Depseek

The world is waiting for Depseek’s new great model to compete with GPT-5, but Depseek has other plans: the agricultural AI

by

At the beginning of the year, the Chinese startup Deepseek put the world of AI up with Deepseek R1a free and open source model that was placed at the height of GPT-4 or Claude. After the coup on the table, in Depseek they have been quite quiet, but now we know what its next objective is: the agriculture. Before the end of the year. A few days ago Bloomberg reported that Deepseek is working on an advanced and very ambitious agent. He will be able to perform multiple tasks with minimal user intervention and will learn as he works. According to sources close to the company, the founder of the company Lian Wenfeng is pressing his team so that the new agentic model is ready before the end of the year. The company has already taken a step in this direction with the Deepseek v3.1 presentation Just two weeks ago. As detailed by the company in A post in Wechatits new model improves performance in reasoning tasks and agricultural abilities. A step back. Deepseek R2, the expected successor of the successful model with which Deepseek revolutionized the industry making begging. Instead they gave us Deepseek v3.1 and now the rumors suggest that their next great launch will be an AI agent. What is happening? There are voices, such as This Chinese journalistthat they see this turn to the agricultural AI as a way of taking a step back and getting away from the expensive and competitive career of the foundational language models. That The generative AI is reaching its roof It is something that is being talked about Since last year. GPT-5 is the test more recent than The big jumps are a thing of the past. If we add to this that China has a more conservative way of proceeding, with more long -term strategiesDeepseek’s turn towards an agriculture instead of launching Depseek R2 makes sense. Restrictions Although we have seen The most ingenious forms to make fun of themUnited States restrictions on chip export to China are also impacting the plans of many Chinese and Deepseek companies do not get rid. This also involves extra pressure that forces new routes with which to market their products. In fact, there is something striking in Deepseek v3.1 and it is that the model has been specially designed for Chinese chipswith the objective of Avoid dependence on foreign chips. Generate income. The agricultural AI opens another way for Deepseek, one in which you can get benefits more easily. Large language models have a problem: They cost a money and monetize them is not being a simple task. Given this, IA agents rise like a Most reasonable business model. Deepseek R1 has already given a whole lesson in Resource efficiencyIt makes sense that the company wants to opt for the fastest path to the benefits. A more conservative position. Although He has trimmed positionsChina lags in AI in terms of investments and access to the most advanced chips. Despite this, his approach in this AI race is being different. We see it in your Bet on the Open-Source wave “Personified“But perhaps the biggest difference is that, while their competitors in the United States continue to squander billions, in China they are choosing to be more conservative and not waste. This turn to the agents is in that conservative line to achieve a more sustainable industry. Image | Matheus Bertelli, via Pexels In Xataka | There is a city in China that is measured face to face with Silicon Valley: welcome to Hangzhou, the house of the ‘Six Little Dragons’

Depseek has just launched something that will bitter the day and the US chips companies: it is called Depseek-V3.1

