Is it China’s tool as good as it seems?

Deepseek, a Chinese application to-chaatbot that was launched last week, has caused chaos in US markets and raised questions about the future of the Domain of the AI ​​of the United States. The BBC analyzes how the application works.

Deepseek looks and feels like any other chatbot, although it leans towards being too talkative.

As with the OpenAI or GEMINI chatgpt of Google, open the application (or the website) and ask you questions about anything, and do everything possible to give it an answer.

Give long answers and will not be attracted to express an opinion, however, one is requested.

The chatbot often begins his answer by saying that the issue is “highly subjective”, is it political (is Donald Trump a good president of the United States?) Or soft drinks (what is tastier, Pepsi or Coque?).

He did not even undertake if it was better or not that the artificial intelligence assistant (AI) of Openai, but weighed the pros and cons of both: the chatgpt did exactly the same and even used a very similar language.

Deepseek says that it was trained in data until October 2023, and although the application seems to have access to current information, such as today, the version of the NO website.

That is not different from the previous versions of Chatgpt and is probably a similar attempt to safeguard: stop the chatbot by throwing erroneous information bombarded on the web in real time.

It can be quite fast in your answers, but it is currently moaning under the weight of so many people who rushed to try it as you have gone viral.

But there is an area where his rival of the United States is not similar: Deepseek is censored when it comes to questions about prohibited issues in China.

Sometimes an answer begins, which then disappears from the screen and is replaced by “let’s talk about something else.”

Obviously, a taboo subject are the 1989 protests in the Tiananmen square that ended with 200 civilians killed by the military according to the Chinese government; Other estimates have oscillated between hundreds to thousands.

But Deepseek will not answer any questions about it, or even more widely about what happened in China that day.

The chatgpt developed by the United States, in comparison, does not contain its answers about Tiananmen Square.

Kayla Blomquist, researcher at the Oxford Internet Institute and director of the Oxford China Policies Laboratory, says “relatively speaking”, the Chinese government has been “no doubt” with the application.

“I would say that there is a change since we have seen an announcement in a large investment of the central government just in the last week, so it will probably indicate a change in the future.”

Deepseek comes with the same warnings as any other chatbots with respect to precision, and has the appearance of the US assistants already used already used by millions.

For many, especially those who do not subscribe to top -level services, you probably feel more or less the same.

Imagine a mathematical problem, in which the true response runs to 32 decimals, but the shortened version extends to eight.

It is not so good, but for most people, that will not matter.

It may be the case, it has managed to reduce costs and calculate, but we do know that at least it is built on the shoulders of the giants: it uses Nvidia chips, although cheaper and cheaper versions, and uses the open source of goal architecture of Flames, as well as the equivalent Qwen from Alibaba.

“I think this absolutely challenges the idea of ​​monetization strategies that many US firms have had,” said Blomquist.

“It is pointing to possible methods of development of the model that are much less calculating and intensive in resources that would potentially point out a change in the paradigm, although that is not confirmed and remains to be seen.

“We will see what they bring the next two months.”