From ChatGPT to DeepSeek: What About Europe?

China surprised the world with the introduction of DeepSeek. Amid this news, US technology markets collapsed, with President Trump calling it a wake-up call. But what does history and economic analysis tell us about this news? And should Europeans be worried or not?

AI, a bit of history: in the early 1990s in the USA, people wondered whether artificial intelligence would turn out to be really useful or become a copy of chaos theory (a butterfly flapping its wings in Beijing changes something in New York, but nobody knows what), mathematically fascinating, but devoid of practical content (at the time). I can say this with confidence because this is the kind of comment I received when I presented a research paper at the National Bureau of Economic Research in Boston, in which, together with Pietro Terna and Sergio Margarita of the University of Turin, I discussed how to use artificial neural networks (ANNs) to model a financial market populated by rational agents who make investment decisions based on a reasoning scheme characteristic of ANNs. It was an example of a research program, summarized in a 1995 book (Neural Networks for Economic and Financial Modelling, Kluwer University Press), the thirtieth anniversary of which is now being celebrated. To give another example, I submitted the same paper to the finance department at Columbia University in New York and received extremely negative feedback, which led me to abandon AI as a major area of research. In thirty years, Americans and the rest of the world have gone from complete skepticism to faith in this computational tool, turning it over the past fifteen years into a tool of competitiveness, cited in the Draghi report as a major factor in the per capita income gap that has accumulated between Europe and the USA over that period.

AI, productivity and income: does AI drive growth? Studies say yes. Manufacturing companies that use predictive analytics (essentially a combination of artificial intelligence and big data) are more efficient. However, this phenomenon was relatively limited until the proliferation of GenAI (generative artificial intelligence), aka ChatGPT, which was to the AI sector what Microsoft Windows was to the use of computers. When Windows 1.0 appeared in 1985, the world of computers was reserved for programmers, the chosen few who communicated with machines using Fortran. Since 1985, it became possible to process data without knowing how to program. Of course, it took decades for Windows to become widespread and computers to be used professionally – to the point where Nobel Prize winner Bob Solow wrote that the impact of computers in the USA can be seen everywhere except for labor productivity statistics – but the cumulative effect is undeniable. We are at the dawn of a similar phenomenon in the field of artificial intelligence. Today, anyone can use ChatGPT without knowing how an artificial neural network works, just as anyone can drive a car without knowing how an engine works (of course, theoretically we should know this since it’s part of the driver’s test preparation course, but we know that no concept is learned until one is convinced of its actual usefulness). In the service industry, experiments conducted by large consulting firms show that people’s productivity increases by up to 30% after they start using ChatGPT. For now, the impact is still limited, but it will increase significantly in the future, according to Nasdaq, which evaluates the companies most affected by AI based on future earnings over the next decades.

ChatGPT and Deep Seek. The news of the spread of the “Chinese ChatGPT,” dubbed DeepSeek, caused a stir. Nvidia shares lost 16% in one day, recovering only eight percent in the next session. So why did the news that a Chinese scientist managed to create a tool comparable to ChatGPT become such a sensation? After all, we’ve known for years that China is investing as much, if not more, in artificial intelligence than the USA, so why be surprised? In addition, we know that in terms of attention to user privacy, China is less advanced than the USA and certainly less advanced than Europe (as the famous saying goes, in the US they invent, in China they adapt, in Europe they regulate), to which we will return shortly. Feeling more freedom to respect users and technology applications can only increase the efficiency of product development and the power of the algorithm. Perhaps there are two surprising points. Firstly, a product like ChatGPT can be created at a much lower cost, although on this point we have no certainty about the costs incurred and the time it took for this software, also considering that (a) it is easier to imitate than to create, (b) the article that “attention is all you need,” the foundation of GenAI using transformer technology, was published worldwide (hence in China) in 2017. Secondly, again, the Chinese algorithm seems to be much more efficient in terms of chip utilization and the energy needed to power it. This is an important point: in July 2024, Google wrote in its non-financial report that heavy use of data centers has led to a 50% increase in electricity consumption over the past five years, casting doubt on its 2030 goals (which after Trump may still be a dream). Many are therefore asking questions about the relationship between sustainable development and AI, and any news suggesting a softening of this trade-off is very positive (and compatible with a rise, not a fall, in tech company share prices).

China, USA, and Europe. What are the implications for future growth? Again, unfortunately for us, nothing new. It was inevitable that the Chinese would temporarily overtake the Americans in the technology race, which would be good for American companies, giving them a chance to rethink their business models and learn from the Chinese (Federer has always said that he wouldn’t have been so successful without Nadal, and Sinner is his own worst enemy, as he will be stronger in 2030 if he finds a real contender on the tour soon). Great news for global productivity. If it is confirmed that energy and component consumption has been reduced, the news will be even better (in the movie Back to the Future, the time machine at one point runs on a banana peel retrieved from a trash bag). The usual bewilderment remains: an authoritarian regime that has the least respect for individual rights has an advantage over those that respect them, and this can have negative consequences if it gains more competitive power. But there’s nothing new here either. And unfortunately, this is also no news for Europe: we have great goals (saving the world), great universities with cutting-edge research products, great talent and potential, but for some reason we can’t turn all this into companies or rather sustainable businesses. In short, Europeans should be much more concerned about DeepSeek than Americans, even if the main concern at this point will be the regular DeepSeek (which is actually not available in our Apple Store in the days following the news).