Recently, there are so many AI (artificial intelligence) breakthrough that it is almost impossible to follow all – at least without using AI. However, among all these breakthroughs, Deepseek’s R1 model OpenAI was as strong as O1, but it was managed to remain on the agenda for a long time with the development of a much cheaper development. The Chinese venture, which develops the model, claims to have spent only $ 5.6 million to train Deepseek, and this is far below $ 100 million that OpenAI spends to train the GPT-4.

Deepseek has created a shock effect in the artificial intelligence industry, but recently spoke to Google Deepmind CEO and founding partner, Speaking to Bloomberg at the Artificial Intelligence Action Summit in Paris. “It is a very impressive model, a very impressive work, and I think the team probably the best team from China, Hass said Hassabis.

Hassabis explained that he believes that the low cost of developers and the low cost to the headlines is only the cost of “the latest education work, and said that this is only a small part of the full cost of educating an AI model from start to finish. He also claims that Deepseek was “trusting Western models to improve, and OpenAI made the same suggestion in the days after Deepseek’s launch.

OpenAI told Bloomberg at that time, “We know that the People’s Republic of China -based companies and others are trying to constantly transform the models of leading US artificial intelligence companies,” he said.

Finally, Hassabis said that, although Deepseek is impressive, Google does not see this model as a “fantastic solution ındaki in terms of new technologies or techniques. “This is impressive, but not a new excessive value in the productivity curve. For example, Gemini is more efficient in terms of deepseek in terms of education or cost from performance or cost. “We are not just talking about it.”