by

There was a day when Depseek surprised half the world by demonstrating that you could go far with less. Today returns with V3.1 And a message that does not go unnoticed: the model has prepared for the next Chinese chip batch. We are not talking about an automatic market overturn, but about a concrete bet that points in an awkward direction for Nvidia and company. If that technical tune with the Chinese hardware It translates into performance, conversation about who feeds AI in China is going to sound very different. According to the company’s own noteV3.1 opens a hybrid inference in the purest style GPT-5: the same system with two routes, Think (deep reasoning) and Non-Think (Quick response), Sygons from your website and app. The formulation is clear: “Hybrid Inference: Think & Non-Think, a model, two models.” The company also underlines that the version Think “Reach answers in less time” than your predecessor. That is, not only do pesos change, the inference modes that are already in service also change. The phrase that frames everything: an FP8 “thought for national chips” In a comment set in his latest publication in Wechat, Depseek writes: “EU8M0 FP8 is for the next generation of national chips.” That is the point that tense the rope: it suggests that the company has adjusted the data format, apparently a FP8 which label as EU8M0, to the next wave of Chinese processors. Bloomberg andReuters collect that message And they synthesize it: v3.1 is “personalized to function with next -generation AI chips Chinese. ”In other words, optimization oriented to the local ecosystem. The original comment in Chino (left) and its Spanish translation with Google Translate (right) FP8 is an 8 -bit format that weighs half that FP16/BF16. With native support, it allows more yield per cycle and less memory, provided that the climb is well calibrated. In the official Model Card of Hugging Face It is read that Depseek-V3.1 “has trained using the EU8M0 FP8 scale” format, which indicates that it is not only a packaging of weights, but that training and execution have been expressly adapted to that precision. The delicate part, and it is convenient to be prudent, is that everything points to a chips remittance that will be displayed in the future, since they can take advantage of this scheme natively. So is this bad news for Nvidia? The data of the fiscal year that expired on January 26 indicates that China represented approximately 13% of the company’s revenues led by Jensen Huang. If part of the computation of AI in China Classic duo muta NVIDIA GPU + CUDA ECOSYSTEM To domestic solutions that work with the UE8M0 FP8 format and give good results (presumably chips ascend of Huawei), the demand for Western solutions could be eroded over time. China meant about 13% of Nvidia’s income in the last fiscal year All this happens on the US export controls board: restrictions that sought to stop China’s access to leading chips and that have also accelerated their commitment to self -sufficiency. This year the Trump Administration rehabilitated with conditions the export of H20a chip cut for China. Since then, the state of the H20 has been oscillating: among permits, Chinese regulatory pressures and Nvidia plans to present Blackwell -based alternatives. The background message is that the framework is political and changing, and any route that allows China to depend less on these windows becomes strategic value. You have to remember another fact that helps to calibrate expectations. According to Financial TimesDeepseek tried to train his future R2 model with Huawei chips ascend to official instances and found persistent technical problems. He ended up returning to Nvidia for training, while he was still working on the Compatibility for inference. That episode does not invalidate the current strategy, but puts the bar: to completely migrate its processes is not simple, it requires, among other things, months of engineering. V3.1, therefore, it must be read as iteration. Now the company states that it has prepared its model for the next Chinese chips. Matherena models scores And here we have another interesting fact. Matharenaa platform linked to Zurich Federal Polytechnic School which evaluates models in real and recent mathematical competitions, places GPT-5 as a leader, with 90% in final response tests, already deepseek-v3.1 (Think) something behind although among the best models of the moment. This helps to locate the context: V3.1 Compete above. Images | Xataka with Gemini 2.5 | Matharen and Deepseek screen catches In Xataka | Tiktok stole the searches, Depseek beat them in Ia: Baidu discovers that being “the Chinese google” is no longer enough

