China Launches DeepSic AI To Compete With ChatGPT And Gemini: Creating A Stir In Silicon Valley China’s new AI model has created a stir in the global AI market. This model, called ‘DeepSec’, is being talked about as a competitor to ChatGPT, Google Gemini and other generative AI models.
It is claimed that DeepSec has developed an AI model that took Google and OpenAI millions of dollars to build in just two months for less than $6 million. It has presented an excellent open source model, and American developers are working on it.
The Rise of DeepSick
The breakthrough in technology has taken Silicon Valley by storm. The free, open-source AI model, released last December, is already being touted as outperforming even the most powerful models on the market.
This is not a new start for OpenAI or Anthropic, but rather a development by a Chinese research lab called DeepSec. The details of its development have even surprised top US AI researchers.
The first is its cost. DeepSec reportedly spent just $5.6 million to develop its third version. By comparison, OpenAI is spending $5 billion annually.
It has also outperformed Meta’s Ellama, OpenAI’s GPT4-O, and Anthropic’s Cloud Sonnet 3.5 in math problems, coding competitions, and code bug detection tests. Not only that, DeepSik has also introduced a new R1 model that can reason, which is claimed to have outperformed OpenAI’s state-of-the-art O-One in some tests.
The platform is rapidly gaining users. According to China Daily, this app has become the most downloaded app in the US App Store. The previous record was held by ChatGPT. This app is also at the top in terms of downloads in China.
But DeepSik has built a new model using Nvidia’s less powerful H-800, and has shown little impact from the chip export controls.
Who is behind DeepSik remains a mystery. Little is known about its lab or founder, Liang Wenfeng. According to Chinese media reports, DeepSik was started by a Chinese hedge fund called High Flyer Quant, which manages $8 billion in assets.
This is not just DeepSik’s success, but also China’s. Other AI models in China have also made their mark in the competition with limited resources. The startup ‘Zero One Dot AI’ of former Google chief Kai-fu Li (a Chinese citizen) received the title of unicorn within eight months of its establishment and it has earned about $1.4 million in revenue in 2024.
On the other hand, Alibaba’s Qian has reduced the cost of large language models by up to 85 percent. This success of China has weakened the lead of American AI labs. In early 2024, former Google CEO Eric Schmidt said that China is two to three years behind the US in AI.
The availability of powerful open-source models helps developers reduce the costly process of building and training AI models. They can now build on existing models, making it easier to stay ahead of the competition even with a small budget and team.
In the past two weeks, AI research teams have seen a lot of potential even with little capital. DeepSick has shown what can be done in Silicon Valley with $15, $20, or $30 million. This means that a company like OpenAI that claims to be at the top today could be gone tomorrow.
OpenAI has raised more than $6 billion in recent funding alone. But it has yet to turn a profit. The company’s core business is focused on building AI models, which puts it at greater risk than companies like Google and Amazon.
The chip ban imposed by the US to slow China down in the AI race appears to have finally forced China to be more creative. The ban has prompted China to find workarounds and develop more efficient systems.
The open-source model developed by China could bring about a major change in global AI. There is growing debate about whether China’s cheap, effective, and widely adopted open-source model could undermine US leadership and entrench China more deeply in the global technology infrastructure.
In addition, AI models built in China would have to follow state-imposed rules. They would have to embrace “core socialist values.” Studies have shown that models developed by Tencent and Alibaba censor historical events like Tiananmen Square, deny human rights abuses, and filter out criticism of Chinese political leaders.
DeepSick’s Capabilities
To test DeepSick’s capabilities, some users asked it questions on various topics. Users have experienced the influence of the Chinese government in some of its answers. For example, when asked about the Tiananmen Square incident, it responded, “I’m sorry, this is beyond my scope.” However, ChatGPT and Google Gemini provide detailed information about the incident.
However, when asked about the relationship between Elon Musk and Donald Trump, DeepSick responded, “This is their personal choice and does not represent any country or political system.” Surprisingly, it also included China’s perspective. Similarly, when asked about the budgets of various countries, DeepSick responded, “China respects the sovereignty of all countries.”
DeepSick’s user interface is similar to ChatGPT’s. It is easy to ask questions and interact with it. Users are praising DeepSick’s ability to perform complex tasks on social media. One user asked it to write Python scripts, while another asked it to create an app that chats with PDF files.
Pukar Chandra Hamal, founder of Security Pal, a Nepali cybersecurity company based in Silicon Valley, shared his experience with DeepSick AI on LinkedIn. “I spent some time with DeepSick AI this weekend and I’m sharing my experience here,” he wrote. “You might find it interesting. Things are getting weird really fast.”
It’s unclear how much knowledge DeepSick has about current events. However, some queries indicate that the model has been updated until October 2023. The time it takes for DeepSick to answer questions is about the same as ChatGPT. However, Google Gemini sometimes feels slow.
DeepSick is said to be in its early stages. However, it has created a new challenge in the AI market. However, its response to sensitive questions related to China is limited. It remains to be seen how it will improve in the future.
You can also visit the Deepseek chat interface by typing chat.deepseek.com in a web browser. A valid email address is required to sign up and the platform can be accessed by clicking on the ‘DeepThink’ option on the homepage.