The good and the bad in America-China AI war

On January 27, a Chinese startup announced a free artificial intelligence (AI) chatbot model called Deepseek R1. It was said to be as good as, if not better than, ChatGPT in math, coding and natural language reasoning. Deepseek R1 has joined the competition against larger, billionaire-backed AI companies, including OpenAI, Microsoft, Google and Anthropic.

The biggest surprise about Deepseek is that the developers reported that it was developed for under $6 million in under two months. This overturns the belief that AI development is immensely expensive, justifying the billions of dollars invested in cutting-edge computer chips, data centres and energy, which caused demand and increased stock value for AI companies. For example, Meta had planned to use $65 billion on AI this year.

This achievement has sparked both awe and concern, as the Chinese firm accomplished it without access to the latest computer chips and semiconductors due to restrictions imposed by the United States of America. The Deepseek developers had to find a way to work around this limitation by using older chips while maintaining higher efficiency.

Trillion dollars

Also, Deepseek is open-source software publicly available for download, use, modification and distribution. This allows others to utilise the model, potentially making it a global standard. Since the announcement, Nvidia and other tech companies have lost a trillion dollars in the stock market, with Nvidia responsible for half of that. Meanwhile, Deepseek has overtaken ChatGPT as the most downloaded app on both Apple and Android platforms.

Deepseek has also solidified China’s place as a force to reckon with in the AI race, shattering the USA’s perceived hegemony in AI development. If China is currently number two to the USA, as many believe, it is shrinking the gap faster than anyone imagined. This might usher in an exciting China-USA AI race akin to the nuclear arms race, space race or the aviation race during the World Wars and the Cold War likely to spur innovation.

When OpenAI released ChatGPT in November 2023, it was a watershed moment in technology that could potentially change how human beings conduct various activities, including education. Today, many understand that AI models have been trained on billions of texts on the Internet and programmed to learn which words are most likely to follow other words in a sentence about any subject, enabling them to produce life-like results.

So far, these systems have been known to provide information not based on facts, merely reflecting back to us stuff that is already on the Internet. Bots do not know the difference between a scientific publication, poem or a short story, and they can draw inspiration from any content, thus making it bad for academic research, for example.

AI chatbots

In the long-term, there are concerns that these AI chatbots will diminish our ability to learn, research and employ critical thought. Unlike the open web searches we are accustomed to, which provide us with the results we query, leaving it to us to choose what to read and believe.

With the USA-China AI race underway, the safety and responsible development, design and deployment of AI may be put on the back burner. For example, Google has reneged on an ethical pledge not to develop AI for weapons and surveillance. Another concern is that this technology is controlled by a small group of billionaires who may not prioritise the world’s best interests. As such, many argue that it is time to set up rules regarding AI, whereas critics say that regulation may stymie innovation.

Nevertheless, Kenya must invest in cheap and reliable electricity, computer literacy, and digital public infrastructure to benefit from the AI revolution.

Mr Kiprono is ICJ Kenya deputy executive director and chairperson of Media Complaints Commission. [email protected].