Deepseek, Chinese chatbot low cost that opens a new stage in artificial intelligence

Jane Anderson
Jane Anderson
Logotipo DeepSeek

The actions of the Microprocessor Manufacturer Company Nvidia They fell yesterday almost 17%, which generated the loss of 589,000 million dollars of stock market capitalization in a single day. The reason? The launch of the Chinese start-up of artificial intelligence Deepseek and its “low cost” chatbot.

The company has gained traction in record time, monopolizing the eyes of the business ecosystem, mainly due to the fact of having developed artificial intelligence models to a significantly lower cost than the big companies in the sector, such as OpenAi, Anthropic or Google.

Nvidia had positioned at the head of the most valuable quoted companies

To boost their artificial intelligence applications they use The nvidia chips, which has built its market positioning about the need to have state -of -the -art microprocessors to develop powerful models. So much so that until now he has occupied the first position in the list of the most valuable quoted companies in the world.

Deepseek: same performance, but cheaper

However, Deepseek’s arrival questions that business model and demonstrates that models and tools of AI can be launched with equally good operation as chatgpt or gemini, but with lower costs of creation and use.
Like the rest of its competitors, the Depseek assistant answers questions in conversational format, can summarize documents and use reasoning to resolve user consultations.

Differences reside, especially, in which it is a open source technology, what contrasts with the secretism that surrounds the models of large companies, and is gratuitous, so it is not necessary to subscribe to use it. The company’s business model is based on the collection of users who seek professional use of the tool.

The company’s most recent and powerful models are The R1 and the V3, and that would match in power and performance at GPT-4o of OpenAI. But, while large companies train their chatbots using around 16,000 chips, from Depseek they claim to need only 2,000. In addition, the development of the aforementioned models would have cost around 5.5 million dollars, compared to the nearly 80 million that required, for example, the training of ChatgPT-4.

Cost reduction, as collected The countryit is found in two different techniques. On the one hand, Depseek I would have learned from previous private models Applying the technique known as “distillation”, which allows asking questions to the previous model, save the answers and learn. And, on the other, use Less computing and energy capacity When offering responses to user consultations.

China in the career for artificial intelligence

Deepseek’s success, which was launched in 2023 by Entrepreneur Liang Wenfeng In Hangzhou, it is understood as a Chinese response to the progress of the United States in artificial intelligence. Moreover, its development arose to the Export restrictions of chips imposed by the US government and that aimed to limit foreign competition.

The hatching of the Chinese firm, in addition, occurs alongside the announcement made by Donald Trump about the Stargate project, through which it intends to invest up to 500,000 million dollars over the next four years with the intention of promoting artificial intelligence projects.

Now Depseek has opened the door to a new stage in the artificial intelligence industry. In recent weeks it has become one of the most downloaded applications in the United States, above Chatgpt, and its market impact threatens the advantage with which great names such as Nvidia and Openai departed, which would aggravate the commercial war between the great powers.

At the same time, the cheapest artificial intelligence in Deepseek implies the progress in the technology comfort, that will be even more accessible to the general public, as well as for companies. It also opens new debates about security and privacy. Moreover, at the time of writing this news, the alert platform for large -scale malignant attacks on its services, which could be generating friction in the use of the chatbot or the registration process.