicated to the sustainability of the space. More specifically, at pressing space debris challenge and the increasing risk of collisions in orbit. Betis 1 – Space trash 130 million. Earth orbit congestion may not be the main concern of green and white fans, but it is a danger for the satellites we use every daywhether with the car navigator, to see the weather forecast or when we turn on the broadcast of a football match. Thousands of operational satellites coexist with up to 130 million fragments of space debris: pieces of dead satellites and rocket remains that travel at hypersonic speeds and have triggered the evasion maneuvers of the active satellites. It is “one of the great challenges that humanity faces in the orbital environment,” says Miguel Ángel Molina, of GMV. Monitor and prevent. This is where the new 2.7 meter satellite dish installed at the Betis training center in Seville comes into play. Its mission is to track space debris and predict collisions in order to avoid them. To this end, GMV internally developed a system called Focusear. It works by “listening” to the signals that the satellites themselves emit in the Ku band (the same one used by satellite television) from the geostationary orbit, about 36,000 km high. Nanosecond precision. Upon receiving these signals, the system uses radio frequency triangulation techniques (TDoA and FDoA) to determine the position and orbit of the satellites with a margin of error of about three meters, equivalent to 10 nanoseconds. These data are vital to inform satellite operators, who are in charge of managing the evasion maneuvers of their fleets. But also to expand the European Space Surveillance System (EUSST), a catalog of objects that helps prevent large-scale collisions. Why Betis. The Sevillian club had created the Forever Green foundation, whose name has a double meaning. In addition to being green for its kit, Betis has become the most sustainable club in LaLiga (and the second in Europe) in terms of energy efficiency, recycling and water reuse. Expanding this vision of sustainability to space is literally taking its environmental commitment “beyond the Earth,” says Rafa Muela, manager of the foundation. But there is something else. Seville is the headquarters of the Spanish Space Agencyso the choice is not accidental. Somehow the Andalusian capital must be placed on the map of national spatial development. Image | GMV, Real Betis Balompié In Xataka | Three large pieces of space debris reenter every day: “one day our luck will run out and they will fall on someone”

Spain steps on the accelerator in its particular chip race. And it does so with a total commitment to integrated photonics

by

The Council of Ministers has approved the award of 4.4 million euros to the IMDEA Networks institute within the european program of Integrated Photonics. It may not seem like a lot of money compared to the fortunes invested by the technology giants, but be careful: it is the last element of an eye-catching strategy. Fifth successful bidder. The IMDEA Networks institute thus joins four other entities that were awarded last July in the same call. The aid granted by the Government is then matched by the European Union, which causes the budget to double in all cases. Thus, we have: Institute of Photonic Sciences (ICFO): 23.1 million euros were awarded, it will receive 46.2 million in total Polytechnic University of Valencia: 16.5 million awarded, will receive 33 million in total National Microelectronics Center (CNM): 15 million awarded, total investment of 30 million University of Vigo: 7.5 million euros awarded, 15 million total investment IMDEA Networks Institute: 4.4 million euros awarded, 8.8 million in total 133 million for integrated photonics. With this new award, the Government and the EU will invest a total of 133 million euros to “promote research and development of faster chips with lower energy consumption, thanks to the use of light (photons) instead of electrons.” Integrated photonics? This technology focuses on using photons (light) instead of electrons to transmit and process information within chips. With this it is possible to obtain higher data transmission speeds and lower consumption and heat dissipation. What integrated photonics seeks is to take advantage of optical components (such as lasers, modulators and detectors) with traditional electronic circuits to combine the advantages of both components. Technological sovereignty. Although the figure may seem modest in the context of global mega-investments, it is part of an ambitious strategy focused on the research and development of disruptive technologies. The ultimate objective is to promote a key sector for Spanish economic and digital sovereignty, and here the commitment is total to integrated photonics, which is seen as the future of data processing. The PERTE is still there. The importance of this investment goes beyond research. It is a fundamental pillar for the PERTE of Microelectronics and Semiconductors (PERTE Chip), the strategic plan endowed with more than 12,000 million euros to try to position Spain as a relevant actor in this value chain. This investment is framed not in chip manufacturing, but in scientific capacity and design strategy. The idea is to ensure that Spain has its own talent and technology to develop new generations of components. Competence centers. To those 4.4 million awarded to IMDEA Networks another 3.9 million euros are added to create two competition centers co-financed by Europe through the JU Chips (Joint Undertaking). The ‘PIXSpain Competence Centre’ will receive one million euros and the MicroNanoSpain Competence Center will receive three million. Both will provide Spanish companies in the sector – especially SMEs – with access to technical knowledge and experimentation spaces. To compete with TSMC or NVIDIA, nothing. This is not about Spain going to start creating chip factories that can compete with TSMC, far from it. The idea is not to try to create a Spanish-style NVIDIA either. In both cases the resources needed would be astronomical. What is sought is a leadership position in a niche with high added value, which is photonic interconnection technologies. Goodbye to copper cable. By focusing on integrated photonics, Spain aligns with the work of giants like Intel, TSMC or Ciscowhich have been investing heavily in this technology for some time to solve the challenge of interconnections in data centers. Everything indicates that integrated photonics could end up replacing copper cables in high-speed communications in the next decade. In Xataka | “They lead and AI follows”: seven Spanish universities tell us how they are implementing AI in class

