DeepSeek places China in the leading AI position
As I write this column the leading app on the AppStore is DeepSeek, a competitor to ChatGPT. DeepSeek is developed by a Chinese artificial intelligence lab, which develops open-source large language models. Over the weekend I played around with the DeepSeek app and I was surprised by the quality of its answers in comparison to other leading AI models.
One major difference between DeepSeek and other leading AI apps is that it’s not powered by a US-based company but by a Chinese based entity. It’s something that is surprising to those who assumed that all answers will come from Open AI and the Google AI giant, DeepMind. No one could have predicted that a major AI technology could come from China. How did this AI platform come into existence?
The answer to that question is important now when one considers the dispute between Elon Musk and Sam Altman about the formation of Open AI. For those who missed this important piece of technology history, in the OpenAI name lies the intention to build an open source entity to power the development of AI technologies for the benefit of all. In other words, OpenAI was not built to be a profit seeking entity. It was designed to rely on open research or wisdom of the crowds to benefit from researchers across the globe and not from a single entity.
However, Musk argues that OpenAI has betrayed the promise to create an open entity and instead its co-founder is building a closed and profit seeking entity. DeepSeek on the other hand has embraced everything that is open source. The entity has been forced by circumstances to take the open source route. The ban by the US designed to stall the development of AI in China has forced Chinese entities to be innovative. DeepSeek is a classic example of developing under difficult circumstances when resources are limited and conditions are less perfect.
The Chinese model has been developed by young people from Chinese universities backed by a hedge fund billionaire Liang Wenfeng. One of the key differences between DeepSeek and others is that it’s powered by research focused people. The entity is staffed with PhDs from top Chinese schools, Peking, Tsinghua and Beihang Universities rather than experts from US institutions. Beyond just relying on a corporate team their approach is to expand their workforce within the open source community.
They released a detailed paper on how to build a large language model on a bootstrapped budget that can automatically learn and improve itself without human supervision.
The DeepSeek story tells us that no country has monopoly on innovation. It also tells us that attempts to strangle progress will fuel innovation. It remains to be seen how DeepSeek will evolve beyond these early stages.
OpenAI also came with a major hype. It’s one thing to create a product and placing it on the market. Its adoption and market relevance over a period of time is what matters. For now we know that China has an AI answer which is a game changer in the AI race. Let’s hope DeepSeek inspires others to innovate. Africa has a lot to learn from this app’s approach. As you download this app, let's hope that we will not see other attempts to ban another Chinese creation as we’ve seen with TikTok and Huawei.
Wesley Diphoko is a technology analyst and Editor-In-Chief of FastCompany (SA) magazine.
BUSINESS REPORT