Silicon Bidena curtain was in ruins. This movement of the Chinese gives an amazing advantage. What will Poland do?

– It is not possible to defend the billions of investment rounds and Stargate projects for a long metal, if you can have a free model at the level of the most developed American services – this is how technology expert, Artur Kurasiñski, comments for Next for NEXT .Gazeta.pl wypuszczenie przez Chiny za darmo potê¿nych modeli sztucznej inteligencji.

In recent weeks, the AI ​​industry has been experiencing a critical period for her. , on which artificial intelligence models are placed. The American Nvidia won 90 percent for itself. The market, so limits for the sale of AI chips are really severe for the whole world, even for Nvidia itself. Then, it seemed that the Americans chewed the whole world and for many years secured a strategic advantage in the area of ​​artificial intelligence. Not only that their algorithms seemed to be ahead of the competition by several lengths, they also dominated the hardware market and planned huge investments that would allow them to escape even more from the rest of the rate. The terror of the whole world and the excitement of Americans did not last long, because less than a week after Trump’s swearing in, China decided to break the silicon curtain.

Great Chinese model. They stole the fire of the gods from the Silicon Valley

First, which in several tests exceeded Chatagpt. Its authors claimed that training it cost him barely $ 6 million. . However, this is still about the government of size less than the American Big Techy spent – Meta invests about $ 65 billion a year in AI, and Openai (creators of Chatagpt) – at least 50 billion.

China, based on data stolen from Opennai

– says Artur Kurasiński, a technology expert, and technology author, technology expert in an interview with Next.gazeta.pl.

Investors reacted to this information with panic. . China may have made a favor to the world and pierced the speculative bubble that has accumulated around artificial intelligence. Time will tell.

After a few days, Deepseek stuck another pin into the Americans and released another AI model, this time multimodal (i.e. able to do more things than just generate text), which.

Here we have stolen the fire of the gods from the Silicon Valley, we are your Prometheus

– Leszek Bukowski, a philosopher, an expert in AI, wrote on X.

The Chinese showed the world that you can create an efficient AI for less money than the Silicon Valley does. What kind of deal could they have in this? Artur Kurasiński suspects that maybe in this way they wanted to show that preventing China from buying Nvidia cards will not stop them from developing their technology that despite blocking their purchase of hardware they are able to defeat the US in their game.

Deepseek has shown that the technological dam created by the American government does not work – it is not important for how much high -flyer (i.e. the hedge fund, which is the creator of this model), bought cards in front of the ban, it is also not important if he bought them quietly when he quietly In theory, the Chinese were cut off from Nvidia cards, it does not matter how much he trained its model. It is only important that a small player was able to do it, a company employing 160 scientists with a budget smaller many times than such powers as OpenAI, Meta or Google

– comments Kurasinki.

Chinese lesson on democracy

Deepseek has been made available to both AI models under an open source license, which means that everyone can download and use the code created in China without restrictions. This is another blow to hyperkapitalist big techs that stand on the eyelashes to monetize their AI models. Chatgpt shows its full capabilities only when we pay for access to the Pro version.

China democratized AI, releasing the first such powerful language model, which is 100 percent. Open Source

– says Kurasiński.

Thanks to the Chinese, everyone can download and put a powerful AI model on their own server, paying nothing to anyone and cut off from external suppliers, which prevents them from following our interaction with the model. Of course, if we secure it ourselves, because on our own server it is our worry – explains Kurasiński. He assures, however, that if the model is placed locally, no one has the opportunity to spy on us on a large scale. Kurasiński claims that this is a real revolution, because putting such a model on his own server provides us 100 percent. Safe and stable, and most important -free infrastructure.

The server, however, will be placed by few, most often, this opportunity will take advantage of various types of entities, such as corporations, research centers or foundations, less often private people. An ordinary Kowalski, Smith or Li will probably use an online agent, set on Deepseek servers, as today we use the chatgpt application or an online agent in a browser on OPENAI servers. There is probably a difference between the collection of our data by the Chinese start-up, almost certainly associated with the regime government and their gathering by American Big Techy. But how great this difference is, everyone must decide for themselves.

American Big Techy also spy on and control us – but they do it in a different way, under the applicable law in the USA

– says Kurasiński.

– Sharing the model as Open Source (but so real, not fake as the finish did it), gives the incredible advantage of the competitors of the largest American concerns. In the long run, billion investment rounds and Stargate projects cannot be defended in the long run, if you can have a functioning model for free at the level of the most developed American services, and putting locally, on your own server and having control over it costs 6,000. dollars. Silicon curtain (division into better and worse countries in terms of the supply of Nvidia cards), which the world of the American government Biden wanted to divide in ruins after less than a week – summed up the expert.

How will Americans answer China? What can Poland do?

Artur Kurasiński suspects that each company is already trying to experiment locally with the Deepseek algorithm, he learns his possibilities to introduce some improvement to their services in a moment.

This is a classic arms race. You can be a month or a year – then someone will chase us away

– says Kurasiński, when asked what the Americans will do now.

What can Poland do? – Look how the world leaves – unfortunately that’s all. If we had aware problems – really important – the ruling class, we would have put tens of millions for the development of research, and above all to support the commercial sector in terms of AI. Currently, we even have a language model of Bielik, which was created as a result of the work of several hundred people, but you still can’t see much will to use it, for example in administration – says Kurasiński. When asked if we can and should build our own model based on Chinese technology, he said that, of course, “the Polish government should sponsor the creation of our language model so that we would not be dependent on anyone.” He does not see any obstacles to this, except that Deepseek one day can change the distribution model and forbid using it.

Of course, this is a fundamental threat and it is worth having your own national model that depends on China or the USA

– summed up Artur Kurasiński.

Source: Gazeta

You may also like

Immediate Access Pro