Everyone is developing chips that compete with NVIDIA’s. They are in the wrong race

by

Qualcomm advertisement on Monday that it is working on AI accelerator chips, which means there will be new competition for NVIDIA. The company that dominates the AI ​​hardware landscape is seeing a large group of competitors try to erode that position, but the problem for all of these companies is not the chips, but something else. A CUDA call. what has happened. Qualcomm has announced the AI200 chip, which will begin selling in 2026, and the AI250, which will do so in 2027. Both will be able to work in rack-type systems that have liquid cooling. Qualcomm servers may have up to 72 chips based on the Hexagon NPUs of the company’s Snapdragon SoCs. Inference yes, training no. The company has revealed that its chips focus on inference (the execution of AI models) and not training. Their rack-based systems will have lower operating costs than cloud system providers, Qualcomm says. Each rack consumes 160 kW, a figure comparable to the consumption of some racks based on NVIDIA GPUs. There are no details about the price of these chips, the cards or the racks that will integrate them, nor about how many NPUs can be offered in each rack. What we do know is that Qualcomm’s accelerator cards will support up to 768 GB of memory, more than what NVIDIA or AMD offer in their current models. according to CNBC. Chips for third parties. The other important point is that Qualcomm will sell its AI chips and other components separately, allowing large AI companies to “customize” their own racks based on Qualcomm chips. It is an identical philosophy to the one they have adopted in the world of their mobile SoCs. Investors viewed the news with exceptional optimism, and Qualcomm shares rose 11% in Monday’s session. NVIDIA dominates with an iron fist. In the AI ​​chip segment, the king is NVIDIA. The company is the absolute protagonist of this market and according to CNBC it maintains a 90% market share, which has allowed it to skyrocket its valuation to 4.5 trillion dollars. That dominance could now be threatened by the avalanche of chips that are arriving from various manufacturers. All against NVIDIA. AMD has its excellent Instinct, Google has your TPUsAmazon their TrainiumMicrosoft their Maia and Huawei has your Ascend. All of them make really striking proposals for NVIDIA chips, and little by little these solutions are being integrated into more and more data centers. But the real problem is not in the hardware, but in the software. The great challenge is to defeat CUDA. The de facto standard in the AI ​​industry that developers use It’s CUDAa platform that allows you to take full advantage of the capabilities of NVIDIA chips in the field of artificial intelligence. This hardware+software combination is much more mature than that of its competitors, who have the hardware part resolved (or are on the right track) but do not have a platform comparable to CUDA. AMD has ROCmwhich is especially interesting because it is Open Source, but at the moment its features still do not reach those of CUDA. Reinvent the wheel? CUDA has been on the market for almost two decades, which means that the majority of academic research and pioneering models—such as ImageNet—were written for CUDA. It is not a language, it is a vast collection of libraries, optimized frameworks (like cuDNN), debugging tools and a huge community. Developing a competitor is basically like reinventing the wheel, and migrations are expensive and companies and startups will not have an easy time assuming it. China is also in the fight. And of course, if there is another great protagonist in this race, it is China. The Asian giant, previously dependent on NVIDIA, is seeking to get rid of this manufacturer, and along with the development of advanced AI chips they are also trying to have its own AI software to surpass CUDA. In Xataka | AI is the best thing happening to nuclear fusion. The construction of ITER is already accelerating

