Technology heads largely agreed that the risks of Deepseek on Openai are still limited at the present time.
Bloomberg Bloomberg Gety pictures
CNPC's senior officials in the field of technology said that the technological progress offered by the Chinese artificial intelligence laboratory Deepseek shows that the game continues when it comes to competing the United States in artificial intelligence.
In a series of interviews at the summit of artificial intelligence in France, the leaders of many major technology companies told CNBC that the appearance of Debsik indicates that China cannot be considered a dangerous player when it comes to creating artificial intelligence.
Last month, Dibsic shocked global markets with a technical paper saying that one of the new artificial intelligence models was created at a total training cost less than $ 6 million – much lower than billions of dollars spent by adult technology and Western prosecution laboratories such as Openai and Anthropic.
Chris Lean, chief international affairs official at Openai, told CNBC that the advanced and low -cost Deepseek model confirms that there is “a very real competition between AI a small democracy led by the United States and CCP (Chinese Communist Party), China, and tyranny Amnesty International.”
Many Deepseek critical referred to clear control by the model when it comes to sensitive topics. For example, when asked about the 1989 Tiananmen Square massacre, the AI Deepseek Assistant application responds: “Sorry, this is outside the current range. Let's talk about something else.”
“There are two countries in the world that can build this widely,” Lien said at CNBC Arjun Kharpal on the sidelines of the Paris AI summit on Monday. “Imagine if there were only countries in the world that can build electricity widely. This is a kind of way you should think about it.”
“For us, what really enhances Deepseek and reaffirms is that there is this very real competition with very real shares,” Lehan added.
However, technology heads have agreed to a large extent that although the penetration of Deepseek shows that China is more than them in the global artificial intelligence race than previously thought, the threat posed by Openai remains limited at the present time.
“The game on”
Deepseek says that the new R1, Open source thinking formHe was able to compete with the performance of the OPenai-similar O1 model only using a cheaper and less powerful process.
This prompted experts to question the prevailing wisdom in the West from the past few years, which is that China is behind the development of the United States due to export restrictions that make it difficult for companies in the country to put their hands on the most advanced NVIDIA graphics, or graphics processing units.
Graphics processing units are necessary to train and operate artificial intelligence applications because they excel in parallel treatment, which means that they can implement multiple accounts simultaneously.
Red Hoffman, co -founder of LinkedIn and his partner at Greylock Partners, CNBC Monday, stated that the new Dempsic model is “a big deal to show that the game is continuing.”
“The competition is in full swing with China, adding that the R1 Deepseek is” a reliable and implemented model. “
CNBC, the founder of the Geocolitical Business, told CNBC that Dibsic shows that the West's understanding of China is still limited.
“The supposed place in America is no longer as the technological leader of the world is no longer acceptable.”
“This is the current current situation now, as the area between the United States and China almost narrowed – but it did not narrow overnight, years of progress has passed,” said Prakash.
“If there are one ready -made meals for the West, then this is that their understanding of China is incredibly limited – and we do not know what will happen after that.”
There is no useful threat to us, Amnesty International – so far
However, Amnesty International Executive Command is not convinced that Dibsic raises any kind of significant risk to artificial intelligence laboratories such as Openai and Anthropor yet.
While experts in general agree on Deepseek from artificial intelligence, it was impressive, doubts about the startups about the cost.
A report issued by the semiconductor research company was estimated last month that the expenses of devices in Depsic are “higher” than 500 million dollars on the company's history. Deepseek was not immediately available for comment when contacting it by CNBC.
The report found that the costs and costs of research and development in Dibsic related to ownership are important and that “artificial data” of the model that must be trained requires “a large amount of account.”
Some technicians believe that Deepseek may be able to achieve such a high level of performance by training their models on the largest American AI systems.
This technique, known as “distillation”, includes the presence of more powerful models of artificial intelligence assessing the quality of answers created by a newer model.
It is a demand that Openai itself had indicated, as CNBC told a statement last month that it is reviewing the reports that Deepseek may be “inappropriately” that uses the output data from its models to develop the artificial intelligence model, a method referred to as “distillation”.
“Most of the fear of the market around (Deepseek) is in fact in place,” Hoffman told CNBC. “It still requires large models – distilled from large models.”
“I think the brief answer that everyone should take is: but the big models are still really important,” he added.
Victor Rebpily, CEO of Synthesia, CEO of Ai Video Platform, told CNBC that although Deepseek challenged, “The model is that scaling brute force is the only way to build better and better models”, the idea that companies will suddenly change important sums of work burden Amnesty International is misleading.
“I still think that when you look at the users of these technologies, and all the workflow tasks, I think that when we look back within three months, I think 0.01 % of these will be transferred to Deepseek from Openai and Anthropor,” Riparbelli said.
The Deepseek's development does not move the needle much for the industry because the market momentum is still widely supporting the largest artificial intelligence models. The signal institution is a non -profit organization that supports the explanation of the encrypted correspondence.
“This is not something that will disrupt the concentration of power or geopolitical balance at this stage,” Whitaker told CNBC. “I think we must watch the ball there and realize that this model is” the biggest is the best “that is not reduced through the gains of efficiency historically, and this leads this focus.”