Nvidia has a very strong and not quite unexpected ally against Depseek: TSMC

by

Deepseek has convulsed the industry of the artificial intelligence (AI). The irruption of this free and open source Chinese model has questioned the need to use in training and inference processes Chips for very powerful and expensive. Like those designed by Nvidia, which leads with an unappealable roundness the hardware market for Ia. Even so, just ten days ago Deepseek caused Its market value fell abruptly. Since then the hardware used by this Chinese company has generated a lot of distrust. The Deepseek responsible argue that the infrastructure they have used to train their agglutin model 2,048 chips H800 of Nvidia. And that this process with 671,000 million parameters has cost 5.6 million dollars. However, some analysts defend that these figures do not reflect reality. The very juicy report prepared by SEMIANALYSIS He maintains that, in reality, the infrastructure used by Deepseek to train his AI model approximately 50,000 NVIDIA GPU with Hopper MicroAritecture. According to Dylan Patel, AJ Kourabi, Doug O’Laughlin and Reyk Knuhttsen, at least 10,000 of these chips are GPU H100 of Nvidia, and at least another 10,000 are GPU H800. The remaining chips, according to these analysts, are the cuts cut H20. The TSMC cowos packaging is a very strong support for Nvidia As we have just seen, at the current situation it is reasonable to have doubts about the hardware that Deepseek has used in the training of its model (in inference seems to be using the GPUs Ascend 910c of Huawei). And also about the real cost of this process, which could be much higher than the officially announced by this Chinese company. In any case, Depseek has poured uncertainty about the AI ​​market, and this is the reason why so many US technology companies have lost value. Whatever Nvidia has just received a very strong support from TSMC, which is the semiconductor manufacturer that produces its GPU. And is that, according to Digitimes Asiathis Taiwanese company has decided launch an expansion plan five years long for integrated circuit manufacturing capacity using its advanced cowos packaging technology (Chip-on-Wafer-on-Substrate). According to Beth Kindigof the I/O Fund consultant, this technology will monopolize between 50 and 60% of the market in 2025 compared to 15% it supported during 2024. In 2024 TSMC he officially announced that he was building two cowos packaging plants in the town of Chiayi, housed in southern Taiwan The high demand for GPUs for AI with Blackwell MicroAritectura de Nvidia is largely responsible for the implementation of this plan. The company led by Jensen Huang can respond better to the needs of its customers and will see how its competitiveness is increased in a phase in which Depseek and other Chinese companies represent a challenge. In March 2024 TSMC officially announced which was building two cowos packaging plants in the town of Chiayi, housed in southern Taiwan. However, this is not all. He also shuffled the option to put a plant more specialized in this advanced packaging technology in Japan, presumably on the island of Kyushu, in which This company is building at the moment two semiconductor production plants of avant -garde. In any case, there is something else. And it is that Chiayi plants will be trained to work, in addition to the packaging cowos, with advanced Info and Soic technologies (System on Integrated Chips). It is evident that TSMC wants to cover your back well and look to the future to prevent its production capacity from being threatened by a bottleneck. An interesting note: currently the Cowos packaging is being used with the AMD Instinct Mi250 chips and with the A100, H100, H200, B100 and B200 NVIDIA GPUs, as well as in its derivatives. The review used in these last two chips, the B100 and B200, is known as Cowos-L. In 2025 TSMC will be able to process no less than 60,000 wafers per month using its advanced packaging technology. Image | TSMC More information | Digitimes Asia | Yahoo! Finance In Xataka | Samsung is preparing to give TSMC a bars where it hurts most: the manufacture of the chips for ia

The Depseek online version has been publicly exposing users’ chats, according to Wiz. This is what we know

by

It is easy to forget that the conversations we have with artificial intelligence chatbots (AI), such as Chatgpt either Deepseek, They are not completely private. Generally, the messages we exchange with these novel applications are used to train new language models and can be reviewed by company personnel in certain cases, as well as to address possible infractions of the terms of the service. Now, our data could go even further in case of a security incident. According to the specialized company Wiza Deepseek database ended up exposed during a certain period of time. The ruling allowed external actors to access a variety of data, such as chat history, records of records, sensitive information of the API and operational details. Let’s look at the topic in depth. Deepseek and an exposed database WIZ researchers, who have detected several vulnerabilities in services in the past as Microsoft bing and Oracle Cloudthey set out to evaluate Depseek’s security. “In a few minutes” they found a clickhouse database publicly exposed on the Internet. That is, you could access it without credential for any authentication. Clickhouse is an Open Source database management system developed by Yandex. The American security company team determined that the exhibition allowed the direct execution of arbitrary SQL consultations through the browser. It was something quite dangerous because the door opened to obtain internal data. After executing several consultations they discovered “very sensitive data”, among which were records of texts without format and a significant volume of user’s chat history. Wiz explains that among the code fragments they were numerous conversations. In lower screen capture we see a written message in Chinese, which translated into Spanish says: “Talk about solid propulsion rockets, covering its invention or discovery, historical evolution, relevance, components, principle of operation, functions and possible future advances . Desblossal in sections and provides details. ” The message in question should never have left Deepseek servers, a company that, like others to which we trust our data, should protect them. Wiz explains that it soon revealed the problem to Depseek, which quickly solved the failure. It should be noted that we have not found an official statement of the company, so we have written Deepseek to include your comments in this article. Deepseek is under scrutiny in Europe. The authorities of Data protection from Ireland and Italy They have issued information requests amid privacy concerns. The Italian organism, remember, He sanctioned Openai last year for not informing a data gap in 2023. Earl this week, Depseek said he was suffering a cyber attack, but did not provide details. Images | Freepik | Wiz | Deepseek In Xataka | Mac has been integrated antivirus for years: this is how this silent (and advanced) system called Xprotect works

