Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Technological chiefs agree to a large extent that OpenAi’s risk proposals are limited for now.
Bloomberg | Bloomberg | Getty images
The technological advances that the Chinese artificial intelligence laboratory has shown Show The Game is in time in the American-Sino competition in AI, the main technology executives said to CNBC.
In a series of interviews at the France Artificial Intelligence Action Summit, the leaders of several important technological companies told CNBC that the appearance of Deepseek demonstrates that China cannot be counted as a serious player when it comes to the innovation of innovation of Ia.
Last month, Depseek surprised global markets with a technical document that says that one of its new AI models was created with a total training cost of less than $ 6 million, much less than the billions of dollars they spent The great technological players and the Western AI. Laboratories such as OpenAi and Anthropic.
Chris Lehane, director of Global Affairs of Openai, told CNBC that the low -cost advanced Deepseek model confirms that there is a “very real competition between democratic AI and the democratic PCCH (Chinese Communist Party) led by the United States. AI.”
Many Deepseek critics have indicated the apparent censorship of the model when it comes to sensitive issues. For example, when asked about the Tiananmen Square massacre of 1989, the assistant application of the Deepseek responds with: “I’m sorry, that is beyond my current reach. Let’s talk about something else.”
“There are two countries in the world that can build this on scale,” Lehane told Arjun Kharpal from CNBC out of the AI Summit in Paris on Monday. “Imagine if there were only two countries in the world that could build electricity on a scale. This is how you have to think about it.”
“For us, what Deepseek really reinforces and reaffirms is that there is very real competition with very real bets,” Lehane added.
Even so, the technological chiefs agreed to a large extent that despite the fact that Deepseek’s progress shows that China is further in the global career of what was previously thought, the threat it represents for Openai remains limited for now.
‘The game is underway’
Depseek says that its new R1 model, a open source reasoning modelHe was able to rival the performance of the similar O1 model of OpenAI, only using a cheaper and less energy intensive process.
That led experts to question the predominant wisdom in the west of recent years, which is that China is behind the United States in the development of artificial intelligence due to export restrictions that make it difficult for companies in the country to have your hands in the most advanced NVIDIA graphics or GPU units.
GPUs are necessary to train and execute AI applications because they stand out in parallel processing, which means that they can perform multiple calculations simultaneously.
Reid Hoffman, co -founder of LinkedIn and partner of the Risk Capital firm Graylock Partners, told CNBC on Monday that Deepseek’s new model is “a big problem by demonstrating that the game is underway.”
“The competition is underway with China,” said Hoffman, added that Deepseek’s R1 is “a credible and processable model.”
Abishur Prakash, founder of the strategic advice firm The Geopolitical Business, told CNBC that Deepseek shows that China’s understanding of the West remains limited.
“The assumed place in the United States as technological captain of the world is no longer the acceptable belief,” Prakash told CNBC in a telephone interview.
“That is the new status quo now, that the space between the United States and China has been reduced almost during the night, but it has not been reduced overnight, years of progress have passed,” Prakash said.
“If there is a conclusion for the West, it is that their understanding of China is incredibly limited, and we do not know what will come later,” he added.
There is no significant threat to us, AI, still
Even so, the main executives of AIs are not convinced that Depseek presents any significant risks for the business of AI laboratories such as OpenAi and Anthrope yet.
While experts in all agree, the advances of the Deepseek have been impressive, doubts have been raised about the startup claims about the cost.
A report by the Semianalysis semiconductor research firm last month estimated that Deepseek hardware spending is “very high” than $ 500 million about the company’s history. Depseek was not immediately available to comment when CNBC contacted.
The report found that the Research and Development costs of Deepseek related to the property are significant and that generating “synthetic data” so that the model will be able to require “a considerable amount of computation.”
Some technologists believe that Depseek could have achieved such a high level of performance by training their models in the US systems. UU. Larger.
This technique, known as “distillation”, implies that the most powerful AI models evaluate the quality of the responses generated by a newer model.
It is a statement to which OpenAi herself has alluded, telling CNBC in a statement last month that is reviewing the reports that Deepseek may have “inappropriately” output “output data of its models to develop its AI model, a known method as “distillation.”
“Most of the fear market (Deepseek) is out of place,” Hoffman told CNBC. “It still requires large models, it was distilled from large models.”
“I think the short response that everyone should take is: game, but large models still really matter,” he added.
Victor Riparebelli, CEO of the AI video platform, Synthesia, told CNBC that, although Depseek challenged the “paradigm that the gross force scale is the only way to build better and better models”, the idea that Companies will suddenly change significant amounts of their AI workloads are wrong.
“I still think that when you look at the users of these technologies, all workflows, I think that when we look back in three months, I think that 0.01% of them will move to Deep Speedai and Anthrope,” Riparebelli said.
Meredith Whitaker, president of the Signal Foundation, said that Deepseek’s development does not move the needle for the industry much, since the market impulse is still widely in favor of the largest models. Signal Foundation is a non -profit organization that admits the sign of the encrypted messaging application.
“This is not something that will interrupt the concentration of power or geopolitical balance at this stage,” Whitaker told CNBC. “I think we have to monitor the ball there and recognize that it is really this bigger ‘paradigm is better’ that is not reduced through efficiency profits historically, which is promoting this concentration.”
