This week, two geopolitical news are based on war: an analogical and unfortunately classic image, the return of tens of thousands of Palestinians from above and below to a devastated northern Gaza; and one that is more invisible with human eyes, with a dimension that is not yet well perceived: the business of artificial intelligence, competition and the commercial battle. Modern society knew that ChatGPT was not the only artificial intelligence, but we had already begun, they were, in the public media of this country, to humanize the technological device and call it Geppetto, by supposed grace. Beyond ChatGPT, modern society knew that there were more competitors in artificial intelligence – such as Perplexity AI and Notebook LM – but it had the appearance, and it was assumed, that ChatGPT would become the equivalent of what it is today in the browser and search engine business. There was little doubt that the Americans would control the entire artificial intelligence pie.

At the end of January, the tortilla takes the form of what it has turned, or can turn, into: A Chinese company has published artificial intelligence called Deepseek and turned everything upside down. In just one week, in the Google app store, it has had one million downloads worldwide, and last weekend it was the most downloaded app in the US, above the second ChatGPT. By the time the modern West has begun to talk about it, several multinationals have been hit hard on the U.S. stock market. In fact, China has shown that it can do better and cheaper programs than Americans.

Various details about Deepseek

Without the pretension of advertising, the new tool shows obvious differences compared to ChatGPT. First of all, it is necessary to create an account in the application in order to start using it. It can be downloaded to the mobile, to any smart device, but it can also be used in the browser.

In general, as reported in the Wall Street Journal, it is a tool with better performance, or at least similar. "Enough for normal people," with no apparent difference, and free, as has been defined in many places. Not only can it produce text, but it can also produce better images, according to the Reuters news agency.

Some errors in text bodies have also been reported in various digital portals. In English, for example, he has eaten ergative in one or another of the tests:

With more in-depth exploration, it’s inevitable that it uses open source. With open source, anyone can explore the interiors, and the functionality allows you to adapt and re-create to your needs. The practice of open source has put the market model on the kinka. Because algorithm chips are expensive, so far only large companies in Silicon Valley had the ability to make massive investments and influence the market.

Horde to Silicon Valley

Hit the stock market: Nvidia, which has become the jewel of Wall Street, has seen a 17% drop – a loss of $600 billion in just one day.

In fact, it seemed that Nvidia was a company that over the years has produced and designed the graphics cards of numerous computers, an indispensable company for ChatGPT and all the artificial intelligences known until now: it had the potential to produce a series of concrete microchips, the little card of algorithms that are the basis of artificial intelligences. Nvidia is located in Silicon Valley, an area where all the sauce was moving. Like I said, however: so far, Deepseek does not use these microchips. It is simpler in terms of production and therefore cheaper. According to several reports, the algorithms of the new tool are 90% cheaper.

The company behind Deepseek is also different from the monsters in Silicon Valley. El Salto tells in this article that it is a Chinese start-up, founded in 2023, with about 200 employees, which started from an investment of 10 million dollars. ChatGPT’s Open AI, on the other hand, has 4,500 employees and its initial funding was $6.6 billion.

The technology revolution has moved to Silicon Valley. Following the emergence of Chat GPT, starting in 2022, the stock market values of several multinationals multiplied: Nvidia increased by 840%, including Meta 580%, Google 105%, and Microsoft 85%. U.S. public investment is also on track: Donald Trump has announced $500 billion in grants to build and maintain data centers. The appearance of Deepseek has been a blow to the table.

Who is lying?

A large amount of information, sometimes contradictory and generally in accordance with particular interests and oppressive, has been disseminated in recent hours. For example, Deepseek says it has not had the economic power to buy Nvidia’s algorithm chip, and has trained its own algorithm, “better and cheaper,” hard over the past year. But several have said that China has vetoed Nvidia, which is why they have not bought the chip.

Julián Estévez, a professor at the UPV/EHU, made the following statements in the newspaper ElDiario: "China imposes sanctions on companies that buy Nvidia products, but it is suspected that Singapore has been used as an intermediary to train Deepseek, that is, that Singapore has bought chips from Nvidia and Deepseek has bought chips from Singapore."

In addition, the executive director of ScaleAI, a company dedicated to training and developing algorithms, a 28-year-old billionaire with Chinese parents, has come to understand that Deepseek is lying, that 50,000 Nvidia algorithm chips have been used to create the tool, and that it is no coincidence that it was once popularized in the U.S. when the president just changed. I mean, it suggests that China is planting a trade war.

If we could use less water...

The environment does not understand trade wars. And, if I had to put it on someone’s side, the Deepseek model would be the best of the worst. In fact, if what the Chinese company said is true, the artificial intelligences of this mold would not need giant data centers. That’s why Deepseek is cheaper: because it doesn’t need that chip, nor a huge maintenance.

Artificial intelligence such as ChatGPT hides an unsustainable social model, as explained several times in the Net Cerca section of LUZ: because this supposedly virtual and ethereal cloud or fog that is growing without limits to collect information needs enormous energy to survive physically, all the way from mineral extraction to having data center servers up and running. In response to the enormous demand for computing capacity, hyper-scale data centers are being created here and there.

To learn more about the ecological impact of artificial intelligence in the GPT Chat template, here is the article by Urko Apaolaza: