A new Chinese AI chatbot, DeepSeek, has taken the tech world by storm since its launch last week, raising eyebrows not only for its capabilities but also for the potential shift it signals in the global AI landscape. The app has already triggered turbulence in US markets, particularly for major players like Nvidia, and has sparked questions about the future of American dominance in AI technology.
On the surface, DeepSeek appears much like other popular AI assistants such as OpenAI’s ChatGPT or Google’s Gemini. It allows users to ask questions on virtually any topic, and it responds with detailed answers, often engaging in friendly conversation. However, DeepSeek quickly differentiates itself from its Western counterparts in several ways.
A Chatty AI With Boundaries
Similar to ChatGPT, DeepSeek delivers long-form responses and avoids offering personal opinions. If asked about subjective topics like politics or preferences, it often starts by acknowledging that the subject is “highly subjective” before proceeding with a balanced answer. For instance, when questioned about the political legacy of Donald Trump or the merits of Pepsi versus Coke, the chatbot hesitates to make any definitive statements, offering a neutral overview instead.
DeepSeek also refrains from stating whether it is superior to ChatGPT, instead comparing both models by listing their respective strengths and weaknesses. This cautious approach is reminiscent of how ChatGPT has addressed similar comparisons.
The chatbot was trained on data up until October 2023, which aligns with the practices of early versions of ChatGPT, designed to mitigate the risk of disseminating real-time misinformation. Despite its limitations in terms of up-to-date knowledge, DeepSeek can still access current information like the day’s date, although its website version lacks this capability.
Censorship: A Key Difference from Western Competitors
However, DeepSeek is not like its US counterparts when it comes to sensitive topics, particularly those that are politically sensitive in China. One of the chatbot’s most controversial features is its self-censorship. If asked about topics such as the 1989 Tiananmen Square protests—one of the most significant and tragic events in modern Chinese history—the chatbot simply refuses to respond. Instead, it redirects the conversation by saying, “Let’s talk about something else.”
This self-censorship marks a stark contrast with ChatGPT, which does not shy away from discussing the Tiananmen Square protests and other politically sensitive issues. In fact, ChatGPT’s unfiltered responses on such topics highlight one of the key differences between AI tools in China and the West, where political sensitivity and government oversight remain crucial factors in AI development.
As DeepSeek continues to gain traction, the extent of government involvement in its development has raised questions. While the Chinese government has largely refrained from directly interfering with the chatbot, recent reports suggest that it may soon ramp up its investment in the project, potentially leading to stricter controls and further regulation. Kayla Blomquist, a researcher at the Oxford Internet Institute, speculates that this shift could signal an increased alignment between DeepSeek and state priorities, which may affect its development and functionality.
Cost-Effective AI: A Challenge to US Models?
While DeepSeek shares some similarities with its Western counterparts, it also stands out in terms of its cost-effectiveness. Despite using older Nvidia chips and Meta’s Llama architecture, alongside Alibaba’s Qwen model, the chatbot seems to have been developed with an emphasis on reducing resource consumption, making it more affordable to run.
Ms Blomquist pointed out that DeepSeek challenges the monetisation strategies of leading US AI companies, which have often relied on expensive computational resources. If DeepSeek proves to be as capable as its rivals, despite its cost-saving approach, it could potentially change the dynamics of the AI market. She speculates that the Chinese model might point the way towards more efficient AI development methods that are less resource-intensive, but she acknowledges that this remains to be fully confirmed.
“We could be witnessing the emergence of a new paradigm in AI model development, one that is not as reliant on vast amounts of compute and resources,” Blomquist said. “How this plays out over the coming months will be crucial.”
The Road Ahead for DeepSeek
As it stands, DeepSeek is rapidly gaining popularity, with many users testing its capabilities. The app has captured attention due to its polished interface and the striking contrast between its innovation and the increasing scrutiny faced by its Western counterparts. Its impact on global AI development remains to be seen, but for now, it is clear that DeepSeek has the potential to shake up the AI sector, especially when it comes to the political and economic implications of China’s growing influence in technology.
Whether DeepSeek will continue to challenge AI giants like OpenAI and Google depends on how effectively it can balance its cost efficiency, censorship policies, and technological advancements. The coming months will likely determine whether this new AI tool can truly rival its Western competitors or simply remain a niche offering in the broader AI ecosystem.
