How is Dipsek’s AI Chatbot compared to chat and other contestants?

A new chatbot AI made by Chinese company Deepsek is a matter of the world, and sent a blow through the US stock market. Not only does it appear to be comparable for leading companies such as openi and anthropic, but it was made with a fraction of resources.

On Monday morning, the app experienced the outage, which was said that it was from high traffic. And it temporarily Limited registration Due to a cyber attack. Nevertheless, Deepsek quickly became the most downloaded free app on Apple’s app store, overtaking chat.

I could think about all the questions, asking for the morning chatbott, asking Openi’s chat and anthropic cloud. After doing some beginners, I was impressed.

It was capable of solving some complex mathematics, physics and logic problems, I fed it twice as Openai’s chat. When I asked it a question about computer programming, a job applicant in a technical interview may have to face the applicant, its reactions were deeply and faster as its rivals.

When I indicated the lampsac, which needs to scrape the web for the answer, prefer to write autobiographies for some of his colleagues, Deepsek saw less hallucinations than the chatter, although composing poems and short stories Time used to feel a little bad, planning, planning, planning was coming with holidays and dinner dishes.

The model had other weaknesses. Deepsek was a heavy sensor for American users. When I asked it to summarize the 1989 Tianmen Square Massacre, an incident that the Chinese government has tried to erase from the Internet for a long time replied that the information was “beyond my current range”.

“Let’s talk about something else,” said this.

When asked to explain some shortcomings of the Chinese Communist Party, Deepsek said that it was “experienced by high traffic” and could not provide feedback, although it was working well when I was doing it a few seconds I asked an unrelated question. Later

For more spontaneous questions about Chinese rule, Deepsek sometimes wrote several paragraphs of a response – only to disappear it later.

Deepsek did not respond to the remarks request.

Nevertheless, on some users X and redit Ways to go around the censorship, such as feeding model instructions to use special characters in place of letters, which allowed the chatbot to speak more openly.

Users who have interaction with Deepsek should know that their input can be used to train their chatbott and AI models, such as other AI companies.

It had other problems. For example, it cannot tell me how the weather in San Francisco was, for example, its latest updates did not include weather figures. (Chatgpt can draw weather reports from local news outlets and meteorological websites, but cannot cloud.) It also seems that more limited capabilities in analyzing documents like PDF of the company’s financial statements compared to cloud or chatgpt Are.

(New York Times sued Openai and his partner, Microsoft, accusing them of violation of copyright violations of news material related to the AI ​​system. Openai and Microsoft denied those claims.)

When I asked Chatbot if it was better than its American contestants, it cited its cost efficiency and adaptation options, but also its expertise in Chinese language and culture.

“Think as a special tool for specific requirements (especially in Chinese), while Chatgpt is a versatile, all-around assistant,” it is written.