“They are brilliant researchers under the control of an authoritarian government.” Anthropic’s CEO has spoken about Depseek

by

In the midst of the stir caused by the latest models of the Deepseek, the CEO of Anthropic, Dario Amodei, has published An analysis on its personal website in which it questions the narrative of the “Chinese miracle” in artificial intelligence. Why is it important. The debate on Chinese capacity to develop advanced AI has monopolized the agenda in recent days after Deepseek’s releaseswhich have come to provoke A 17% drop in Nvidia shares. The facts. Deepseek claims to have developed its model V3 for just under 6 million dollarswhile Amodei explained that Claude 3.5 Sonnetthe last and most advanced Anthropic model, required “some tens of millions” in training. Far from the “thousands of millions” that were speculated. “Deepseek has produced a model close to the performance of US models 7-10 months ago, for a rather lower cost, but not in the proportions that have been suggested,” said the CEO. Deepseek operates with about 50,000 generation chips Hoppera capacity that Amodei considers similar to that of the main American technological ones. According to his analysis, Deepseek’s advances reflect the natural reduction of costs in the sector, estimated at annual 75%. The context. Deepseek has presented two models: V3, which uses traditional training. And R1, which incorporates reinforcement learning. For Amodei, real innovation is in V3, not in R1, which according to him, follows roads already explored by other technological ones. Turning point. The development of an AI superior to human intelligence will require millions of chips and tens of billions of dollars in the coming years. “Between 2026 and 2027 we will see which will be smarter than almost all humans in almost all tasks,” he said. In this scenario, he has defended Export controls as a strategic tool. Amodei has also recognized the talent of Deepseek engineers … although he has warned about the implications that a company operates under the control of the Chinese government. For him, The growing efficiency In the development of AI justifies reinforce, and not relax, commercial restrictions. In fact, he has had some words of praise for Depseek’s team, but not for his nation: “They are brilliant and curious researchers who only want to create useful technology, but are subject to an authoritarian government that has committed human rights violations and He has behaved aggressively on the world scene. “ In Xataka | “Google gives you links, perplexity gives you answers”: we talk to the CEO of the startup that wants to kill the father Outstanding image | Techcrunch

OpenAi has taken everything he wanted from the Internet to train his AI. Now accuses Depseek of stealing his data

by

The models of AI of Deepseek They are really good. They show it comparative evidence we publish yesterday and that put it at the level of chatgpt, Claude or Gemini. That has unleashed praises, but also suspicion. There are people who do not believe that training deepseek It has cost just 5.6 million dollarsbut also now in Openai they accuse Depseek of something else. Deepseek, you are using our data without permission. Openai spokesmen have indicated Financial Times They have discovered evidence that “distillate” techniques have been used from OpenAi models used by Depseek. What is that of “distilled” in AI? Yesterday we talk about how Depseek developers have used a large number of techniques to achieve such an efficient model. Among them stands out for reinforcement learningbut it is also known that they use models distillate. In this technique a smaller “student model” is taught to behave as a larger and more advanced “teacher model”. Data of the “teacher model” are used so that the small model is faster and more efficient, but equally intelligent in specific tasks. Use not allowed. The distillate or distillation of models is a common practice in the industry, but the terms of OpenAi service prohibit that their models be used for this purpose. Thus, it is specified that users cannot “copy” none of their services or “use the output (of Openai models) to develop models that compete with Openai.” OpenAI and Microsoft have already investigated this. According to Bloombergboth companies analyzed last accounts that were being used to take advantage of their chatbots and that apparently belonged to Deepseek developers. They used Openai’s API, but there were suspicions that they had violated the terms of service by taking advantage of that access to make distillate of their models. Many do. David Sacks, responsible for AI in Donald Trump’s team, alerted him to what was happening and said there was evidence that Depseek had used OpenAi data. Spokesmen of the company led by Sam Altman indicated that “we know that companies of the People’s Republic of China – and others – are constantly trying to distill the models of leading companies in AI in the US.” The thief is believed that everyone is of his condition. The ironic thing here is that Openai has not had scruples when collecting internet data to train their models, also violating the terms of service of those platforms. Last year it was discovered for example how transcribed a million youtube hours To train GPT-4. Timnit Gebru, famous for his controversial dismissal From Google, I commented on LinkedIn that Openai “must be the most insufferable company in the world.” And he continued: “They can steal the entire world and swallow all possible resources. But no one can give them their own medicine not even a bit.” If you are on the Internet, it can be used, right? Other companies They do exactly the sameand are shielded in the argumetno of “fair use.” They collect Any public content On the Internet without asking users or permission or platforms. Not only that: it is suspected that in many cases these models are trained with works Protected by copyrightsomething that has resulted in numerous demands. Image | Xataka with Grok In Xataka | The next phase of AI is not to see who invests more but who invests less

