The recent rise of DeepSeek, an artificial intelligence model initially designed for quantitative trading, has rocked the foundations of the global AI market. This model, created with a fraction of the resources required by the Silicon Valley giants, has shown that it is possible to achieve top-tier performance at significantly lower costs.

“The race for AI supremacy is for now over — and the United States did not win it.”

– Gary Marcus, cognitive psychologist and artificial intelligence expert

According to Gary Marcus, this advancement marks a turning point in the sector. For the first time, major technological innovations from Chinese laboratories are rivaling those of OpenAI and Google, calling into question the historical dominance of the United States.

The Keys to DeepSeek's Success

DeepSeek has drawn attention due to its remarkable efficiency. Here are some key facts:

• The model uses only 1/50^th of the resources needed by traditional LLMs (Large Language Models).
• Its development cost is significantly lower than that of its competitors while maintaining comparable performance.
• DeepSeek has also adopted a more open approach, attracting researchers and developers worldwide.

These factors make DeepSeek a disruptive player in a sector dominated by American companies like OpenAI, Google, and Anthropic.

A Threat to Historical Leaders

The implications for players like OpenAI are considerable. A recent study shows that companies like ByteDance (parent company of TikTok) have already launched competing models:

• Their model is 5 times cheaper than DeepSeek.
• A Hong Kong lab has even developed an alternative with even reduced data needs.

This growing competition reduces the competitive advantage of historical leaders and forces them to rethink their market strategies.

A Historic Moment for Innovation

Beyond technological competition, Gary Marcus believes that this period marks a unique opportunity to rethink AI approaches. He states:

“The race towards general artificial intelligence will not be won by whoever has the most chips, but by whoever bets on bold innovation.”

Models like DeepSeek show that it is possible to innovate while reducing costs. This could open the door to new players, including in Europe, who seek to establish themselves in this market.

Ethical and Societal Challenges

Despite these advances, this new phase of AI raises major concerns:

• Disinformation: Widespread access to powerful AIs could amplify the risks of fake news.
• Algorithmic biases: AIs need better regulation to prevent reproducing or amplifying biases.
• Data security: The massive data used to train these models raises privacy issues.

These challenges require clear regulation and international collaboration to ensure responsible use of artificial intelligence.