DeepSeek versus ChatGPT: How should Europe respond?
A new chatbot from China, DeepSeek, has surprised the AI world. The chatbot has similar capabilities compared to OpenAI’s latest models, but requires only a fraction of the cost to train it: 6 million dollars as compared to 100. It’s also cheaper to operate.
The US stock market reacted fast; the AI sector took a serious hit. Nvidia shares fell sharply when the market opened on Monday. It lost nearly 600 billion dollar in market value. Altman was impressed with the technology.
Later the enthusiasm cooled down: OpenAI claimed that DeepSeek stole its technology. DeepSeek used so-called “distillation” — training an AI on the output of another. A smaller model learns from a larger one. Altman’s complain, however, led to a lot of mockery, since OpenAI itself has been accused of stealing from creatives. Critics had a good laugh. Altman is not amused: did OpenAI lose its edge forever? As Futurism put it: “AI Companies Tremble”.
What does all this mean for Europe?
Europe lags behind in innovation and investment, when it comes to AI and in general. Last year the Draghi report already warned that the EU does not keep pace with its rivals the U.S. and China. We were mainly worried about the U.S., but now the gap with China also seems to widen. Embarrassing. Worrying.
The competition narrative is influential. Are we going to be the losers of the AI race? Or an AI war? We live in times of trade wars, perhaps a cold war. The metaphor is powerful. Is Europe on its way down? Is it already defeated? Is it even relevant?
But there is also good news. The DeepSeek shock is also a chance. DeepSeek shows that you don’t need enormous investments. Smaller players, also in Europe, can join. Less investment is needed than we thought.
And perhaps more importantly when it comes to AI ethics: the case also shows that it’s possible to train and run such models in a way that is not only more economical but also with a lot less environmental impact. This may not be of interest to the current U.S. administration, or to Altman for that matter, who is still on a mission to deliver more compute and doesn’t really care about cost efficiency, but it certainly resonates in Europe.
Instead of letting itself be drawn into an AI arms race between the U.S. and China, then, Europe should aim to become a global leader in ethical and responsible AI. That includes environmentally friendly AI that has a smaller footprint in terms of electricity consumption and carbon emissions. Regulation such as the AI act can help. But proactive, bottom-up initiatives, creative innovation, and genuine ethical entrepreneurship are very much needed.
They are better because the start from a vision that makes sense: the belief that ethical and environmental AI are good for the economy and good for the people and planet. An increasing number of AI people in Europe share that vision. They already knew that it is desirable to develop AI that is climate friendly and better for the environment. One thing we can learn from the DeepSeek case is that it is also possible.