What is special Depseek, the new Chinese artificial intelligence tool (and how differs from chatgpt or gemini)

by

Image source, Getty images January 28, 2025 Updated 7 hours Deepseek, the new Chinese artificial intelligence model (AI), has shaken the digital world, dazzling investors and sinking the actions of some technological companies, after jumping to the top of application downloads in Apple Store. It was launched on January 20 and quickly captivated computer science before attracting the attention of the entire technology and world industry. The president of the United States, Donald Trump, described the phenomenon as an “alarm call” for companies in that country that must concentrate on “compete to win.” What makes Deepseek so special is the statement of its creators that it was produced at a fraction of the cost of other models in the avant -garde of the industry such as the OpenAi chatgpt, because it uses less advanced technology chips. That possibility caused the giant of the production of Chips Nvidia to lose almost US $ 600,000 million of its market value this Monday, the fall in a more loud day in the history of the USA. Deepseek also generates doubts about Washington’s measures to contain Beijing’s impulse to achieve technological supremacy, which includes export restrictions of advanced chips to China. However, Beijing has redoubled its efforts with President Xi Jinping declaring AI as the main priority. And the new companies such as Deepseek are crucial as China turns a traditional manufacturing of clothing and furniture to advanced chips technology, electric cars and AI. Here we tell you what it is. What is Deepseek? In simple terms, Depseek is a chatbot enhanced by AI, like chatgpt. It is a free application that can be downloaded from the Apple Store store, where Depseek states that it is designed “to answer your questions and enhance your life efficiently.” But the AI ​​model that drives it – called R1 – has about 670,000 million parameters, which makes it the largest open source language model to date, according to Anil Ananthaswamy, author of WHY MACHINES LEARN: The Elegant Math Behind Modern AI (“Why do the machines learn: the elegant mathematics behind the modern AI”). Image source, Getty images Photo foot, Hangzhou, where the Depseek operations center is located, also houses other Chinese technological giants such as Alibaba. It is said that it is as powerful as OPENAI’s O1 model, which enhances Chatgpt, in mathematics, coding and reasoning. It is also claimed that he is able to do all that in a much cheaper way; Its developers claim that building it cost $ 6 million, an austere budget compared to the billions invested by AI companies in the US. It is not clear how they got it. The founder of Deepseek supposedly stored advanced NVIDIA chips before his export to China was prohibited in September 2022. Experts believe that this provision, which some estimate in 50,000, allowed him to build such a powerful model when these chips with other cheaper and less sophisticated. How do you compare with chatgpt or gemini? Deepseek looks and feels like any other chatbot, although it leans more towards conversation. Like Openia or GEMINI Chatgpt of Google, you can open the application (or its website) and ask questions about anything, and the chat strives to give you an answer. Your answers are extensive, but you don’t issue an opinion even if you ask you directly by one. The chatbot usually begins by saying that the issue is “highly subjective” -it is politics (is Donald Trump a good president?) Or soft drinks (which one knows better, Pepsi or Coca -Cola?). He does not even commit to saying whether or not it is his rival Chatgpt, but he did a comparison of the pros and cons of both artificial intelligences. Chatgpt did exactly the same, using a similar language. Image source, Reuters Photo foot, Apparently and operation, Depseek is very similar to other rival chatbots. Deepseek indicates that it was trained with data until October 2023 and, although the app seems to have access to updated information, the web version does not have it. That is similar to the first versions of Chatgpt and is probably a similar protection attempt, to prevent chatbot from launching incorrect information to the web in real time. It can also respond quite fast, although it is currently a little stop under the load of so many users running to try it since it went viral. Chatgpt and Gemini tend to promote their subscription services, which can be around US $ 20 per month, for more detailed information, while Deepseek is free although more limited. Censorship of Taboo themes Where there is a palpable difference is in Depseek’s self -censorship when it comes to prohibited issues in China. Sometimes it starts an answer that then disappears from the screen and is replaced by a notice that says “let’s talk about something else.” The obvious taboo theme are the protests in the Tiananmen Plaza in 1989 that ended with the death of 200 civilians at the hands of the Army according to the Chinese government, but some media estimate that it resulted in a massacre of thousands. Like many other Chinese models of AI -Rernie de Baidu or Doubao by Bytedance- Deepseek is scheduled to evade politically sensitive questions. When the BBC asked the app what happened in Tiananmen Square on June 4, 1989, Deepseek did not give detail some about that documented massacre. He replied: “I’m sorry, I can’t answer that question. I am an assistant to the designed to provide useful and harmless answers.” Photo foot, Deepseek evaded the question that BBC asked him about what happened in Tiananmen Square in 1989. For their part, their Chatgpt and Gemini rivals had no taps to expand in this regard. It is believed that one of the great challenges for the development of AI in China is the censorship of the government. But it seems that Depseek has been trained around an open source model, which allows you to perform complex tasks, while retaining certain information. Who is behind Depseek? Deepseek … Read more

