Deepseek, the Chinese A-Chatbot app that was launched last week, caused chaos in US markets and asked questions about the future of the American AI dominance. BBC looks at how the application works.
Deepseek looks and feels like any other chatbot, though he tilts that he is overly chatty.
As with Openai Chatgpt or Google’s twin, you open an app (or website) and ask her questions about anything, and does your best to give you an answer.
It gives long answers and will not be attracted to expressing opinions, but it is directly sought.
Chatbot often begins its answer, saying that the topic is “very subjective” – whether it is a policy (is Donald Trump a good US president?) Or soft drinks (which is tastier, pepsi or coke?).
They would not even be obliged to say if it is better or not, or not, the CHATGPT assistant (AI), but measured the benefits and disadvantages -Chatgpt did exactly the same and even used a very similar language.
Deepseek says it is dressed for data until October 2023, and although the application seems to have access to current information, such as today’s date, the version of the website no.
This is no different from the earlier versions of Chatgpt -Ai is probably a similar attempt to protect – to stop the chatbot that is a real -time misinformation pump on the Internet.
It can be quite quickly in his answers, but at the moment the moaning under the weight of so many people who rushed him to try him out because he became viral.
But there is one area where nothing is similar to his rival – Deepseek censorship when it comes to issues on topics forbidden in China.
Sometimes the answer begins, which then disappears from the screen and replaces it “Let’s talk about something else.”
One obviously taboo topic was protests in 1989 at Tiananmen Square, which ended up with 200 civilians killed by the army to the Chinese government – other estimates ranged from hundreds to many thousands.
But Deepseek will not answer questions about it or wider about what happened in China that day.
Chatgpt developed in the US, in comparison, does not refrain in answers about Tiananmen Square.
Kayla Bluquist, researcher of Oxford Internet Institute and Director of Oxford China Politics Laboratory, says “relatively speaking” Chinese Government “took away” with the application.
“I would say there is a shift because we have seen the announcement in huge investments from the central government only in the last week – so this is likely to signal a change forward.”
Deepseek comes with the same warnings as any other chatbota in connection with accuracy, and has the appearance and sense of more established American assistants of AI already used by millions.
For many – especially those who do not subscribe to top services – it probably feels almost the same.
Imagine a mathematical problem, in which the right answer moves to 32 decimal places, but the shortened version is moving to eight.
It’s not so good – but for most people it won’t matter.
He may have been able to reduce the costs and calculate, but we know that it was built at least partly on the shoulders of the giants: the uses Nvidia chips – although older, cheaper versions – and uses the meta open source of Llam architecture, as well as Alibabin Equivalent Qwen.
“I think this absolutely provokes the idea of a strategy of licking that had many US companies,” said Ms. Bluquist.
“It indicates potential methods of developing models that are much less computer and intense resources that would potentially signal a shift in paradigm, although it is unconfirmed and remains to be seen.
“We’ll see what they bring for the next few months.”
