José Carlos Feliciano Nishikawa
Associate Director of the Center for China and Asia-Pacific Studies
Universidad del Pacífico (Lima, Peru)
What are the main technological differences between DeepSeek and ChatGPT?
In general terms, DeepSeek is comparable to the latest models developed by OpenAI for ChatGPT. DeepSeek R1 achieves results similar to OpenAI o1 and OpenAI o3-mini in benchmark tests for Large Language Models (LLMs) and even surpasses them in certain categories such as solving complex mathematical problems. The main difference lies in the cost of these algorithms. DeepSeek has indicated that the cost of developing the R1 model reached 6 million dollars, in contrast with the more than 100 million dollars invested by OpenAI in the development of ChatGPT-4. The launch of DeepSeek and its cost also drew significant attention after Trump announced the Stargate project, in which 5 billion dollars would be invested in AI infrastructure development in the United States.
How does the competition between DeepSeek and ChatGPT impact the global digital economy?
The positioning of DeepSeek as a market competitor has impacted the economy in two main ways. On the one hand, it has cast doubt on the claims of leading AI companies that assert the need for billions of dollars in investment to continue developing LLMs. DeepSeek, with its lower development cost, has led investors worldwide to question the position of these companies in the AI market. This was reflected on February 27, when many leading technology companies registered record losses in the stock market. On the other hand, the fact that DeepSeek R1 is openly accessible has pushed companies such as OpenAI to rethink their business models. ChatGPT has dominated the market thanks to more advanced LLMs. If it loses this advantage, its position in the industry could be severely affected.
What ethical challenges arise in the development of AI in geopolitical contexts?
The development of AI always entails ethical challenges. This is because training models requires vast amounts of data extracted from the internet, which may include the work of other people, such as books, articles, social media posts, and even works of art. This constitutes a legal grey area, since it is argued that these individuals did not explicitly forbid the use of their data. Fortunately, some legislations has already begun to address this issue.
In the context of the geopolitical rivalry between China and the United States, DeepSeek faces two main concerns. First, it has been accused of distillation, that is, training its model based on larger ones, such as ChatGPT. This is not new, as several Chinese AI companies have been suspected of allegedly using U.S. models to develop their own. Furthermore, concerns have been raised regarding censorship in DeepSeek’s responses about sensitive topics related to China, something that can indeed be easily verified. However, this geopolitical rivalry could also strengthen the U.S. monopoly in AI by presenting DeepSeek as a national security threat, as was the case with TikTok.
What role do international academic collaborations play in the evolution of technologies such as DeepSeek and ChatGPT?
Academia has played a more passive role in the evolution of AI, although its sources, such as books and articles, have been fundamental for training many models. However, it has already begun to analyze both the risks that AI represents for academic work and its benefits for research and teaching. Through international collaboration, academia has the opportunity to influence the development of these technologies by sharing knowledge, approaches and standards, providing access to specialized databases, and peer review.
What does it mean that the University of California has implemented an educational version of ChatGPT in its academic community?
On the one hand, this agreement reflects a commitment to the role of AI in academia. While it is easy to imagine its potential negative impacts on learning and knowledge production, it is still unclear how or if it will affect academic services. On the other hand, it reinforces the position of AI companies in the technology services market. This agreement between OpenAI and the University of California could set a precedent for companies such as DeepSeek to seek offering their products to universities and research centers.