Elon Musk is trying to win the AI ​​race by creating the Wikipedia of AI. We have many questions

by

Grokipediathe new online encyclopedia created by xAI, is now available. The project that Elon Musk has been talking about for some time is just what we expected: a version of Wikipedia in which the content has been generated by Grok, the AI ​​model developed by Musk’s company. And that is precisely the problem. What is Grokipedia. Basically, a copy of Wikipedia in which, as we say, the writing of the texts is done by Grok. The design is simple, with a home page that is a search engine. The articles follow the design of Wikipedia and its structure of different headings and photos. At the moment there do not seem to be any photos in those articles, and Grokipedia does not currently allow users to edit those pages either. If AI makes mistakes, how can we trust AI? The essential question that determines the validity of the idea of ​​Grokipedia is precisely that. Considering that AI makes things up and makes mistakes, what can you expect from an online encyclopedia created by an AI model? Grokipedia on the left, Wikipedia on the right. The PS5 article is an absolute copy of the Wikipedia original. Content “adapted” or directly copied from Wikipedia. Some Grokipedia pages display the message that the content has been adapted from Wikipedia taking advantage of the Creative Commons Attribution-ShareAlike 4.0 license. This happens, for example, with the article dedicated to MacBook Air. In other articles such as that of the PlayStation 5 That message falls short because the article is basically the same as Wikipedia’s. An encyclopedia with biases. In Grokipedia there are signs that the theoretical neutrality and objectivity that should be fundamental pillars of such a project are faltering. As indicated in Wiredthere are worrying examples such as the one that talks about the slavery of African Americans in the US in which they talk about “ideological justifications.” In an entry about “gay porn“false information is shown indicating that the proliferation of these contents fueled the AIDS epidemic in the 1980s. In the entry on the genre, Grokipedia indicates that “gender refers to the binary classification of humans as males or females based on biological sex.” Wikipedia start entry stating that “Gender is the range of social, psychological, cultural and behavioral aspects of being a man (or boy) or woman (or girl), or a third gender.” In the image and likeness of Elon Musk. and the article about Elon Musk It contains 11,000 words and 300 citations/references compared to the 8,000 and 523 of its Wikipedia version. In both encyclopedias there are curiosities about that article, and for example in Wikipedia there is a section dedicated to Musk’s controversial greeting which is not on Grokipedia. And on the opposite side, Grokipedia does have mention of the “fart guy” controversy which is not available on Wikipedia. This is just the beginning. This version “0.1” of Grokipedia contains 885,000 articles, while Wikipedia has more than 8 million entries. In 2017 Elon Musk posted a tweet in which he praised the work of Wikipedia, but over time that perception changed, probably due to the comments included in the entry about him on Wikipedia. This year tweeted the message “Stop financially supporting Wikipedia until balance is restored!” The danger. Although Elon Musk assures that Grokipedia is open source and anyone can use it for free, it remains to be seen the ability that its users will have to edit articles created by AI. The risk is that this project poses a new attempt to control the conversation, and as he says entrepreneur Gary Marcus, “whoever writes the encyclopedia controls the narrative.” Jimmy Wales warns. The creator of Wikipedia, Jimmy Wales, indicated in an interview in The Washington Post a few days ago that he was curious to know what Grokipedia would end up being, but that he did not have too many expectations about the result. For him, AI language models “are simply not good enough to write encyclopedia articles. There will be a lot of errors.” Lauren Dickinson, spokesperson for the Wikimedia Foundation, explained in The Verge how “Wikipedia knowledge is and always will be human.” Problems for the free and human-created encyclopedia. Even so, Wikipedia is threatened by AI. Not only because this legendary online encyclopedia has been the great manual for training AI, but because it is suffering a traffic crisis. The xAI project is the latest attack on that source of knowledge and information, which, from being under control and editing completely carried out by human beings, now cedes those editing and writing tasks to xAI’s AI model, Grok. Image | dvids In Xataka | There is a reason why Wikipedia resists as the last human bastion against AI: because its editors rebelled

