The recent launch of DeepSeek, a Chinese AI-chatbot app, has stirred up disruption in the US markets and spurred discussions on the future of America's AI dominance. The BBC has delved into the workings of the app.
DeepSeek functions like a typical chatbot but tends to be excessively chatty in its interactions.
Similar to OpenAI's ChatGPT or Google's Gemini, users can ask DeepSeek questions on various topics upon opening the app or website, receiving detailed responses without expressing personal opinions.
The chatbot often deems topics as "highly subjective," whether discussing politics or preferable soft drinks like Pepsi or Coke.
Despite avoiding a direct comparison to OpenAI's ChatGPT, DeepSeek assesses the strengths and weaknesses of both in a similar manner, using comparable language.
Trained on data up to October 2023, DeepSeek seemingly accesses up-to-date information, though the website version may lack this feature, akin to past versions of ChatGPT.
DeepSeek's reluctance to discuss sensitive topics forbidden in China diverges from its US counterpart, as it self-censors responses to questions on contentious subjects like the 1989 Tiananmen Square protests.
While DeepSeek may respond quickly, it currently faces delays due to a surge in users exploring the app during its viral wave.
Unlike US-developed chatbots, DeepSeek refrains from addressing banned topics in China, redirecting conversations when such themes arise.
A significant point of variance between DeepSeek and its US competitors is its avoidance of controversial subjects censored in China, contrasting with the more candid responses from US-developed AI like ChatGPT concerning events like Tiananmen Square.
Kayla Blomquist of the Oxford China Policy Lab noted a seeming hands-off approach from the Chinese government towards DeepSeek amidst increased investment in the sector.
Despite sharing similarities with established US AI assistants, DeepSeek is subject to the same scrutiny regarding accuracy and functionality.
By showcasing possibly more cost-effective model development, DeepSeek introduces an alternative to the resource-intensive methods employed by leading US AI firms, hinting at a potential paradigm shift yet to be fully confirmed.