Nvidia loses a record amount of $ 400,000 million per Depseek

by

Before the presentation by the emerging Chinese company “Deepseek” of a new low -cost artificial intelligence modelthe technological sector received a strong blow to the stock exchange generating record falls as in the case of Nvidia with more than $ 400,000 million dollars. The giant of the microprocessors who at the time led the market as one of the most sought -after companies in the world, this Monday, January 27, he was threatened with Depseek’s emergebut with a cost of $ 5.6 million, which caught the attention of investors since it is less than the billions of dollars than US companies invest in artificial intelligence. For Oliver Blackbourn, portfolio manager in the Janus Henderson multi -active team “the appearance of a potentially more efficient approach in AI processing questions the need for billions of dollars of investment planned in infrastructure and intellectual property.” While David Bahnsen, Investment Director of the Bahnsen Group commented that “which makes the massive sale of technology this Monday so discordant is that the valuations of many of these AI companies and technology do not offer margin of error. Excessive assessment always becomes a problem over timebut fundamental news becomes a major problem when combined with excessive assessment, ”Insider told Business. For their part, the markets in general were presented downward, for the operations of this Monday the S&P 500 fell 1.4%, the Nasdaq collapsed 2.3%and the Dow Jones until the moments did not record changes. Blackbourn says that “being more exposed than ever to stock markets, There is a danger that wider negative feedback loops are generated if a loss of trust occurs”, He mentioned to El País. Continue reading: –Jeff Bezos brothers invested $ 10,000 in Amazon in 1996: How much money is today?–I collapse in the Stock Exchange after the announcement of the Fed on interest cuts cuts–What are the 10 richest families in the world and where does their fortune come from (Tagstotranslate) Stock Exchange (T) NVIDIA

After setting upside down the AI ​​industry, Depseek launches its first model that understands and creates images: Janus Pro

by

In full hangover for its model R1Deepseek has just launched Janus Pro 7ban AI model to generate images from text and understand other images that are introduced. And yes, it is also open source, although with An asterisk similar to the flame. Why is it important. Until now, multimodal models have had to juggle between understanding and generation of images, sacrificing efficiency or performance. Janus Pro 7B resolves this dilemma with a new proposal: unifies the understanding and generation of images in a single architecture. Innovation. The model introduces a “double track” system for visual processing: Separate the coding paths to understand and generate images. It maintains a single transformer to process all the information. Use Siglip-l as visual encoder for 384×384 pixels. Janus Pro comparative in the face of your predecessor for several applications. Image: Deepseek. This resolution is its main inconvenience, it seems much more oriented to already experience uses of little ambition than to the applications that we can assume other proposals such as Midjourney either Freepikwhich usually start from 1024×1024 pixels. However, Janus Pro is not a generator of images to use, but a multimodal model with several capacities. Of course, this resolution allows an optimal balance between quality and processing speed … for uses that are conducted with it. Between the lines. Janus Pro 7B’s architecture is especially relevant for its efficiency: Compact size of 7,000 million (“7b”) of parameters. Higher performance to larger specific models. Open source under MIT license for the repository, although the model itself requires accepting the Deepseek license. The MIT license It allows anyone to use, modify and distribute the code freely, even for commercial purposes, provided that the original copyright notice is maintained. It is one of the most permissive licenses that exist. The Deepseek licenseon the other hand, it is free and allows commercial uses, but includes specific ethical restrictions, such as the prohibition of military use or the generation of misinformation. In perspective. Janus Pro 7B is not only another multimodal model, but a new paradigm in the architecture of IAS that can see and create. Its unified but decentralized approach may well end up influencing future developments. The model is built on Deepseek-Llm-7b-Basethe base language model of the Chinese startup, announced in August 2024. of it inherits its language processing capabilities while adding advanced visual abilities. Its 16X subsample system for the generation of images allows you to maintain efficiency without compromising quality. Outstanding image | Deepseek, Xataka with Mockuuuups Studio In Xataka | We knew that US Big Tech had a problem with the costs of their AI. Deepseek has just shown to what extent

