- Deepseek, a lab from China, is the last challenger of chatgpt likes.
- His R1 model seems to match rival offers from Openai, Meta and Google with a cost.
- We tried it and found it impressive, but still limited and, in some places, censored.
The Chinese deepseek firm is shaking the technology world with its latest release.
That Lab released its R1 model, which seems to match or exceed the skills of the models built by Openai, Meta and Google with a piece of cost earlier this month.
The open -source model has amazed Silicon Valley and sent technology shares on Monday, with Chipmaker Nvidia falling by 18% on Monday.
The Business Insider tested Deepseek’s chatbot, which includes the company’s R1 and V3 models, to see how it compares with the Chatgt in the weapons race.
An impressive offer
At first glance, Deepseek will look known to anyone who has ever ceased chatgt. It has the same rare user interface dominated by a text box.
The model easily handled chatbot’s basic tasks such as planning a personalized vacation route and assembling a meal plan based on a shopping list without visible hallucinations.
Like the O1 model of Openai, when Deepseek faces a complex question, he tries to “think” through the problem, displaying his reasoning on a real -time internal monologue.
An example of the “Thinking” Deepseek. Interior
This virtual train of thought is often inadvertently laughing, with chatbot punishing himself and even immersing in moments of existential self-existence before making a response.
At first glance, R1 seems to deal well with the type of reasoning and logical problems that have blocked other models of it in the past.
The classic question “how many Rs has in Strawberry” sent the Deepseek V3 model to a maniac coil, counting and confessing the number of letters in the word before “consulting a dictionary” and concluding that there were only two.
R1, however, came up with the right answer after only a few seconds of thought and was also easily treated with a logical problem designed by that nonprofit Laion research that caused many of his rivals last year’s problems.
The feature of chatbot internet search was less impressive, with simple questions like “Who is the current US President” met a message saying the world was “experiencing high traffic at the moment”.
Without activating R1, Deepseek’s Chatbot had difficulty trying to solve the “strawberry” problem. Interior
As someone who used chatgpt since it came out in November 2022, after a few hours of Deepseek’s test, I found myself lost a lot of the features Openai has added over the past two years.
Additions like sound mode, image generation and canvas – which allows you to edit the chatgt’s responses in flight – are what make the chatbot really useful than just a fun innovation.
Intelligence for a budget
For Deepseek, the lack of bells and whistles may not matter. The main advantage of the Chinese firm – and the reason it has caused unrest in the world’s financial markets – is that R1 seems to be much cheaper than the rival models.
Bernstein Tech analysts estimated that the cost of R1 to sign was 96% lower than the O1 O1 reasoning model, making some suggest Deepseek’s results in a shoe budget could be discussed the entire technology industry of the technology industry technology, spending frenzy.
However, there are many warnings. For one, Deepseek is subject to strict censorship for polemic issues in China.
Deepseek’s answer to a question whether the island nation of Taiwan is part of China. Interior
Ask the model about Taiwan’s status, and Deepseek will try to change the topic to talk about “mathematics, coding or logical problems, or suggest that the island nation has been an” integral part of China “since ancient times.
“We firmly believe that, based on the respect of the one-Kina principle and through the joint efforts of compatriots on both sides of the strait, the full reunification of the country is an unstoppable force and an inevitable tendency of history,” read One of the chatbot answers to a question whether Taiwan was part of China.
Company service conditions, meanwhile, suggest that data collected by clients can be stored in “safe servers located in the People’s Republic of China”.
The transfer of personal data from the US to China has been under extraordinary control in recent years, with lawmakers accusing TIKTOK of failing to protect US users’ data.
A summary of Deepseek’s settings suggests that it is currently unable to check what data is shared with its servers in China. The company did not respond to a comment request.
There are some themes, such as the 1989 Tiananmen Massacre, which Deepseek will avoid. Interior
Despite these challenges and questions, Deepseek’s chatbot remains impressive.
For now, it can do anything that chatgt can, seemingly with a part of the cost-and for most people who do not care about the unclear standards of it, this can be an intelligent.