Deepseek sparks global AI collaboration

DeepSeek was a name that, until recently, was unknown even to the most avid followers of AI technology. It has rapidly become a household name, shocking the world with its powerful AI chatbot, an unexpected achievement from Asia as the world continues to adopt this revolutionary technology.

Widely used across China as a virtual assistant and source of quick information, DeepSeek has become an overnight sensation, with young people increasingly turning to the chatbot for therapy and mental health support, creating a new cultural phenomenon.

What has sparked global interest in particular is that it has been produced for just a fraction of the cost of its Western rivals, such as OpenAI's ChatGPT.

The surprising popularity of DeepSeek has disrupted international markets, causing shares of chip company Nvidia to plummet dramatically and triggering a wakeup call for Western AI firms.

DeepSeek founder Liang Wenfeng, like many other Big Tech CEOs, has an academic background from a top university. He graduated in computer science and electronic information engineering from Zhejiang University and also has experience in finance.

In the financial sector, the tech tycoon owns High-Flyer, a hedge fund that uses AI to make investment decisions based on historical data.

Liang's background has positioned DeepSeek to compete effectively with US rivals like ChatGPT, disrupting stock markets and shaking confidence among industry leaders.

The upheaval stems partly from the industry consensus that building such a platform required hundreds of billions of dollars, with OpenAI currently valued at over $150 billion.

DeepSeek was created at just a fraction of the cost, casting doubt on the perceived value of high-performance chips needed for such technology. Chip makers like Nvidia saw their share prices tumble as a result.

DeepSeek's cost-effective power comes from an efficient technique called Reinforcement Learning, or RL, a machine learning method that allows AI to learn and make decisions through trial and error to achieve optimal outcomes.

This natural, energy-efficient method mirrors human learning, where interaction with the environment leads to observed results. When decisions yield desired outcomes, those pathways are reinforced, while unsuccessful choices are discarded.

The concept of RL can rely on delayed gratification models, where the best strategies involve short-term sacrifices to reach a final goal. Ensuring that learning programs have this long-term oversight, rather than being short-circuited by seemingly simple quick steps, is an achievement in itself through the coding of learning models.

Despite DeepSeek's initial success, it is important that a country's self-reliance does not lead to technological isolationism. Sam Altman, CEO of US-based OpenAI, stated in a recent interview that he would like to work with China. Progressing such an important technology together is not only beneficial for both parties but also ensures that regulation of AI's power can be carried out with universal standards.

Open-source community learning could also help other nations advance their AI development. India has fallen behind China and the US in the AI race, with problems ranging from lack of capital, with a mere $1 billion compared to $500 billion in the US Stargate project, to a diverse range of languages that all need high-quality data sets.

Given AI's potential to revolutionize areas such as energy and medicine, a future of limitless fuel sources and curable cancers may be much closer if other nations are given the tools to help develop the AI we need to make this become a reality.