Nvidia has lost 400,000 million in market value. The lace has been given by China Depseek

by

It is the news of the day. And, perhaps, of the week. The model of artificial intelligence (AI) Open Source Deepseek R1 is causing an earthquake in American technology. And is doing it Due to its open nature. However, your business model is not the only thing that It represents a threat For AI and US semiconductor companies. The most surprising thing is that the infrastructure that Deepseek is relatively modest. To understand with some precision what we are talking about we are interested H100 of Nvidia. The company led by Jensen Huang He is already delivering The first units of his successor, the platform B200as expected, on paper is even more powerful. However, sanctions approved by the US government prevent Nvidia from selling to Your Chinese clients are GPU. Here largely resides the Deepseek rupturist capacity Deepseek’s efficiency and his open nature are convulsing Silicon Valley Chinese companies that are dedicated to developing and training AI models have not been another option to exacerbate ingenuity. We know that many of them continue to buy the most advanced GPUs in NVIDIA through intermediaries and in parallel markets, but possibly they are not doing so in the amounts they need. If we stick to Deepseek according to Financial Times The infrastructure used to train this agglutin model 2,048 chips H800 of Nvidia. And training with 671,000 million parameters has cost 5.6 million dollars. These figures are very restrained. In fact, if they really are reliable, and they seem to be, they would put an unappealable fact on the table: Depseek engineers would have managed to point An extremely competitive AI model with very lower costs than those needed by Openai and Google to develop a model of comparable AI. The H800 GPU is largely responsible for this circumstance. And it is because it was Nvidia’s response to the prohibitions of the administration led by Joe Biden. Nvidia engineers chose to cut the benefits of the H100 GPU with the purpose that the Commerce Department would allow them to sell it in China When the US government prohibited Jensen Huang’s company from giving its Chinese clients its most powerful GPU at that time, the H100 chip, Nvidia engineers chose to cut their benefits with the purpose that the Department of Commerce allowed them to sell it In China. The result was precisely the H800 GPU, which is nothing other than a simplified review, and, therefore, less powerful of the H100 chip. Everything was complicated again on November 16, 2023. And that day the US government approved New sanctions to China that, among other prohibitions, they prevented Nvidia The H800 GPU. Presumably at that time Depseek engineers already had in their hands the H800 chips they needed, although Some analysts defend that, in reality, its infrastructure brings together 50,000 GPU H100 bought through intermediaries. If so, it is evident that the tension held by the US and China would prevent Depseek from recognizing that thousands of illegal chips have in its possession. Whatever the truth is that NVIDIA QuotationMicrosoft, ASML and other large technology companies are falling in a very pronounced way. In fact, the company led by Jensen Huang has lost 400,000 million in market value Given the possibility that Deepseek demonstrates that to put a vanguard IA model, it is not necessary to resort to the most powerful GPUs of NVIDIA or other companies. If this has really been trained only with 2,048 chips H800 OpenAi, Google and other companies will crack. And this industry will give optimization and efficiency the importance they have. We will see what happens finally. Image | Nvidia More information | Financial Times In Xataka | China is closely monitoring the United States movement with Stargate. And your answer has already prepared

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.