The race to put a humanoid robot in our house has begun. It’s an absurd race

by

A robot that walks around the house picking up what we have left lying around, loads the dishwasher and even starts the washing machine. It is not a science fiction movie, it is the advertisement of the Figure 03 and it is not the only company interested in sell us the idea that soon we will all have a home robot. Detective Spooner doesn’t like this. Robots for everyone There are people convinced that in a few years Humanoid robots will be as common in homes as robot vacuum cleaners are now. One of those people is Elon Musk, who assured that In five or six years we will all be able to afford a personal robot. Peter Diamandis, well-known writer and “futurologist” predicts that the first humanoid robots will reach homes as early as 2026. It is not an obsession of the West, In China they are also obsessed with robotics, although from a different approach. The government wants robots to have transformed the industry by 2035, but it also contemplates creation of robots as accompaniment within the home. We do not know if this future will materialize or if humanoid robots will end up being an eccentricity for a few. Regardless of whether they succeed, These are the companies that want to make it possible. Figure AI Figure 03 Based in California, it is the company that has shown the most progress in creating a humanoid robot for the home. Its latest model, the Figure 03, is presented to us as a kind of robotic butler that does all the housework. Until now the previous models did not go much beyond the “wow” effect of the video, but this time it is different because Figure has a plan to mass produce them. The first year They hope to produce 12,000 robots a yearalmost nothing. Figure is the spearhead of robotics in the United States. Its valuation is 39,000 million dollars and among its investors are NVIDIA, Salesforce, Qualcomm, Intel, Microsoft and Jeff Bezos himself. At the moment it is not for sale nor do we know the price it will have. tesla Tesla Optimus Gen 2 No introduction needed. The first time we learned that Tesla wanted to make a humanoid robot it was in 2021. In 2022 they had a functional prototype and in 2023 they presented the Optimus Gen 2. Although we have not seen him doing household chores, they did show how he was capable of handling fragile objects like an egg. According to Musk, the Optimus will be cheaper than a car (between 20 and 30,000 dollars), but the reality is that we are in 2025 and The promise has not yet come true. Musk continues determined to build “an army of robots” and just showed your worry about who will control him. In Tesla’s latest earnings call, he stated that he wants to maintain strong influence over this hypothetical army. 1X Technologies Neo Gamma It is based in California, but it is a Norwegian company. 1x’s goal has been the home from the beginning and its goal is a robot that does cleaning, organizing and even running errands. A year ago they presented the Neo Beta robot and in February of this year they presented the Neo Gammaits most advanced model. It is capable of interacting with humans, can manipulate all types of objects, and is covered in soft materials. 1X’s plan is to start deploying its robots in homes this year, but in a pilot project. The company has been set as a goal manufacture 100,000 units in 2027 and “millions more in 2028”. We don’t know anything about how much it will cost, although 1X says it is “expected to be priced competitively within the home robotics market,” whatever that means. The company is valued at 10 billion dollars and between your most powerful investors There are OpenAI and EQT. Unitree Robotics Unitree H2 Based in Hangzhou, it is one of the companies that form the ‘Six Little Dragons’ and leader in robotics in China. We knew her for her quadruped robotsbut recently they have moved on to humanoid robots. Its most advanced model, the Unitree H2was announced just a few days ago and is capable of dancing and even doing kung-fu, but it is not as focused on the home as other proposals. In China, spectacular demonstrations of robots that dance or box have become very fashionable, but for the moment They are not showing practical applications for these humanoid robots. Of course, it is the only one that already has humanoid robots for sale and at very competitive prices. The Unitree G1 costs $16,000, but the Unitree H1 costs 131,000 euros. Deep Robotics DR02 It is also a Chinese company and part of the ‘Six Little Dragons’, which are the six most cutting-edge companies in the country in AI and robotics. Like Unitree, they also launched quadruped robots and recently switched to humanoids. Its focus is the creation of resilient models so that they can work in sectors such as industry, logistics or public services. Their latest model is the DR02, a robot resistant to water and dust and is designed to work outdoors. In the future the company also wants to expand to other areas such as the home. What is the point of a humanoid robot? There are other voices at the opposite end of these visionaries, such as that of Rodney Brooks, the co-founder of iRobot. Brooks believes that humanoid robots are a fantasy and they are a format that is anything but practical. Keeping such a robot standing requires a lot of energy and can be a huge risk if it falls. Furthermore, he states that Imitating the dexterity of a human hand is practically impossible. For Ehsan Saffari, robotics engineer, There is no point in making human-shaped robots. At least not if we want them to be efficient. To illustrate this, he gives a very good example: “Imagine that instead of building a … 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.